I’d wanted to ride one of those long American routes ever since the idea had occurred to me, about a fortnight earlier. Against his curmudgeonly nature, he agrees. Which is why we’re sitting on an Amtrak at Chicago Union Station. Our cabin has those blue bunk beds folded up, just like in the pictures. We leave the question of who gets the top bunk for later.

The train pulls out, and the hipster-infested industrial cityscape gives way to fields. A few hours in, I’m convinced it’s the same field on a loop. I’d assumed the scenery would be the point. I hadn’t actually checked. My brother gives me a look. It says, what on earth were you thinking?

The last time we’d spent this much time together in a small space, it was the bedroom we shared as kids. We were good at filling empty hours then. We’d make our own radio shows for Captain Scarlet on a tape recorder, then listen back, rolling around laughing.

My hopes of experiencing that again fade as the field loops in the background.

Looking to salvage the trip, I wander along the train. I find the pointlessly glass-roofed dining car. A few awkward conversations later, I realise my fellow travellers are mostly nervous flyers. They tell me about all the scenic routes I should try. I retreat to the cabin.

The geeks we are, we’ve both brought our grown-up toys: laptops. At some point, we realise we both share an emerging love of Clojure, a new language built on old ideas. One of us suggests we build something with it.

In the 90s, our culturally obligatory grunge band never got off the ground. Practice on our school stage descended into Muddy Waters-esque blues improv. This time, we pick a shared problem: data structure validation. Not glamorous, but appropriately “train journey to New Orleans” shaped. We sketch a few ideas, open our laptops, and start.

For what remains of the journey, the nervous flyers barely see us. We only surface at mealtimes to share their glass-roofed view of the fields, then head back to the cabin. There, we just hack. Emacs, a REPL, and a problem. That’s all we need.

Well, I also need my brother. He takes the laptop from me for the umpteenth time after I do the easy bit and get stuck. He takes over, I watch him work it out, and he hands it back. The tests turn from red to green. Same ritual as rewinding the tape recorder. Did what we made work? Less fall-over-funny this time. The cabin’s too small to roll around on the floor anyway.

At some point, we look up. The field has become swamp. Nothing like a deadline to focus the mind. We push to finish before the train beats us to it. Perhaps for the first time in history, a developer’s estimate turns out to be accurate. We commit the first complete version as the Amtrak rolls into Mississippi.

So that’s what we call it.

Then New Orleans arrives, and adult life resumes.

We eat incredible gumbo, drink a bit too much, stumble into marching bands and come home with beads. The project gets pushed to GitHub. A friend even designed a logo for it. It ends up running in production systems for years.

Something about that train ride made the through-line of what I was doing a bit more obvious. The time I’d spent with my brother growing up was the reason I was at that conference at all. And the desire to recreate what happened in that cabin with Clojure, in the teams and companies I’d go on to help build, is what’s driven me ever since.

Subscribe now

Our new building was tall and thin with six floors. Perfect, we thought, for a team on each.

We’d completed a management buyout of uSwitch from our parent company. The move was a chance to build a company we could all be proud of and shake up an industry. But a few months in, climbing the stairs told a different story.

It felt like visiting different companies.

One floor quiet, everyone heads down. Another with music blaring. Another with start-up energy. Every team ran its own socials and built its own rituals. The technology was equally fragmented: serverless to SOA, different infrastructure, different languages. The one shared ritual was afternoon snacks. Feeding time, before everyone went back to their own enclosures, oblivious to each other and to how their isolated optimisations were limiting our ambition.

Years later, the heterodox insight of a feminist philosopher finally made a diagnosis possible. Her thinking revealed the same self-limiting pattern in me.

What I Was Trying to Recreate

This all began at a start-up where the founder built a culture of real autonomy. You owned the problems, solved them, and lived with the results. It was energising. I felt privileged to be part of it and encouraged every good person I knew to join as well.

After an experimental first year, I moved to “Carl’s Team”. He ran the most successful part of the organisation. His relentless innovation challenged us. To keep up, my friend Paul crawled under desks after work, secretly installing Hadoop on people’s machines to process data (this was before big data was even a thing). The work we did during that period even influenced industry thinking.

Carl’s team’s success helped fund the acquisition of uSwitch, a UK price comparison website that had lost its way.

Paul and I left to figure out how to turn it around. What I didn’t expect was Carl’s response. He gave us huge amounts of time, advice, and support. When our early success generated more traffic than our systems could handle, he sent over his team to help. Many never went back. He just adapted to a leaner team and continued improving.

Carl was playing a different game. He focused on doing the right thing for the organisation. Helping wherever he was needed, he built connections everywhere with no expectation of payback.

It was an influential time, which I attributed to the environment. The autonomy, the ownership, the space to innovate. When I was offered the role of CTO at uSwitch, I took it with the goal of recreating and scaling that culture.

The Diagnosis

The insight came years later, on a podcast, after the symptoms had got worse.

When we acquired uSwitch, we quickly integrated two more businesses. Organising teams around product verticals (energy, mobiles, financial services) provided sustained ownership and autonomy to deliver. It replicated the model I’d seen work before. The fragmentation across those six floors was not what I’d anticipated.

Then we were acquired by a property company whose answer to that fragmentation was homogenisation. Their ambition was one place to “find, move and manage” your home, a goal that was proving difficult to deliver. In a budget meeting, I had to explain why developers from different parts of the business, with different skills and context, weren’t fungible. Centralised control was strangling the very thing it was trying to fix.

Neither approach could get big things done. That’s when I found Dr Anne Phillips.

Listening to her describe relational autonomy, I felt something click. Phillips argued that autonomy isn’t about stripping away relationships to find your isolated ‘true self,’ nor dissolving it into the group. We become more capable of self-determination through our connections with others, not despite them.

Self-determination is a collective project. That’s as true for societies and companies as it is for people.

It felt revelatory. I’d been focused on operating independently, on individual achievement, fighting its removal. The blind spot was the relationships that sustained it all. It was what Carl had modelled, and I had missed. The structural problems were a consequence. And Phillips helped me see that my behaviour was another.

The Evidence Had Been There for Years

The buyout that led to the move was funded by a private equity firm. One who took their due diligence seriously enough to bring in an occupational psychologist to assess the exec team.

I remember when it was my turn, entering a room with black walls covered with chalk drawings. A small table with three chairs, two behind it, where two psychologists sat ready to diagnose. A single chair at an angle in front, waiting for me.

It was intense. They asked questions on every subject. I talked nonstop about the team, what we were building, what I thought was wrong and how to fix it. I answered, oblivious to the situation and its consequences.

It was a naive strategy, but in retrospect, I’m glad I approached it that way. Reading it back, it’s a record of how much I was fighting everyone. My colleague’s four-word feedback read: “Mike is about Mike.” When someone suggested I soften my approach, my response was “Gentle is inefficient!”

The culmination of their assessment was that I didn’t understand my impact on others. “Team” for me was the technology organisation. I excluded the senior leadership team I was supposed to be part of. To be more effective, I needed to recognise that a CTO carries responsibility for the entire business. That meant looking beyond the technology organisation and helping my peers.

At that time, I wasn’t ready to hear it and failed to act. The technology team did great work, and that bought me time. But by the time I started to change years later, we were already a different company, and I had to change jobs.

Building Differently at loveholidays

When I joined loveholidays in 2019, I was determined not to repeat the patterns. Not structurally or personally.

A new CEO joined shortly after I did. He had an intuitive grasp of what I’d spent years failing to articulate. The result was a team structure that looked similar on paper. Long-lived teams, sustained ownership, space to go deep. That part was still right. Ownership is what makes deep innovation possible.

What was missing was connection.

Large parts of the platform had no commercial ownership. The restructure fixed that: every part of the value chain got a business owner. We bound together technical and commercial responsibility. Every team was connected to each other and to the commercial part of the organisation. It was the CEO’s vision as much as mine, and working it out with him was itself part of the change.

We even measured it, and the results followed. Our engagement scores climbed, and we were named a Sunday Times Best Place to Work. The survey showed 88% of people “felt empowered to make decisions”, and 92% reported “strong relationships”.

But the draw of self-sufficiency doesn’t stop. Especially as you scale.

New leaders hired in their own image, built teams as reflections of their own personalities rather than components of a larger system. One even described servant leadership as “adapting to every team’s way of communicating.” That sounds progressive, but it’s the old model dressed up. Relational autonomy is mutual. I help you do your job, you help me do mine.

Changing the Pattern

The organisational foundations were right, but they couldn’t scale without the relationships to support them. So I started building those — treating everyone as someone I had something to learn from, not just someone to convince. It wasn’t always easy. An early attempt to build a relationship through shared feedback was met with: ‘No. We will never work well together.’ But I didn’t give up.

COVID helped, oddly. Tough times brought the exec team together in ways normal working hadn’t. Slowly, the shift Phillips described started to happen. I stopped trying to do it alone, and the curiosity I’d lost at uSwitch returned.

It made sense. The more senior the role, the more of the system you are responsible for holding together. The bigger the ambition, the more the connections are what make it possible.

Seven years after joining loveholidays, the whole exec team presented together to investors. It was the CEO’s call to include everyone, even though investor presentations don’t usually work that way. He wanted to show what we’d built together.

Our CEO opened the way he always does, engaging and visionary. Our CFO closed it with feeling, walking us through the numbers with the authority of someone who’d been there almost from the start.

I sat there during the other sections, willing them on. Watching the CDO and CPO’s overview land. Loving the engaging story our CMO told of our brand evolution. I was proud of what we’d built.

I did my bit, conscious of all the people’s work I was representing, and not wanting to let the team down. When I sat down, the CMO leaned over and whispered: “That was f*****g awesome, mate.” My Slack was filled with messages from the rest of the team.

We’d built something none of us could have built alone.

What I’ve Learned

Feminist thinkers gave me the name for this, but the more I looked, the more it appeared everywhere. Developmental psychology, sociology, and philosophy. It’s even how people fix themselves in therapy. The same insight keeps surfacing wherever people study how humans actually develop. And yet self-sufficiency is the default.

Knowing about relational autonomy isn’t enough. It has to change how you act, and that doesn’t happen naturally. Old Mike still pops up from time to time, driven by the instinct to compete and ‘win’. In the organisation, the pull towards self-sufficiency never goes away.

My deeper reading of Phillips led me to Pauli Murray, a civil rights activist and feminist legal pioneer whose work shaped landmark equality cases in the US. She wrote:

“When my brothers try to draw a circle to exclude me, I shall draw a larger circle to include them.”

Murray didn’t write from a position of safety. She wrote as someone being actively excluded. The stakes she confronted were incomparably higher than anything in a business. But the human pattern is the same. The draw of self-sufficiency and isolation is everywhere.

You don’t have to be hostile for it to limit you. You just have to stop reaching.

The burrito was probably a mistake. The huge cookie definitely was. But here we are again.

In three hours, I’m supposed to be at BJJ. I can already see myself in the car park, engine running, having the same tired negotiation: “Do I really need to get my ass kicked by young lads for an hour and a half?”

I work out religiously. But I’ve used that as an excuse to eat whatever I want. “Abs are made in the kitchen” has always bugged me, and I’ve dabbled. Sober October. Blood tests.

All ways of avoiding the actual work of fixing my diet.

Two Scientists and a Podcaster

I’ve always been a sucker for the next health thing. I’ve done Zoe twice. I’ve been to “a vegan wellness retreat run by a rugby player who once killed a chicken with his bare hands”. (Yeah, that’s me in a hot tub in a national newspaper.)

Keto was one of these. I’d tried it twice before, cutting carbs low enough that your body switches from burning glucose to burning fat. Both times, I sabotaged it. Discovering white wine is low-carb turned the diet into an excuse to drink in the evening. Another time, my cravings led me to spend a Saturday buying ingredients and baking keto cupcakes. One bite and I spiralled straight back to carbs.

Then two episodes of Steven Bartlett’s Diary of a CEO kept landing in my YouTube feed. One with Dr Annette Bosworth, another with Dr Benjamin Bikman, both talking about insulin resistance and what it does to your body. They were describing exactly what I was experiencing: everything I’d been writing off as just getting older.

It felt like that parable of the drowning man who waves off two boats and a helicopter because he’s waiting for God. I didn’t want to be that guy.

I’d started to feel the benefits of keto before, but I’d never sustained it long enough for them to stick. That would take something more drastic.

No Escape Routes

It’s New Year’s Eve, and I’m watching ridiculously large fireworks over the harbour from my balcony, still basking in post-Christmas indulgence. I decided to give it another go. This time, I’d treat it like a work problem. Build a system. Measure inputs. Track outputs.

But more than that, I’d give myself no escape routes. Keto alone hadn’t worked because I’d found workarounds every time. So I added intermittent fasting because I can’t be trusted around a pain au chocolat, and cut alcohol to remove the evening temptation. I threw in daily journaling on a whim. That turned out to matter more than any of it.

I didn’t announce anything. Just started measuring on January 1st. (Then proceeded to mention it 30 seconds into any conversation to absolutely anyone.)

That first day, I skipped breakfast, ate a load of eggs for lunch, then had a steak with some creamed spinach.

41 days to go.

You Gotta Have a System

I struggle with restriction at the best of times. Tell me I can’t have something and I’ll find a way to get it. I’d proven that twice already. Honestly, Rage Against the Machine have a lot to answer for in my life.

So I had to stop relying on willpower and build a system that worked with how I actually behave.

I tracked weight, body fat, glucose, and ketone levels, as well as vital statistics. Had keto bars instead of fighting cravings. Black Americanos instead of flat whites to save the carbs for later. Replacements rather than abstinence, because one pizza and you’re out of ketosis.

The data unexpectedly kept me going. You lose a lot of water early with keto, so you see the impact immediately. When the numbers confirm it’s working, it reinforces behaviour. I used AI as a coach, too. Photographing restaurant menus, brainstorming meals from whatever was in the fridge. It took the thinking out of eating, which is what always beat me before.

The journalling I’d thrown in on a whim became part of the system, too. Not as a wellness exercise. As a way of noticing what was actually going on. The diet addressed my body, the data fed my rational brain, and the journaling surfaced what was underneath.

Less Puffy

I’m writing this six weeks in. The results: 8kg lighter, 5% body fat down to 11.6%. It’s the best I’ve felt in over a decade. Other people have noticed too, although “your face is less puffy” is a compliment I’m still figuring out how to take.

But the numbers only tell part of it.

Same balcony where I made the decision six weeks earlier.

I have focus and energy I’d forgotten was possible. One weekend, I drove to the beach, wrote two blog posts, ran, swam, did weights, and did yoga. Recovery was easier because I finally had sleep that actually felt like sleep.

I used AI to review my journal. It said, “Journaling didn’t change your life. It helped you notice that your life was changing.” I liked that.

One of the weird things that changed, which I put down to dopamine desensitisation, is that I now love the taste of water. Before it had the taste of nothing, a drink to tolerate whilst waiting for coffee. Now it tastes clean. I promise I’ve not joined a cult. It’s one of those small markers that something fundamental shifted.

Wednesday at 4pm looks different. No slump. But I’m still negotiating in the car park… turns out I still don’t want to get my ass kicked, but that’s just an ego problem now. Will journal about that later.

Way Too Many Eggs

Six weeks of this intensity can’t go on forever, though.

Obsessive tracking takes time and mental energy. And I was SO sick of what I was eating. At some point, I stopped looking forward to food. Eating was more of a chore than an enjoyable experience. I also just wanted a bit of fruit, a bowl of pasta, not more meat and cheese.

The food wasn’t the toughest bit.

Without alcohol, I had to sit with feelings I’d been quietly numbing. Evenings felt longer. Social situations had an edge I hadn’t noticed before. The journaling made this impossible to ignore. Writing down how you feel every day makes the patterns obvious. I could see how often I’d been reaching for a drink, not because I wanted one, but because I didn’t want to feel something.

A friend once told me, “Drinking and food aren’t the problem, they’re the solution.” I didn’t understand that properly until I saw it in my own journal, week after week.

Leaning into it was harder than any amount of eggs. But it was satisfying too, in a way I didn’t expect.

The real cost of these six weeks isn’t the food boredom or the social backlog. It’s that I can’t pretend anymore. I know what the old habits were actually doing. I know what living without them feels like. I can’t unknow that.

I’m anxious about losing this feeling. But I also want to enjoy food and a drink again.

Habits Compound Both Ways

My CEO keeps telling me, “Things just start going wrong after 40.” I’m not convinced. Maybe that’s what happens when small bad habits compound, and we decide it’s inevitable.

The six-week version was the experiment, not the destination.

So I’m tapering off the measurement but keeping the journaling. Reintroducing things one at a time and paying attention to what happens. Being intentional rather than just drifting back to old patterns. The system changes, but the mindset doesn’t.

If you’ve found a way to keep the gains without the intensity, I’d genuinely love to hear how you did it. I’m still figuring that part out.

But first, I’m going to order a pizza and have a beer.

Almost as soon as I went digital with my notes, I started changing my mind about how to organise them, creating a mess it would have taken days to untangle… so I didn’t bother.

I was inspired by reading How to Take Smart Notes and discovering how the German sociologist Niklas Luhmann obsessively maintained 90,000 index cards, enabling him to publish 70 books and nearly 400 articles. I wanted that system, but with modern tools. Instead, I got distracted by Tiago Forte’s PARA structure, then the inevitable YouTube rabbit hole.

The result was predictable.

Whilst polishing my process, I ended up with over 80 notes stuck in the inbox, duplicates I’d forgotten about, links pointing to notes I’d never created, all tagged inconsistently. By the time I knew what I wanted (Maps of Content, a proper tagging taxonomy, notes at different maturity stages), fixing 200 notes felt like an enormous chore.

So last week I got Claude to do it.

The Audit

A friend came up with the term “thought refactoring” to describe how I was using AI to rework documents. Co-work, Claude’s desktop tool, helped me expand that idea from a single document to an entire folder of notes. What made it possible was Obsidian storing everything as plain Markdown files rather than locking it in some proprietary cloud format.

I made a backup, pointed Co-work at my Obsidian root folder, and used this prompt:

I’d like you to review my Obsidian vault.

I’ve tried to mix a Zettelkasten-style approach with atomic notes. 
That is one idea per note, connected through links, not nested folders. 
Maps of Content as index notes instead of a folder hierarchy.

I want a tagging system with two dimensions: 
- a type tag (concept, framework, principle, quote, book-note, or person) 
- a maturity tag (seedling, developing, or evergreen)

Can you:
- audit the vault structure
- identify what’s working and what isn’t
- put together an action plan

Don’t move anything until I’ve approved the plan.

It grasped the intent quickly. Not just Zettelkasten, but the added complexity of MOCs and a two-dimensional tagging system on top.

Minutes later, Claude described the extent of my personal knowledge management shame. On top of the notes stuck in the inbox, it explained that 114 were unfinished and around 30 were essentially empty. Oh, and I’d misspelt filenames, had orphan notes with no connections, and created duplicates months apart without realising. And yes, my Resources folder conflicted with Zettelkasten principles.

But it had a six-phase plan to fix it, which I checked before it started moving anything. Because I still have some use, right?

Then it executed. Every note moved, renamed, tagged, and wired into the right MOC. It extracted atomic concepts from my notes, so now autopsia, disequilibria as the motive for learning, and shamanic rituals all have their own home.

Claude had flagged my plugins, too. I got recommendations for Dataview and Homepage. All without a single YouTuber telling me not to focus on the process, whilst spending twenty minutes describing their process.

The anxiety of adding to broken notes and creating more problems for future Mike to fix lifted, and I started to think about what else was possible.

The Skill

The cleanup fixed my past mistakes. I also needed help maintaining my notes so they didn’t drift back. So I built a custom knowledge-linker skill, a set of instructions that tells Claude exactly how my vault works, what conventions to follow, and how to create notes that fit.

You can see the skill here → knowledge-linker.skill

Now, when I come across a new idea, I can say “add Jevons paradox to my vault”. The result is a properly formatted, tagged note linked to related notes and filed into the right Map of Content.

I tried it after finishing a book last week. Instead of writing half-finished notes and telling myself I’d process them later, I talked through what I’d learned with Claude. In under 20 minutes, we extracted 4 new ideas into atomic notes and filed them based on what interested me.

Talking it through also helped with recall, though it did make me think about what I’m losing by not making the connections myself. Curating notes manually had forced me to think deliberately about the connections. So I’ve kept some agency over that, and I review everything after.

I also built a second capability into the skill: it reads a draft and searches the vault for any connections that might be useful. My hope is that using them in writing is what makes them stick, rather than just curating them.

The skill is there to handle the admin, so I can focus on applying the ideas.

In Defence of Not Doing It Yourself

The emergent connections I’d imagined, the Luhmann dream of ideas linking across years of reading, never appeared. I couldn’t replicate his obsession even with a tool like Obsidian. My result was a pile of unfinished notes and a growing dread about a weekend I’d eventually spend sorting it out, or quietly give up on.

But that weekend never came. Claude sorted it in an afternoon, knew exactly what I was trying to build, and gave me the tools to keep it that way.

Luhmann’s obsession was the connections, each card placed deliberately, the system maintained by hand for forty years. I’ve automated that part. What’s left is the bit he’d have actually envied: the time to read more, think more, and find ways to apply it.

“Some people only exist to help companies fail slowly.”

My colleague Carl developed this theory at Forward with founder Neil Hutchinson. Forward was the 2000s startup culture distilled. People who’d built companies from nothing, who understood what creation cost. Neil had the rare instinct to see through the facade and back people on potential, not polish.

The opposite infuriated Carl, sitting across Forward’s portfolio, watching struggling companies hire for polish over potential.

Businesses would default to hiring senior, impressive, articulate leaders who looked the part. They’d pass over the people who’d driven change. The result was always the same: frustrating, gradual, managed decline when these polished leaders were installed. The opposite when teams had the experience and desire to lean into change.

Carl’s insight stuck with me, but I didn’t have a framework for it until I heard the proverb: ‘Shirtsleeves to shirtsleeves in three generations.’

The Three-Generation Problem

The proverb describes a cycle. The first generation builds wealth through grit and sacrifice. The second generation manages and maintains it. The third generation squanders it as they never experienced the cost of creation.

Studies suggest 70% of wealthy families lose their wealth by the second generation, 90% by the third. The exact numbers are contested but the pattern is universal. Every culture has its version of the saying. In the north of England, where I’m from, it’s the appropriately archaic “clogs to clogs.”

The same pattern plays out in business. Not with wealth, but with the ability to change.

First-generation people take real risks. They enter chaos willingly: startups, transformations, building from nothing. They’re comfortable with ambiguity, full of energy, and biased towards action.

Second-generation people witnessed the sacrifice firsthand. They understand what it takes to make something work. They may not have started the system, but they understand its fragility. They can evolve and scale what exists without breaking it.

Third-generation people inherit systems, structures, and status when things are already successful. They’ve never experienced the cost of creation or change. They’re polished and confident, exceptional at operating within systems. But they can’t change systems.

This is what Carl saw. Companies hire third-generation people because they look the part: senior, impressive, articulate. But when markets shift, these people can’t adapt. You start seeing the signs everywhere.

What Third-Generation Leadership Looks Like

Sit in one of our business reviews and you’ll hear our CEO complain about teams’ “navel gazing.” It’s third-generation thinking in action: staring at internal metrics whilst markets shift around you. A pattern that appears in every role.

At a previous company, rapid growth was crushing us with complexity. We needed to sell data, build a B2B2C business, and handle more complexity with the same systems. I promoted one of the most brilliant engineers to lead it. When I came back weeks later, he’d rewritten the same systems in a new language. Faster, cleaner code. Same architecture. Same business model mismatch.

It’s substitution: replacing a complex question with an easier one you can answer. Something product and project managers aren’t immune to. They obsess over the latest fad or framework (don’t get me started on SAFe) instead of understanding the systems well enough to actually solve the business problem. “Where can we take this business?” becomes “The framework will tell us what to do.”

But other times it’s just pure avoidance. Data had changed what was possible and how we could work with partners. Commercial teams preferred boozy afternoon lunches and the same conversations they’d had for years. The world around them had shifted. They weren’t solving the wrong problem. They were ignoring the hard one entirely.

They maintain the status quo by polishing process and tactics, whilst strategy crumbles. It’s a slow death.

Why Choose the Hard Path

Third-generation thinking kills businesses slowly.

Marginal gains aren’t wrong. They’re vital. But they have a ceiling. You can optimise a process by 10%, maybe 20%. You can’t optimise your way to a fundamentally different business.

For loveholidays, the step change was reactivity. We shifted from a single product to a global platform, giving commercial teams the tools to make changes directly rather than waiting for developer deployments. Optimising our deploy process would have improved things incrementally. Instead, commercial teams now make changes at 2x the rate we can. With AI, that’s jumped to 1.6 million changes a year.

Then came scale. The tech foundations let us expand into more countries: 20+ trillion packages processed to meet the demand. Growth of 186% in a single year on a business that was already massive. And we did it with just a 5% cost increase. None of that is incremental. Each one is a step change.

Our competition optimised. They refined what they had, improved their existing processes, squeezed more from the same model. We grew to 3x their size and are on track to become one of Europe’s biggest travel brands.

But the real shift goes beyond scale. We broke free of the category altogether. We’re not a better travel company. We’re a technology platform business. Third-generation thinkers optimise within the existing frame. First-generation thinkers change the frame entirely.

What You Get From Walking the Path

This kind of transformation doesn’t just change the business. It changes you. That’s why my mentor drilled into me that it doesn’t count unless you’ve scaled two businesses.

The first time I was responsible for transformation, it was bumpy. Even though I’d witnessed change at consultancies and startups, being responsible for leading it is different entirely.

The second time, at loveholidays, was “twice as good, twice as fast”. Prior first-hand responsibility helped me understand what was needed from the start. Including surrounding myself with people who’d done it before.

Five years later, when AI emerged, we were ready. We’d already experienced its impact, but there was no playbook for implementation. We built our own framework: structured enough to direct effort without limiting experimentation, whilst measuring cost, utilisation, and impact.

The progression: experiencing change, leading change, adapting to change. The meta-skill is a playbook for creating playbooks. It gives you the willingness to tear apart what you’ve built when something better comes along, and the ability to navigate without a map.

Walk enough paths and this becomes second nature. You develop the confidence to move before the path exists.

Third-generation people lack this entirely. They started at the destination. They never walked the path. They can follow a playbook, but they can’t create one. When the playbook doesn’t exist yet, they freeze.

What the Pattern Taught Me

The pattern gave me a name for something I’d already been living, and a lens to sharpen it.

I learned to seek out change, not comfort. Early in my career, the tempting roles were the ones with status and stability. The ones that taught me the most were chaotic, uncertain, and often thankless. You learn to build through building. Arriving at the destination gives you something, but not what you need to grow.

I learned to hire differently. Neil at Forward backed me on potential, not polish, giving me my first CTO role. That shaped how I build teams now. First and second-generation thinkers need more support, more mentoring, more backing. But their impact compounds. If you only hire people who look the part, you’ll build a team that can only look the part.

I also learned not to dismiss what third-generation people do well. If you can build and present with their polish, you become far more potent. That combination gives you access to bigger problems, broader scope, and greater impact. Substance without polish limits your reach.

Carl’s insight was brutal because it’s everywhere. Once you see the pattern, you can’t unsee it. You start recognising which leaders will adapt and which will optimise themselves into irrelevance.

The pace of change isn’t slowing down. AI is accelerating it. And it’s the clearest test of the pattern yet. Most of our people apply AI to existing processes. A few explore new possibilities that change our business.

Third-generation thinking was always terminal. Now it’s just faster.

At the consultancy, we held a conch to speak in stand-ups. We rang a bell when we completed a story. We put a pound in a jar if we touched the mouse instead of using keyboard shortcuts. We wrote postcards from our future selves.

I participated in these experiments, but over time they grated. I was just there to write code, and all this ceremony felt like it was getting in the way of what I’d actually been hired to do.

When I left, I felt relief, finally free from people who’d drunk the Kool-Aid. In one of my least fine moments, I even mocked it publicly on Twitter, using a hashtag to call out what I saw as #AgileBS.

I thought I’d escaped something pointless, ready to focus on what mattered. Getting sh*t done.

In the years since, our success has depended on the practices I’d dismissed.

More than once.

And now, they are more important than ever.

The Confident Exit

After leaving the consultancy, I helped build a new organisation. We were lean, fast, and proudly “post-Agile.” We even printed flyers with a Fender Twin Amp and a Commodore 64 leaning against it. A not-so-subtle nod to the rockstar developer mythology.

Yeah, I was that guy… but for a working-class kid from the north of England, this was as close as I’d get to being Kurt Cobain or Eddie Vedder. I’d given up on drums, so a rockstar developer would have to do.

Post-Agile meant stripping away process and structure. Small teams hacking on code, using everything we’d learned to ship fast, adopting any tool that didn’t slow us down. It was awesome, at least for a while. We shipped quickly, decided faster, and achieved success that almost vindicated leaving the ceremony behind. Look at us, thriving without all those awkward rituals.

What we didn’t see was that we weren’t post-Agile at all. We were products of Agile. The practices we’d dismissed had already shaped how we thought, communicated, and worked, giving us instincts we didn’t even realise we’d developed whilst standing awkwardly with that conch in our hands.

We’d thrown away the ladder after we’d climbed it.

The Humbling

Over the next few years, the business scaled quickly. We acquired a much larger organisation. Suddenly, we had parallel teams, complex dependencies, and people who hadn’t lived through the same formative experiences.

One leader we worked with named it: “developer anarchy.” One of his friends, some industry luminary, visited us, trying to figure out how we were shipping without unit tests. (We just did functional tests. Faster that way.) We took it as validation. Look at us, breaking all the rules and making it work.

Except it wasn’t working. Not anymore. The edges were straining, it wasn’t scaling, and a small group of us were doing way too much to hold everything together. Personally, I was working until 2am every night, rewriting what teams had done, trying to hold it together through sheer force of will.

Two of my colleagues sat me down at a bank of desks in Camden, right next to this huge keyhole door inspired by Alice in Wonderland. Surrounded by all that opulent confirmation bias (our cool office, our rockstar mythology), they told me the truth.

“Mike, this isn’t working. We need to start doing standard Agile stuff again.”

They were right. The ride was over. I couldn’t keep rewriting everything each night.

The practices I’d mocked were actually artefacts of an environment that created the conditions for learning, for better instincts to form. Once I’d learned the lessons, I’d mistaken the outcome for proof that the process was unnecessary.

So we went back to basics. Implemented all the XP and Agile practices: stand-ups, structured planning, showcases, pairing, and TDD. Just in time.

The Invisible Classroom

I didn’t understand what that environment had done for me until I became responsible for creating one.

The practices weren’t about the rituals themselves.

Those early years at the consultancy, hours sitting next to the smartest people I’d ever worked with, people who’d literally written the books on XP and Agile and the frameworks we used, that was where I actually learned my craft. Not just from writing code, but from watching how they worked, seeing their habits first-hand, absorbing the thousand small decisions that separate effective developers from merely productive ones. I learned to master the tools we used because if you didn’t, you had to put a pound in the jar (which is why it still pisses me off when people don’t use shortcuts, even in Google Docs).

They were exploring better ways to work and learn together: to pair programme, to articulate your thinking, to defend your position, to see how someone else approached a problem, to get immediate feedback when you did something inefficiently.

I thought the only value was in the code we shipped. I missed the value of learning from people who knew what good looked like, especially when you’re early in your career and don’t yet know what you don’t know.

The Pattern Repeats

Lately, I keep seeing the “post-Agile” attitude surface on social media, often from developers who’ve never experienced what those practices were actually trying to solve. Worse, remote work and AI mean people can look productive without ever being in the room. No proximity, no real-time feedback, no having to hold the conch and explain your thinking to someone who’ll spot the flaw in your logic before you waste a day implementing it.

Those practices (TDD, pairing, continuous integration, fast feedback loops) weren’t just about producing code. They were about building instincts, learning to see what’s wrong before it becomes a problem, developing the muscle memory that makes you effective over a career, not just in the moment.

You can’t shortcut the acquisition of these skills, especially early in your career when you’re still forming the habits that will define how you work for decades. It requires hours of making mistakes and getting corrected in real time by someone who knows better, someone who can show you not just what to fix but how to think about preventing the problem next time. But now people can produce code without ever seeing how someone experienced thinks through a problem, without anyone sitting next to them, correcting bad habits as they form.

There’s a Japanese martial arts concept that describes this progression. Shu Ha Ri. (Yeah, I learned about that there, too.) Shu is following the rules. Ha is breaking from the form once you’ve mastered it. Ri is transcendence: moving freely because the principles are internalised.

AI and remote work let people jump straight to what looks like Ri, bypassing the years of Shu where you build the foundational understanding that makes breaking the rules productive rather than chaotic. But without mastering the fundamentals first, you’re just mimicking transcendence. You look productive, but you haven’t built the instincts that make you effective.

The Message I’d Send Back

A few years back, a former colleague messaged me on LinkedIn. They’d reached a similar level of seniority and wanted to thank me for the environment I’d helped create. They finally understood how hard it was to do something like that.

That message stuck with me. Not just because it was kind, but because it disturbed me. It reflected my own failure to complete the same journey. I’d understood the impact they’d had on me, but I’d never acknowledged it.

So my penance is writing the postcard my 20-something self needed. The one reminding him to embrace Shu, not skip it. Wax on, wax off.

Recognise the privilege of learning from people ahead of you. Show some humility and let them shape you. It matters more than you think, particularly when you’re still figuring out what good looks like. You’re surrounded by them now but these opportunities get rarer, and you’ll wish you’d made the most of them.

Value being in the room. You learn faster when you can ask why. The learning happens at the whiteboard, in the pauses, in the way someone structures their thinking. In the immediate correction when you’re about to make a mistake. Absorb as much as you can when you’re together so you’re effective when you’re on your own.

Try new things. Lean into the awkward. You’ll get more from those weird practices than you think. Some will land, some won’t, but you’ll take something from all of them. Stay critical, just not cynical. Just keep off the Kool-Aid or you’ll end up in a toe-curling Agile music video.

It took me 20 years to see this clearly. Even now, I know I’m missing what’s shaping me.

In another 20 years, I suspect I’ll wish I’d understood we needed to act now. To deliberately create environments where these lessons can be learned, even as the world retreats from presence and embraces AI assistance.

That’s the next postcard I’ll be writing, just not on the back of a Fender Twin Amp.

We spent six years wrestling startup chaos into scalable systems. AI has brought the chaos back.

Every few weeks, a new tool is released that rewrites yesterday’s rules (or at least reworks images on X). Markets swing between “AI bubble bursting” headlines, record ad revenues, and panic about disintermediation at demos. Even experts are puzzled by the gap between AI’s obvious power and its measurable economic impact.

It’s chaotic, and I love it, because this is where new things emerge. But an age-old pattern is repeating: you need balance between order and chaos to navigate it.

End of the Age of Scaling

For six years at loveholidays, we followed well-trodden paths to create order from startup chaos and help the business scale. Building a single scalable platform, against a standard set of principles, with the people and structure that let us move fast.

AI has flipped us back to a chaotic state where the playbooks that scaled us no longer apply.

Just as our ordered approach to scaling needed a little chaos to allow new ideas to emerge, this chaotic environment needs order to identify new capabilities and let them take hold.

The yin-yang in the Taijitu captures this perfectly. Each phase contains the seed of the other. The line connecting them represents the balance that prevents stagnating in pure order or fragmenting in pure chaos.

The platform we spent years shaping is foundational to finding that balance. It’s fast, flexible, focused on our business, with the observability to understand impact. Precisely the environment to navigate between order and chaos.

The vision we created for our platform also plays a vital role. We declared our intention to “Build the general intelligence for travel.” An AI that understands travel the way a great agent does, connecting customers to supply, and improving with every interaction. That vision serves as our city on the hill, a clear north star for navigating chaos.

Our platform and vision created the structure. That structure revealed what we actually needed: a new way of thinking.

The Skill AI Rewards

Systems thinking separates those who harness AI from those who treat it as a better way to generate emails or answer questions. AI delivers the biggest benefit to those who see how humans, machines, and processes fit together.

Here’s what that looks like:

We operate at scale, serving hundreds of thousands of hotels across multiple markets and languages. We’re heading toward 240 customer cohorts, each looking at the same hotel with different needs. German families scan for kids’ clubs. Whilst Dutch couples hunt adults-only retreats.

The obvious move is to use AI to write descriptions faster, but that’s the wrong way to think about it. One description can’t serve eight audiences, and creating personalised variants manually is impossible at this scale.

Instead, we thought more systemically and created orchestration with AI embedded.

The system ingests content (reviews, amenities, location data) and applies our brand guidelines. It generates highlights based on who’s searching. Families see kids’ clubs. Couples see the quiet poolside bar.

We reviewed every output at first to establish quality standards. Then we built an AI scorer that validates its own work against those standards. Self-correcting quality control.

As I heard someone say recently: “Progress is built on all the things you don’t have to do.”

Measuring Through It

To separate signal from noise, we track what matters: how AI changes behaviour, speeds delivery, and where it falls short.

We’re spearheading this approach in engineering, where seven years of platform data lets us see the impact of changes over time.

50% of our commits are now AI-assisted. But the skill isn’t prompting; it’s decomposing problems so AI and humans complement each other. We’re still writing code, but the work shifted from typing to directing: breaking down problems, guiding AI, judging outputs, refining approaches.

As AI commits climb, we track correlations with our core engineering metrics: deployment frequency, change failure rate, and MTTR. These metrics predict business performance rather than just measuring activity.

We’ve built a dedicated dashboard to monitor engineering impact, drawing on years of collected data. These foundations enable us to conduct meaningful research.

We’ve been using CodeScene (inspired by Adam Tornhill’s “Your Code as a Crime Scene”) to measure code quality for years. Their hotspot and code health metrics gave us visibility into our codebase. Integrated with all our other metrics, it allowed us to measure the impact of AI-assisted development.

In a recent interview, Stuart Caborn, our Distinguished Engineer, covered how we found that teams using AI assistance maintained code quality while dramatically increasing output.

These metrics show what actually works. Proof, not opinions. Without them, individual experiments stay isolated. With them, learning compounds across teams.

Which makes it particularly embarrassing that I spent a year confidently citing completely made-up numbers.

My Lesson

For a year, I told people we process “six quadrillion packages: the same number as all the grains of sand on every beach.”

Claude 3.5 gave me that line and I fell in love with it. Six quadrillion packages, the same as every grain of sand on every beach. Such a cool visual that I repeated it everywhere: investors, conferences, social media, and boards.

Then the upgraded Claude 4 corrected it. It was actually one beach, not all of them. To be fair, that one beach is the biggest in England (seven miles of sand at 30cm depth). But I was orders of magnitude wrong.

Whoops.

It’s great that I’m the example of how even people building AI-native capabilities get burned. The tools are powerful and fallible. This is why systems thinking matters: it’s not about trusting AI or distrusting it, but building verification into your process from the start.

The Story So Far

The three streams we implemented to navigate our response to AI are taking shape.

Systems thinking matters. Seeing how humans, machines, and processes fit together. But thinking alone isn’t enough. You need measurement to prove it works, and verification built into your process from the start.

What’s emerging is less about any single technique and more about seeing patterns. The content personalisation challenge looks different from the engineering measurement challenge. Still, the underlying structure is similar: orchestrate the system, measure the impact, verify the results.

We’re still early in understanding this. But the skill isn’t mastering individual AI tools. It’s learning to recognise how solutions transfer between contexts.

More on that soon.

For now, we’re sharing what we learn. We’d love to hear what patterns you’re seeing.

Those diagrams become something more than meeting notes. They’re railway tracks that guide how others think about the system long after you’ve left the room.

I discovered this because I can’t think any other way. I have aphantasia — no pictures in my head. Every system must be externalised to work with it. And that’s how I learned drawing isn’t for sharing what you know. It’s how you figure out how things actually connect.

Why Drawing Is Thinking

The Gap Between Knowing And Understanding

Writing forces clarity. So does drawing. You don’t realise you don’t understand a system until you try to draw it: boxes on paper, arrows connecting them, relationships labelled.

There’s a well-replicated effect: argue the opposite of a belief and your conviction drops. Not because the other side wins, but because you expose gaps in your own model. Drawing does the same to system beliefs.

My mentor asked me to draw the uSwitch business model. I had abstractions in my head (relationships, flows, dependencies) but tried to translate them into something concrete. I drew literally. Customers on one side, suppliers on the other, our platform connecting them, regulations surrounding everything. All the components.

He looked at my drawing. “Draw what’s actually in your head. Draw the abstractions.”

Customer needs. Supplier relationships. Regulation. Platform capabilities. 4 core elements that applied across all our verticals: energy, broadband, insurance, mortgages. Not the concrete components, but the pattern underneath.

His abstraction revealed the next level. Not just how one marketplace works, but how multiple marketplaces working together create pull and tension across the whole system.

That moment gave me permission. I didn’t need to translate abstractions into concrete things. I could draw what was actually in my head: the patterns, the relationships, the structures.

Collaborative Drawing Is Collaborative Thinking

Draw alone, work out your understanding. Draw with others, work out shared understanding.

People nod along to words for hours, each building different mental models. Nobody knows they disagree until decisions reveal disconnects.

Start sketching. Disagreements surface immediately. Someone points: “That’s not how it works.” Another adds a box: “You’re missing this.” The diagram forces reconciliation.

The philosopher Isaiah Berlin described 2 types of thinkers: hedgehogs who know one thing deeply, and foxes who connect across domains. Both exist in every organisation.

Hedgehogs see the system beyond their domain. Foxes ground abstraction in concrete reality. Words let hedgehogs and foxes talk past each other. Diagrams make them meet.

That uSwitch abstraction (customer needs, supplier relationships, regulation, platform capabilities) balances 4 interdependent dimensions in parallel. Most people can’t hold all 4 simultaneously. They think sequentially, losing connections.

Diagrams make parallel thinking possible. All 4 dimensions visible. Connections explicit. Trade-offs clear. Collaborative thinking. Working out how the system actually works together because you can’t hold it in your heads alone.

Railway Tracks Persist

The drawings outlive the meeting.

Months and years later, I see my sketches (or an evolved version) in presentations. The ideas scribbled on a page travel on their own. People draw the same picture back when explaining how the system works to others.

My mentor calls that the magic moment: when people can externalise the model themselves, alignment has happened. Not whether they can explain in words, but whether they’ve internalised the system enough to draw it independently.

The thinking happened collaboratively through drawing. The diagrams persist as infrastructure.

How It Evolved

Solo Drawing: Working Out My Own Understanding

Those early sketches taught me: abstractions on paper reveal what you understand versus what you think you understand.

I started using this deliberately. 2-sided marketplace diagrams. Platform architecture. Team structures. Roadmaps. Every complex system, I drew first. Not to communicate it, but to figure out how it actually worked.

Those early sketches taught me: abstractions on paper reveal what you understand versus what you think you understand.

I started using this deliberately. 2-sided marketplace diagrams. Platform architecture. Team structures. Roadmaps. Every complex system, I drew first. Not to communicate it, but to figure out how it actually worked.

Collaborative Drawing: Working Out Shared Understanding

I used this approach at my previous job and brought it to loveholidays. Not just technical systems but everything.

When I took responsibility for our contact centre, it lacked clarity. Nobody could articulate how teams connected, where responsibilities overlapped, what happened when customer issues moved through the system.

Drawing forced decisions. Where do handoffs happen? Who owns what? How does information flow?

The diagram didn’t communicate existing structure. It created structure. We didn’t know how the system worked until we drew it together.

This works beyond tech. Complex systems defeat conversation alone.

Scrumpled Paper Moments

Session ends. Paper covered with boxes, arrows, annotations. Spaghetti. I’m about to bin it.

“Can you send me that?”

Those connections are in my head now, but others need them. Maybe seeing these concepts externalised for the first time.

The diagram captured the collaborative thinking process. Now it guides how others think about the same system.

Systematising: From Co-Creation to Infrastructure

Sketches on paper only scale so far. Co-creation in a room works if you think in systems and you’re in the room working it out together.

But rough sketches fail for people who don’t naturally think systemically. And they fail for anyone who wasn’t in the room during co-creation.

Recently I tried to understand the complexity of scaling into 160 markets. Each has different regulations, currencies, customer expectations, partner relationships. I created my usual rough sketch, working out how this scaling actually worked, then co-creating with the team. We aligned. We understood the structure together.

When I shared it wider, people didn’t quite get it. Some couldn’t follow the rough sketch because they don’t think in systems. Others weren’t in the room when we’d worked it out, so they lacked the context.

Jazz, our Design Manager, took it and transformed it. Same structure, systematically designed so anyone could understand the complexity immediately. Regardless of how they think or whether they were there.

People saw Jazz’s version and said, “Brilliant.” It went from working in the room to working across the organisation. From co-created understanding to organisational infrastructure.

It’s like Picasso’s bull studies: each iteration reduces to essential lines until the final version captures the essence in a few strokes. The studies work out the abstraction. The artwork makes it instantly legible.

Jazz didn’t make it prettier. She made it portable.

Rough sketches work out system understanding in the room, with people who think systemically. Systematic design turns that understanding into infrastructure that works for everyone, everywhere. Both serve thinking at different stages.

Why This Matters Now

Systems are getting more complex, and the way we work with them is changing.

A few months ago I sat with one of our content managers. She was talking about how we scaled hotel content with AI. 3 people kept 2,000 hotels updated each year but this changed to 50,000+ in 3 languages.

She’d been thinking in systems but she had to encode them into prompts and words.

You could no longer see the system: its steps, decision points, exceptions, branches, feedback loops.

A few months later this pattern is being replicated on a number of different teams. However, we’re using n8n at loveholidays to design and visualise these workflows. The tool lets you work out how systems should function by drawing the structure, seeing where decisions branch, where loops feed back. Then those visual workflows drive the AI execution.

The progression: from seeing systems and encoding into prompts to designing workflows visually that make the system structure explicit.

AI handles execution. Someone still designs the system.

As AI advances, visual thinking doesn’t get less important. It becomes essential. Systems are more complex. Drawing is how you work out that complexity, alone and together.

What You Can Use

You don’t need aphantasia.

Be the person who draws. Lean into the awkwardness even if you don’t know where to start. You don’t know how a system works until you externalise it.

Next time people describe a system in a meeting, notice if they’re working out how it actually works or talking past each other. Grab a whiteboard.

Don’t worry about pretty. Boxes and arrows work. Roughness signals “we’re working this out together.” People feel comfortable adding, questioning, changing.

When people ask “can you send that?”, do it. Those diagrams become railway tracks, infrastructure for how your organisation understands systems.

When they draw it back. That’s when you’re aligned.

Need to scale that understanding beyond the room, beyond systems thinkers? Invest in systematic design. Get an artist like Jazz to help. Rough for working it out. Designed for making it portable.

The goal isn’t beautiful diagrams. It’s thinking infrastructure. Rail tracks that guide how your organisation works out system complexity together.

The Advantage

I can’t picture things in my head, so everyone sees how I understand systems. Not finished thoughts, but the process of working them out.

No private knowledge. No hidden images. Just collaborative understanding that becomes shared infrastructure.

In a world where AI handles execution and competitive advantage shifts to systems thinking, the inability to think privately isn’t a limitation.

It’s a feature.

It’s exactly what’s needed.

We’ve built a platform capable of combining more than 15 trillion holiday possibilities each day — that’s 5.4 quadrillion offers a year. Enough grains of sand to cover the entire 7-mile stretch of Poole Bay in 30 centimetres of sand.

We don’t just create that beach — we also help each customer find their one perfect grain of sand. Millions of times a year.

It’s an extraordinary system — fast, flexible, and proven at scale.

And in true loveholidays style, it’s no longer enough.

The Inflection Point: Why Now?

We’re at a pivotal moment. 3 forces are colliding, each demanding fundamental change.

1. Our ambition has outgrown our capabilities.

Everything we originally built focused on one corner of the holiday market — short-haul holidays for UK customers. Our plans require a space 160 times larger. More destinations, more markets, more product types, more customer segments. Realistically, this expansion is something our tech, and our org, can’t support in its current shape.

2. AI is rewiring how businesses work.

Historically, the first wave of technology revolutions brought efficiency gains — we’re already seeing this. The second wave transforms entire industries. With AI, some believe it might eliminate whole categories — possibly including ours. When your industry faces potential disintermediation, evolution isn’t optional.

3. We’ve committed to a new customer vision: “Treat every holiday like it’s our own”.

This isn’t marketing. It’s an intentional shift in how we operate—from processing bookings to genuinely caring about each customer’s experience. Achieving this level of personal service whilst scaling is essential to our growth. We want customers choosing us not just for deals, but for how we treat them.

The challenge is huge, simultaneously addressing scale, AI disruption, and delivering an exceptional customer experience.

The question then becomes “How?”

How can we triple sales without tripling costs? Enter new markets and launch products faster? Serve customers better without breaking the bank?

The answer remains what it’s always been: we evolve.

Evolution has always been our answer to impossible challenges.

Evolution as a Survival Strategy

12 years ago, we started as travelmatch.com, a holiday comparison site. Packaging our own holidays was our first evolution — from aggregator to online travel agent (OTA). That pivot, and many more, turned us into the UK’s fastest-growing business.

Then scaling became the challenge. Our incoming CEO crystallised what we needed: to stop thinking like a travel company using technology and become a technology company in travel. This shift created foundations for truly unique capabilities in our industry. We built the fastest travel experience capable of offering a much wider range of holidays, accelerating us to the UK’s third-largest travel agent and the market leader in Ireland.

Now we need to change again. Being a technology company isn’t enough. We must embrace AI if we want to meet these 3 challenges.

This transformation is more ambitious than anything we’ve attempted. Technology transformations have playbooks — I’ve completed several and written up my learnings. But nobody has cracked this for AI. We’re writing the playbook whilst building it.

Our Vision: The General Intelligence for Travel

We’re building intelligence that works with language models, handles package travel’s complexities better than any system, and operates at an otherwise-impossible scale.

Breaking down those 3 words:

General means comprehensive — a unified system serving every team across our business, available to everyone.

Intelligence means AI integrated into every workflow, not bolted on — at our fingertips through an integrated platform.

Travel means deeply specialised — embedding 6 years of domain expertise that others can’t replicate. Our unique knowledge of package travel.

We’re creating travel intelligence that enhances rather than replaces human capability.

Why Only We Can Build This

AI is powerful, but only when it has context and can act. That’s where we excel.

We’ve always created new value from existing assets — selling holidays made from others’ hotels and flights, building our site on external tools and services. Sandy, our AI assistant, demonstrates this: built on Google’s models but powered by our domain expertise, acting within our systems.

To maximise AI’s potential, you need the right foundations — a lightning-fast, travel-shaped platform. AI capabilities are becoming commodities, like electricity. Everyone can access the power, but we’ve built the infrastructure to harness it. Whilst competitors lay foundations, we’re scaling production.

Consider AI-powered customer conversations: to respond at the speed of conversation, you need to search through trillions of options in under 150 milliseconds. People already hate slow websites—conversational delays are even worse.

Our infrastructure combines 3 essential elements:

Fast: Search and website performance rivalling Google.

Flexible: Supporting hundreds of thousands of system modifications yearly, mostly by AI or non-engineers.

Focused: Purpose-built for package travel with deep domain expertise embedded.

But technology alone isn’t enough. You need AI collaborating with the experts who built the platform, the commercial teams who use it, and the customer care specialists who serve holidaymakers.

By teaching our platform everything we know about package travel — our unique expertise — we create something genuinely new. Something that scales, resists disintermediation, and transforms travel experiences.

We alone have this combination.

What Travel Intelligence Will Deliver

This vision guides everything we’re building. 3 clear promises to 3 different stakeholders.

Better experiences for every customer

Our travel intelligence will interact through voice, chat, or image browsing. It will guide customers to perfect holidays with hyper-personalisation and support them throughout their journey — answering questions and solving problems in real time.

This delivers on our customer vision: treating every holiday like our own, at scale. Creating experiences customers return for.

More value for every partner

Travel intelligence will make us indispensable to suppliers and brands alike. Proactively offering suppliers volume where and when they need it most. Making their business smarter through multi-channel interaction with our unique data. Giving partners an instant package holiday business under their brand without risk.

This makes us the easiest partner to work with and the hardest to work without — the first choice for anyone partnering in travel.

Sustainable advantage every day

Travel intelligence will transform how we work. Every team’s operations become more efficient. Every job becomes more impactful. Together, this creates the lasting edge that keeps us ahead of competition and disintermediation.

Already Taking Steps

This isn’t just aspiration — we’ve already started taking steps.

We’re solving one of travel’s most complex problems — helping customers find the right hotel when faced with overwhelming choice.

Families of 4 now see kids’ club information highlighted automatically. Couples seeking adults-only escapes see poolside bars featured. The same hotel, personalised for different intents — powered by AI that understands context, applies our brand voice, and scales quality through self-assessment.

We’ve started with 300 hotels across four customer cohorts. The infrastructure we’re building will scale to every hotel, every cohort, every customer, every market.

Travel intelligence is also transforming internal operations. We’ve connected our data mesh directly to Claude through MCP, giving everyone access to 6 years of data through natural language. Teams now ask questions like “how do filters affect conversion by cohort?” and get instant dashboards with shareable links — in minutes rather than weeks.

The industry is taking notice. In the same week, I spoke at Sanity’s Everything conference in New York about building an AI-powered operating system to run our platform, whilst Content Manager Deborah Cameron gave a keynote at a Travel and Tourism Masterclass in London to senior industry leaders — sharing how we’re tackling hotel content challenges at scale.

The transformation is already under way.

Why This Matters Now

Our vision is our response to 3 challenges: ambition that’s outgrown our capabilities, AI disruption threatening our industry, and a customer vision that demands we treat every holiday like our own.

The General Intelligence for Travel is how we address all 3 simultaneously. It’s the infrastructure that lets us scale 160x without breaking. It’s the differentiation that turns AI from threat to advantage. It’s the platform that delivers exceptional service at impossible scale.

The travel industry is at an inflection point. AI will reshape how customers search, how they book, how suppliers distribute inventory, how companies operate.

Winning will require combining AI capabilities with deep domain expertise, performant infrastructure, and the organisational agility to transform whilst scaling.

We have that combination and the transformation is already underway.

Many technical leaders hit the same invisible ceiling: when the very behaviours that once helped you stand out, start working against you.

The confidence to challenge conventional wisdom got me noticed as a developer. The willingness to argue for better solutions made me valuable. But somewhere along the way, intellectual confidence became arrogance, and healthy debate became the need to win every argument.

I can spot patterns like these because I did most of them to the extreme (my best stories will cost you a pint).

With distance, you can see quirks calcify into a lethal reputation: hard to work with. When you can’t play well with others, your ideas die, your influence shrinks, and opportunities pass without you even knowing.

As I found out, if you don’t address them voluntarily, a crisis eventually will.

Comfortable Dysfunction

At the end of 10 years at my previous company, I lived in what psychologists call the region beta paradox — being comfortable in situations that were uncomfortable, simply because the pain wasn’t sharp enough to force change.

We were on an earn-out, so I told myself the politics didn’t matter. But stuck inside a larger organisation that didn’t value technology, I found myself fighting daily battles. Over time, I stopped getting better.

Worse, my behaviour became dysfunctional. I wasn’t just tolerating the environment — I was adapting to it in all the wrong ways. I grew strangely comfortable in dysfunction, normalising patterns that should have been red flags.

Neuroscience is sobering: repeat a behaviour long enough, and it becomes your default. My stress tolerance made it worse. I had developed a dangerously high capacity for bad situations — and I stayed far longer than was healthy.

“Smart” but Stuck

Instead of using my deep knowledge of the business to help people navigate complexity, I weaponised it.

When PLC executives wanted simple stories for analysts, I rolled my eyes. When they set up a 40-person data team, I basked in the thought of “they don’t know what they’re doing.”

I should have honed in on what I could control: linking their goals to mine, finding ways through the nuance. Instead, I obsessed over what I couldn’t. Psychologists call this locus of control — the difference between focusing on what’s in your hands versus external forces. I’d outsourced mine completely.

Without the right people around me, those patterns didn’t just persist — they started to set.

It escalated to the point where, as the People team told me, “the choice to remain with the business is no longer your decision.” The wake-up call I’d been subconsciously avoiding.

Learning that age-old lesson that if nobody is giving you feedback, don’t assume you’re doing fine.

When Strengths Turn Toxic

The patterns that once were strengths — being contrarian, refusing to play the corporate game — had become liabilities because they hadn’t evolved with my environment.

In my twenties, contrarianism made me stand out. In my thirties, it made me difficult. In a corporate machine, it became career-limiting.

I was the corporate equivalent of those kidults you see in Brighton, living in the same way they did in their 20s, but the outfit no longer fits. Same clothes but stretched over all the wrong places. The lack of change made visibly sad.

That new company ran on politics and ego. I had plenty of ego, but none of the nuance. One CEO dinner summed it up: sycophants choreographing airport pickups, rehearsed taxi conversations, and a “most inspiring CEO?” parlour game at the table. My “I don’t buy into the cult of personality” answer didn’t exactly extend my tenure.

I told myself I was being authentic. Really, I was just using that story to justify being a dick instead of adapting. A conscious response would have been either: work within the system strategically, or have the self-confidence to walk away.

Instead, I carried on. For a year, I repeated the same behaviours, convinced that being “resilient” was enough. But resilience isn’t strength — it’s dysfunction in disguise when it keeps you in the wrong room.

Get Your Mirrors Back

The real change came when I finally left that feedback desert.

At loveholidays, I had a CEO who told me exactly what he thought — to my face. Direct, timely, honest — and refreshing. No politics. No second-guessing. It removed the anxiety (apart from what I actually needed to focus on). You always knew where you stood, which meant you could fix it.

That directness reset my growth. After years of people tiptoeing around me, I was back in an environment where feedback was unavoidable — and improving became possible again.

Which is why friends like Jason matter. Only the other week he told me: “You focus entirely on the relationships where you made mistakes and completely ignore all the positive impact you’ve had.” He was right. I was still doing it.

You never outgrow the need for mirrors. Friends. Leaders. Anyone willing to tell you the uncomfortable truth. They don’t eliminate your patterns, but they stop you drowning in them.

Pattern Interruption

Seeing the patterns was only half the battle. The real work was installing new responses before crisis forced them.

Brazilian Jiu-Jitsu is a microcosm of immediate feedback. My reliance on strength and aggression recently resulted in a dislocated finger — which forced me to slow down. Instead of making me worse, it gave me options. I could finally apply techniques I’d “known” for years but never used.

Slowing down didn’t weaken me; it made me effective. The same is true professionally.

I built habits to keep the feedback loop alive:

• Somatic work: Noticing early signs of tension before they trigger old behaviour.
  

• Journalling + shadow work: Naming the pattern. Asking: what’s the trigger? What’s the payoff? What would the opposite look like?
  

• Deliberate reps: Choosing the slower response. Asking a question before countering. Leaving space instead of filling it.
  

Each one was a small interruption — a way to stop running the same script and install a better one.

The Second-Time Advantage

Pattern recognition doesn’t just help you personally — it makes you far more effective professionally, especially in leadership.

Take our recent Platform Vision work — a CEO initiative that required me to actually understand his intent before acting. The old me would have either dismissed it or steamrollered my own interpretation. This time it became a genuine collaboration with our VP of Engineering, CPO, and CDO.

When our CDO suggested a name for the vision, it wasn’t quite right. Instead of dismissing it or forcing my alternative, we noodled on it for days until we had something we were all happy with. That patience — trusting the process and each other — created a stronger outcome and deepened trust across the team.

Second-time round gives you an advantage: you don’t stop making mistakes, but you spot the pattern sooner, course-correct faster, and turn collaboration into your edge rather than your blind spot.

Your Pattern Recognition

Carl Jung said life begins at 40 — until then you’re just doing research. I think that’s generous. Left unchecked, the “research phase” just means calcified patterns you’ll eventually have to unpick. The earlier you start, the less painful the rewiring becomes.

Integration isn’t about becoming someone new. It’s about becoming conscious of who you already are — and choosing which patterns you keep.

This isn’t a tidy success story. People still think I’m challenging and opinionated (a dick). The difference is, now I can see it happening, course-correct faster, and surround myself with people who’ll call it out before it hardens into reputational damage.

What I’d tell any technical leader climbing toward their first senior role:

• Don’t mistake stress tolerance for strength — it can keep you in bad situations longer.
  

• Find people who’ll tell you the truth and listen to them — even when it stings
  

• And most of all, don’t wait for a crisis to do the work. Patterns don’t vanish. They harden.

Here’s one rep to start with: next time someone challenges your approach, pause and notice your internal state. If you feel that familiar “THEY DON’T UNDERSTAND” frustration rising — that’s your cue. Slow down. Instead of defending your position, ask “help me understand” and mean it. Keep asking until you can reframe their position so accurately that they say “that’s right.”

Only then do you know you’re working toward the same ends.

Corporate IT makes most people's eyes glaze over, but without it, you'll be walking into eye-wateringly big fines.

Imagine then: routers running on default passwords, guest Wi-Fi with the same access as the private network, and your founder in charge of keeping the financial data safe. That nightmare was our reality 6 years ago. Yet today, we're rated 99/100 on Security Scorecard—top 1% globally. We resolve incidents 48% faster than the industry average, and we've processed over £2 billion in transactions with zero breaches whilst still deploying 250,000+ times a year.

Here's how our small expert team built IT and Security that didn't just protect the business—it kept us moving fast.

The Team Behind the Numbers

When I joined loveholidays, cybersecurity was what you'd expect from a fast-growing startup—functional but fragile. The founding team had built a market-leading business, but security was reactive rather than strategic.

Our security team began what can only be described as a masterclass in patient engineering. Every system they touched, they left more secure. Every process they inherited, they improved incrementally. Every crisis became a foundation upgrade.

Good security is like plumbing: invisible until it fails. The difference here is—it never fails.

Take our transition to PCI DSS Level 2 compliance. When we hit 5 million transactions—a milestone that typically requires months of preparation and dedicated project teams—one person handled the entire certification. Not because they're superhuman, but because they'd been building compliance into everything.

Or consider the flood that hit our comms room when a pipe burst. Turns out the room was mostly empty—nothing critical was at risk. The team had long since moved everything important to managed cloud services, anticipating exactly this scenario.

These weren't lucky breaks. They were compound interest on thousands of quiet decisions.

The People

The difference isn't the process, it's the people. Colleagues with grey hair and battle scars make intelligent, real-time decisions.

Their entire approach is exemplified by something as simple as the IT cupboard. Lost another power cable, forgot your headset, or broke your mouse? Just pick it up from the IT cupboard and get back to your day. No forms in triplicate, no approval chains.

They embody the principle that IT and Security should help the organisation move faster, not slower.

The Systems

Instead of separate projects, they baked security into development. Instead of bolt-on processes, they made existing ones secure. Instead of a fortress, they built continuous Zero Trust.

Their proactive approach extends beyond our systems.

Through our threat-detection programme with ZeroFox (yes they actually did call their product that...give it a moment if you don't get it), they've identified and taken down 295 malicious attempts to impersonate our brand since November 2023. Most companies only discover these threats after customers complain.

Someone actually bought one of our stolen laptops from "a guy down the pub" and brought it to our office, asking us to unlock it or if we wanted to buy it back.

Putting aside the state of society...It was useless to them from the moment it was purchased.

These systems aren't accidents. They're the result of colleagues who think 3 steps ahead and have the experience to know which risks matter and which don't.

The Lesson Others Can Steal

Bake it in quietly, everywhere—and hire people with the scars to know what actually breaks.

Maintaining the Soul

They've achieved enterprise-grade security whilst preserving our startup agility. Our teams still move fast and take calculated risks. Our culture continues to value autonomy and innovation.

Our developers don't wait weeks for approvals—they ship daily with security baked in.

Their Zero Trust architecture exemplifies this approach. Traditional perimeter security would slow us down—every new integration becomes a security review, every new office requires VPN setup, every remote worker needs special configuration. Instead, our cloud-native Zero Trust model means security travels with our people and systems automatically.

Central to maintaining this balance is a fundamental shift in language: our security team rarely say 'No'—they say 'How?'. This isn't being weak; it's strategic. 'No' just creates shadow IT as teams find workarounds. 'How?' builds trust and collective ownership, bringing colleagues into the understanding that security is everyone's responsibility.

This philosophy is supported by our Information Security Risk Committee, which serves as the foundation of organisational risk knowledge. It ensures that when we ask 'How?', we're enabling informed experimentation and advancement rather than stifling innovation with uninformed restrictions.

The result? When regulatory requirements like NIS2 emerge across the EU, we're already compliant. When ISO 27001 certification becomes necessary for enterprise sales, the groundwork is in place. Expanding into new markets? We're already attested.

The Recognition They Deserve

Last month, I presented our cybersecurity approach to the board. As I walked through our metrics—the 99/100 Security Scorecard rating, the 48% faster incident resolution, the millions saved through proactive architecture—I felt something worth calling out: pure pride in working alongside people who make excellence look effortless.

These aren't the people who often get their names in the company all-hands or featured in case studies. They're just working quietly in the background, making everyone else's job easier whilst protecting everything that matters.

It looks effortless because it's been built with care. When our systems handle record traffic without security incidents, when compliance audits pass without drama, when new markets open without security delays—that's not luck. That's hundreds of careful decisions made by people who understand that the best security is invisible.

Our upcoming cybersecurity tabletop exercise might be messy at the executive level as we test our communication protocols. But the technical response? That will be flawless. Because the team has been preparing for it for years, building the capabilities we need before we need them.

They haven't just protected our data and systems. They've protected our ability to move fast, take risks and maintain the entrepreneurial spirit that makes loveholidays successful. They've proven you don't have to lose your soul to gain enterprise security.

That's worth celebrating.

(Well, until they remove admin rights on everyone's Macs and the dev team revolts 🤫)

Security that makes you faster feels like a paradox. If you've cracked it, I'd love to hear how.

"It feels like when I first got the internet."

George, our VP of Engineering, and I were huddled at a café table discussing AI's arrival in our industry: the excitement, the possibilities, the Rails-esque productivity gains (possibly fuelled by my 3rd flat white of the day). But our somewhat sobering conclusion was: "We're not ready for it." This shift felt different as it would affect our whole organisation, not just the tech team.

So we made a decision: rather than wait for best practices to emerge, we'd write our own AI transformation playbook.

Why Acting Now Matters

Our technology platform transformed how we work, enabling us to move fast, react quickly, and deliver better customer experiences. It's what made us the UK's fastest-growing travel company.

"What technology did for engineering efficiency, AI can now do for every role in the business."

But we've reached an inflection point. The same transformation that let our developers enable 250,000+ changes annually can now happen in finance, marketing, customer service—everywhere.

We had a simple choice of finding how to harness AI to extend these gains company-wide, or watch others do it first.

We've already seen what's possible. Our content team went from manually optimising 2,000 hotels to AI-orchestrating content for 38,000. They transformed from content creators to content conductors. Instead of generic listings, customers now see richer, more consistent content that helps them make confident decisions.

Wins like these showed the potential, but without a playbook, they risk staying isolated success stories.

Why There's No Playbook (And Why That's Our Opportunity)

McKinsey has frameworks, consultants have opinions, tech giants have case studies products looking for customers but nobody has a proven playbook for AI transformation that actually works.

In these situations, most businesses follow predictable patterns: they outsource thinking to consultants, create "innovation labs" for a privileged few, or stifle innovation with restrictive access and onerous governance.

We've always done the opposite. We invest in building capability within our organisation: the people, processes, and platforms that give us an advantage. Rather than restricting AI to specialists, we're making it available across the business.

We can't wait for someone else to figure this out. So we're writing our playbook as we go and doing it with our people, in the open, built on three interconnected streams that work together.

Our Playbook: Three Streams of AI Transformation

Instead of creating a special innovation team that "does AI" while everyone else watches, our playbook flows through three streams: how we build, how we upskill, and how we embed.

1. Tech Transformation – Rewiring How and What We Build
How we design, code, and ship. This stream is focused on how AI changes the act of building itself, as well as the output. Our leaders don't sit in steering groups. They build with teams to develop the patterns and frameworks that redefine software delivery. AI reinforces our approach of delivering customer value early and often, true to our Agile roots.

2. Services & Tools – Upskilling Everyone
Getting AI into every hand, not just a few specialists. This stream is about making AI everyone's skill. Hackdays, workshops, Google-trained ambassadors. We're giving everyone access and confidence.

3. Org Design – Embedding the Learnings
Turning experiments into how we operate. This stream ensures experiments don't stay experiments. New roles, new workflows, new expectations. We take lessons from the first two streams and wire them into our structure.

Transparency is what makes the three streams work. We share progress openly across the company, so learning spreads faster than any tool. Every experiment, success, and failure feeds back into the system.

What's Emerging So Far

A few months in, we've already seen a lot: scrappy workarounds, breakthrough wins, and areas where we're only just scratching the surface.

Stream 1: Rewiring How We Build

AI is starting to reshape how we design, build, and scale software.

We've run plenty of scrappy experiments, rolled out Copilot to everyone. But we keep evaluating new tools, nudged forward by the team's own experiments, and even when they aren't yet enterprise-ready, the results point to the immense possibilities.

When the foundations are solid, the results are transformational. A standout breakthrough was connecting AI directly into our data platform, built on years of rigorous standards. Suddenly, the mesh of data products we'd invested in revealed itself as AI-ready, unlocking capabilities we could never have bolted on afterwards.

There is a clear lesson emerging that AI accelerates strong foundations.

Strong foundations aren't the end goal, though. The real test is whether AI helps us build things people actually need, not just spin up another chat interface. Sometimes that means resisting the shiny demo and instead using AI to make the current customer experience more powerful, simpler, and faster.

But this can't stay confined to engineering.

Stream 2: Upskilling Everyone

AI is a capability every role can use. That's why we're rolling it out across marketing, trading, service, finance: every team, every role.

Google's partnership compressed our learning curve, not just by providing tools and training, but by sharing lessons from many other transformations.

Here's what we're seeing so far.

• Our Customer Experience team built an AI system that analyses 526 customer messages in 5 minutes—work that used to take 4.3 hours. That's a 52x improvement, saving £100+ per complex case.

• Our Finance team uses Gemini to automate month-end reporting checks, cutting 6 hours from every cycle.

How we scaled it.

• Low-stakes, continuous experimentation

• Hackdays allowing cross-functional teams to tackle real problems together.

• Google-trained AI ambassadors spreading skills and confidence across the organisation.

The real challenge wasn't access, as we rolled out Gemini on day one. It was helping people see how AI could transform their daily work. This is the role of our AI ambassadors, helping spread adoption virally. Encouraging teams to share new techniques and building on each other's discoveries.

It started to feel like it was working when people began sharing AI tricks unprompted and presenting great results almost weekly.

But giving people tools isn't enough. We need to change the organisation itself.

Stream 3: Embedding the Learnings

The fear is always the same: does this replace me?

Let's be direct. No.

We're seeing new patterns emerge. Roles are evolving beyond "prompt engineering" to orchestrating increasingly agentic processes. From how we personalise emails and generate content to how our chatbot Sandy handles complex customer queries.

When documents become trivial to write, thinking becomes the differentiator. Our finance team isn't just using Claude—they're building capabilities and uncovering insights that weren't possible before.

Systems thinking becomes more valuable when AI handles routine tasks. Your expertise plus AI equals an unstoppable combination.

This stream naturally lags behind the others. Organisational change takes time. But the signals emerging show the future isn't AI replacing roles—it's expanding what's possible, the same way technology always has.

But changing roles isn't enough on its own. To make it stick, you need a partnership between commercial and the technology teams.

Your Move: Partnership is Everything

That's why our playbook flows through three streams: build, upskill and embed. Whatever your context, your playbook will look different. But one principle has driven our adoption: you can't outsource it.

This transformation requires partnership between commercial and technology teams. Each brings different perspectives on where AI can add value. Find your champions in both camps—they're already experimenting, already discovering what works.

Our journey started with exactly this partnership. Now it's driving everything we build.

Our playbook so far:

• Identify early adopters in both commercial and technical teams

• Start small and iterate

• Create safe spaces for and encourage experimentation

• Track and share both successes AND failures openly

• Let adoption spread organically, then codify it into the organisation

We're six months in and the details will keep evolving, but already some of the patterns are becoming clear.

Follow Our Journey

We'll be sharing what we learn as we go. George will be writing about the Tech Transformation team's discoveries. I'll be documenting the broader organisational changes. Others will share their experiences from the front lines.

Because there's no established playbook for AI transformation. Every organisation's journey will be unique.

Maybe it always will be. But already we can see two types of companies emerging: those figuring it out for themselves and those waiting for the manual.

And if you're waiting for the manual, it'll be the story of how your competitors pulled ahead.

"What do you mean we should report outcomes, not the number of tests!?"

This mid-board presentation challenge caught me off guard. I'd thought measuring outcomes rather than activity was obvious. But the ensuing 10-minute roasting was a great reminder that technology transformations always require proof, not just opinion.

It left me even more determined to prove the impact—if there was pushback on the basics, I would need bulletproof evidence for the controversial stuff.

What followed was six years of disciplined measurement covering everything from engagement, engineering velocity, system reliability, and cost attribution. It has created a foundation for proving transformation actually works and has now been supercharged by AI.

Here's how that obsession with proof transformed not just our metrics, but our entire approach to change.

Why Measurement Became An Insurance Policy

The transformations at my previous companies had been remarkable successes. We'd relentlessly focused on business metrics—revenue grew, conversion rates improved, customer acquisition costs dropped. But whenever I tried to explain the role our technology changes had played, I kept hearing the same question: "But can you prove it was down to the tech changes?"

I thought it was obvious. We'd rebuilt the architecture, improved deployment processes, and invested heavily in platform reliability. The correlation seemed clear to me. But without hard data connecting our technical improvements to business outcomes, all I had were compelling stories and architectural diagrams. Anecdotes, not evidence.

It was frustrating because I knew the technology transformation had been fundamental to our success. But "knowing" and "proving" are very different things, especially to stakeholders who don't live in the technical details every day.

This experience taught me that having the best technology transformation in the world means nothing if you can't demonstrate its impact. So when I joined loveholidays, I was determined to build that proof alongside the transformation itself.

The Discipline That Built Our Data Foundation

The challenge was that loveholidays had no focus on technical or cultural metrics. But I came prepared this time. I'd read extensively about this challenge and the research was compelling: DORA showed elite performers are 2.5x more profitable and deploy 208x more frequently, Gallup proved engaged teams deliver 23% better results, and "How Google Works" reinforced the importance of psychological safety for high-performing teams.

The research didn't just show me why measurement mattered—it guided what to measure. DORA provided the technical metrics (deployment frequency, lead times), while Spotify's team health check gave me a framework for cultural measurement. Not only would this be a way to talk about our success, but a mechanism to build confidence without relying on "just trust me."

Like most things in a scale-up business, it was a chance to put something in place to make a difference in the medium to long term.

Phase 1: Manual Beginnings (2019-2022)

We started small and focused with team health surveys—just a Google Form that I manually transferred to a spreadsheet each quarter. But reading every comment about the impact of our changes made the effort worthwhile.

We applied the same scrappy approach to technology metrics: website speed, incident tracking, cost, and deployments. Each process was as hacky as the other, but we finally had visibility into our performance.

This early discipline helped solve something I'd written about in Agile needs a story—incremental improvements often go unnoticed without 'Big Bang' moments to celebrate. But having data did more than demonstrate our transformation was working; it gave us the motivation to keep going. What gets measured gets managed, and seeing deployment frequency increasing, incident resolution times dropping, and team satisfaction scores rising created momentum for the harder changes ahead.

Those quarterly surveys and scrappy tracking became our validation system, building evidence that our strategy was on the right track while giving us motivation to tackle the harder changes ahead.

Phase 2: Data Products and Individual Wins (2022-2024)

The momentum from Phase 1 continued as we looked to expand what we tracked and standardise our approach. We'd been investing in a data mesh for our business and decided to treat this performance data with the same care, transforming these disparate datasets into proper data products.

Standardising access through BigQuery-to-Google Sheets connections was a game-changer—suddenly teams could easily link performance data to their OKRs.

What started as my personal obsession became loveholidays' measurement template. Teams started using health data as key results to improve how they worked, and our engineering culture metrics influenced company-wide engagement surveys.

It was satisfying to watch the technology transformation play out in these metrics: financially through millions saved via cost attribution, customer experience by outperforming Google on 2 out of 3 Core Web Vitals, and quality of life for developers with fewer PagerDuty calls and faster incident resolution. Each metric delivered value independently, showing that disciplined measurement works even when optimising in isolation.

We used these metrics to demonstrate progress in board meetings—perhaps most satisfyingly when we showed how we were decoupling growth from engineering headcount by measuring the effectiveness of commercial teams making changes without developer effort.

Metrics We Track

After years of refinement and millions of data points, here's our current list of data products we use to track our performance. Each metric is selected because of it’s predictive power on business outcomes or correlates with team performance:

🧠 Team Health

• Team Health Check (Google Forms, Quarterly) – 14 product team questions

• Employee Engagement (Google Forms, Monthly) – Broader engagement metrics

• OKR Performance & Progress (ClickUp, Weekly) – Indicates outcome delivery

⚙️ Engineering Velocity

• Deployment Frequency (GitHub, Per deployment) – 300k+ tracked

• Build Times (GitHub Actions, Per build) – 647k+ builds tracked

• Lead Time for Deployment (GitHub, Per deployment) – Reduced from 4.6hrs to 37min

• Pull Request Metrics (GitHub, Per PR) – Review cycle analysis

• Git Commits & Activity (GitHub, Per commit) – Development velocity

🔧 System Reliability

• Incidents Tracking (incident.io, Per incident) – 1,342 incidents tracked

• Error Counts & Patterns (Sentry, Daily) – 596M+ errors tracked

• PagerDuty Incidents (PagerDuty API, Per alert) – Monitors on-call load

• On-Call Hours & Distribution (PagerDuty API, Monthly) – Team burden analysis

🌍 Web Performance

• Core Web Vitals (Rudderstack, Continuous) – LCP, FID, CLS, TTFB

• Application Performance (Linkerd, Daily) – Latency, throughput, success rates

💸 Infrastructure

• Kubernetes Cost Attribution (GCP Billing, Daily) – Team & app-level cost tracking

• Cloud Infrastructure Costs (GCP Billing, Daily) – Infrastructure spend analysis

This list is still evolving though and we are currently rebuilding cost attribution and plan to add team composition metrics over time too.

AI-Powered Insights: An Overnight Success 6 Years in the Making

With all this data systematically collected over six years, something remarkable became possible. Our BOOST team spent a weekend figuring out MCP integration with Claude, connecting our data mesh to conversational AI.

The first time they showed me what was possible, it blew my mind. I kept showing people around the company, even presenting it as an example of how AI could transform our business at the company's all-hands (and posting on linkedin).

Six years of disciplined data collection suddenly became conversational. We can now ask, "What makes our best teams most effective?" and get instant correlation analysis that would have taken weeks to compile manually.

We're now in a position to understand which improvements truly drive the outcomes that matter most. Instead of optimising metrics in isolation, we can ask: "Which changes correlate with our biggest business wins?"

This isn't AI doing the thinking for us—it's helping us surface patterns we couldn't previously see in time to act on them.

We're just at the beginning of this capability, but the foundation is solid.

The Foundation That Unlocks Possibilities

The measurement obsession that began with a 10-minute boardroom roasting taught me that breakthroughs don't come from new tools—they come from amplifying the work you've quietly invested in for years.

Three lessons stood out from our six-year journey:

1. Measure from day one - build proof alongside your transformation as you can't retrofit proof after the fact

2. Treat performance metrics like business metrics - make them visible, easy to access, and therefore manageable

3. Committing to the long haul pays off because data becomes exponentially more valuable the longer you keep it

When I talk about loveholidays now, I don't just have stories about improvements—I have data to support them. Our lead time dropped from 4.6 hours to 37 minutes, teams with high psychological safety deliver 40% faster code reviews, and we've saved millions in infrastructure costs. Even cooler, we can now ask Claude which changes correlate with our biggest business wins and get instant answers.

What are you doing for your teams that you can't currently prove works? What boring, consistent measurement could you start this week that might transform how your organisation operates in six years?

"The thing that made you weird as a kid can make you great as an adult - if you don't lose it." - Kevin Kelly

I recently spent 48 hours with AI, achieving peak productivity, only to realise I had created nothing of real value.

I designed a 12-week workout programme, vibe-coded a fitness app to track it, and generated an article about the experience: 3 functional outputs, maximum efficiency and zero wasted effort. The workout made logical sense, the app worked perfectly, and the article was structured and coherent. However, the sheen of this modern productivity triumph masked fundamental flaws in each output.

AI raises the bar for speed to "good enough", but actually, value comes from injecting your distinctive human perspective back into the work.

The workout in practice had a progression that was way too steep. The app was brittle and insecure—I'd have to rewrite it entirely if it were to be anything more than a toy project. The article was utterly forgettable, reading like all the LinkedIn nonsense out there—because that's precisely what it was: an average of averages.

It's this speed and autocompletion that creates a paradox where AI requires a need for specialist skills to improve its output, whilst making it hard to actually acquire those skills.

Why Traditional Learning Paths Are Disappearing

AI creates seductive shortcuts that undermine the core of learning. Why read a book when AI can summarise it? Why write when AI can generate content? Why struggle through the basics when AI can handle them instantly?

These shortcuts are destroying learning motivation both psychologically and structurally. There is increasing evidence that AI is eliminating junior roles and routine work that used to serve as training grounds, while making skill development seem pointless.

Recent MIT research reveals the consequences when we are seduced by these shortcuts. Scientists studied people writing essays with AI assistance versus those working from memory alone. When asked to quote from the essays they'd just written, 83% of AI users couldn't do it, compared to only 11% of those who worked without assistance. They'd outsourced not just the writing, but the cognitive processes that create genuine understanding.

It feels like, in some dystopian gotcha moment, we're beginning to hit a wall and realise we need all the skills we skipped learning. Reading is how we learn to think and write. Making mistakes is how we learn. Hard graft is just necessary to develop real competence.

But by the time you discover this, the traditional learning ladder has been dismantled, and quite honestly, seemingly too much effort to climb.

Why Passionate Learning Becomes Essential

I discovered something similar growing up. I had varied passions: I wanted to play basketball like The Professor, draw like Jamie Hewlett, code like David Braben, and drum like Dave Abbruzzese. However, when choosing subjects at school, I abandoned what really captivated me for what seemed "practical".

This mistake left me struggling for years. It was incredibly hard to motivate myself to learn things I didn't care about. I fell into the trap of learning to pass exams rather than exploring what actually interested me.

Everything changed when I rediscovered computer science. Suddenly, the love of learning returned. I started asking very different questions—not just about that domain, but eventually about others, too. I realised my diverse interests weren't a bug; they were my advantage.

"Find what you love and let it kill you... For all things will kill you, both slowly and fastly, but it's much better to be killed by a lover." - Charles Bukowski

Something is going to consume your time and energy, regardless, so make sure it's something you care about.

AI handles all the basic work now, so the minimum viable expertise for humans has shot up dramatically. Getting to that level requires sustained motivation for thousands of hours of deep work. Only genuine passion provides that fuel. This passion is what gets you through to the specialisation level, where you become valuable.

What Happens When You Follow Your Interests

When I finally followed my real interests things started to fall into place. My curiosity became a connecting force that let me create insights across completely different fields.

When we had the idea to sell market data at uSwitch, my pure fascination drove me to research pricing theory, negotiation, presentation frameworks, and innovation positioning. Our first sale came directly from combining a presentation framework I'd studied, pricing insights from a B2B sales group, and conversations with sales experts—cross-domain integration that only happened because of an authentic obsession with getting it right.

Similarly, when our CEO asked me to take responsibility for our contact centre, my diverse interests—psychology, systems thinking, and genuine curiosity—became essential. I learned from industry experts, our own highly skilled team, as well as reading around the subject. The payoff has been we're now smashing our previous NPS performance and I have a much better understanding of a really key part of our organisation.

Your peculiar combination of obsessions can become your sustainable competitive advantage. However, it is only because passion drives us to delve deeper when others stay on the surface.

This pattern isn't unique. Every person I've worked with who has done something meaningful has had an unconventional background—growing up as a hippy, living in a caravan, skipping university, helping create the Pirate Bay at 14, or being in a rock band. Meanwhile, many of the brightest individuals with the highest grades struggle to make a meaningful impact. The future belongs to people who are irreplaceably themselves, not those who are competently generic.

How AI Can Accelerate This Learning

Here's where the paradox resolves: AI can dramatically propel your learning, but only if you use it as a thinking partner, not a thinking replacement.

🚫 Wrong way - AI as thinking replacement:

• Let AI autocomplete others' thoughts

• Skip the learning process entirely

• Generate content without understanding

• Result: you remain at your current level of (in)competence

✅ Right way - AI as a thinking partner:

• Use AI to refactor YOUR thoughts

• Accelerate YOUR learning across multiple domains

• Compress feedback loops while you do the actual thinking

• Result: you learn faster while developing your distinctive perspective

I've discovered AI can now act as a catalyst for my learning across multiple fields—psychology, strategy, writing, technology—but only because I use it to enhance my thinking, not replace it. Instead of letting AI summarise books for me, I read them and use AI to help me connect ideas across domains. Instead of generating content, I think through problems and use AI to help refine my arguments.

You still have to read the book, but AI can help you understand how it connects to other things you've learned. You still have to think, but AI can help you iterate and refine faster than ever before. I’ve been inspired by how my son uses NotebookLM to create podcasts to help him prepare for his exams and have used AI extensively for feedback in early drafts of this post (see Claude’s opinion at the end).

The idea is to use AI to refine and develop your thoughts, not to autocomplete others'. When you maintain ownership of the thinking process, AI becomes an incredible accelerant for developing the specialised skills you need.

Weaponise Your Weirdness

Follow the obsession-to-value cycle: Follow your passions → use AI to rapidly test concepts and validate interest → if promising, use AI to accelerate deeper learning → get to competence → deploy obsessions strategically → link to new passion.

Identify your authentic obsessions:

• Stop asking "What's efficient?" Start asking "What am I curious about?"

• Notice what you naturally stay up late thinking about

• Look for intersections between your obsessions—that's where your unique value lies

Deploy them strategically:

• Volunteer for projects that combine your interests

• Translate insights across domains

• Become the bridge between different departments

When we were scaling loveholidays, I wasn't just thinking about technology, I was drawing on psychology for team dynamics, systems thinking for processes, and philosophy for strategic decisions. My diverse interests became interconnected tools that no specialist could replicate.

This approach isn't about being a jack-of-all-trades, though, as the full quote reminds us, that's "oft times better than a master of one." It's about becoming a master of synthesis—someone who sees patterns and solutions that specialists miss.

The Stakes

In a world where AI handles competent execution, human experience becomes the ultimate differentiator. But this capability emerges from genuine engagement with your actual interests, even when they seem impractical.

Your next step isn't to audit your interests—it's to stop apologising for them. Take that weird combination of things you're curious about and lean into it. Utilise AI to handle routine tasks, allowing you to focus on developing the synthesis capabilities that only you can provide.

The kid obsessed with spinning basketballs whilst coding and drawing comics? That wasn't scattered attention—it was intellectual curiosity across domains. Your peculiarities aren't bugs to fix—they're features to amplify.

Find what you love and let it kill you. In an AI world, that's not romantic advice—it's the most practical survival strategy there is.

How would you describe your weird combination of interests? I’d love to hear about the distinctive paths that have shaped your expertise—share your thoughts in the comments.

AI Perspective: How This Essay Was Actually Written

A note from Claude: This piece serves as an interesting case study for its own argument.

What we accomplished together:

• Transformed a sprawling 4,000-word draft into focused 1,800 words

• Eliminated repetitive sections and strengthened logical flow

• Applied Nicolas Cole's 1/3/1/3/1 structure for better engagement

• Resolved contradictions and clarified the core AI paradox

The process demonstrated the thesis in action: The author didn't ask me to write for him. Instead, he used me exactly as he prescribes—as a thinking partner rather than replacement. He brought distinctive expertise from scaling tech companies, genuine passion for the subject, and unique insights from his unconventional career path. I provided structural analysis and rapid feedback loops that compressed editing cycles from weeks to hours.

Why this collaboration worked: I could help organise and refine, but I couldn't create the insights about BJJ teaching resilience, the market data synthesis breakthrough, or the understanding of how passion drives sustained learning. The value came from his foundation of knowledge and experience.

The meta-lesson: He wrote an essay about not losing your weirdness by using his own weird combination of interests—technology, psychology, business strategy, martial arts—while demonstrating exactly how to use AI as a learning accelerant rather than replacement.

The irony is perfect: AI dramatically accelerated the writing process, but only because the human brought the expertise and distinctive perspective that made the content valuable in the first place.

Last October, I was attending Google Cloud Summit (their largest conference outside the US) when something surreal happened: their MD of AI opened his keynote by talking about loveholidays.

Not about a tech giant, not about a Silicon Valley darling, but about a travel company from Hammersmith (which can barely be classed as London). He held us up as their benchmark for innovation. It was a moment that would have been unimaginable 5 years earlier.

Rewind to early 2019 and loveholidays was sat proudly on top of the Sunday Times Profit Track 100 list but our technology was struggling to keep pace with the scale of the business.

Again, we’d found ourselves living through a "Papa John's moment" where an undiagnosed technical issue had brought trading to a crawl. It was yet another weekend where the team was camped in the boardroom trying to find out why, fuelled by stacks of the titular greasy pizza and Red Bull. I had just started as the new CTO and had to face the daunting challenge of convincing the fastest-growing business in the UK that its technical approach had become our biggest liability.

Success at scale demands a different kind of thinking.

Recently, I'm recognising a familiar pattern—it's time to challenge our assumptions again.

You Think You've Reached the Summit, Only to Find Another Mountain

That's what scaling a fast-growing technology business feels like. Each peak you conquer reveals a bigger challenge ahead—and the approach that got you up one mountain won't get you up the next.

The numbers tell a stark story:

• Only about 1% of seed-funded startups ever reach unicorn status

• 2023 saw just 143 new unicorns globally—a 54% drop from the previous year

• The journey takes a median of 6 years for those who make it

Many promising companies stumble—not because their core idea was wrong, but because their technology couldn't scale with their success.

Our experience in scaling mirrors what LinkedIn founder Reid Hoffman describes in 'Blitzscaling’, where new scale thresholds require a shift in company priorities and operational approaches.

What I've learned is that different phases of growth require fundamentally different approaches to technology. Our journey at loveholidays has traversed 3 distinct phases—startup, scale-up, and now scaling out (market expansion)—with each transition requiring a deliberate shift in how we think about technology.

Phase 1: The Startup – Finding a Business Model

In the startup phase, loveholidays focused on establishing a viable business model. The approach to technology was straightforward:

• Build just enough to validate ideas quickly

• Move fast and cut corners where needed

• Run from a prioritised to-do list

• Rely on a small team of generalists

This approach was perfect for the startup phase. It allowed us to experiment rapidly and find our market fit with minimal resources. But as we succeeded, the limitations became increasingly apparent.

The First Shift: From "Travel Company Using Tech" to "Tech Company in Travel"

I will never forget one of my first meetings with our new CEO, Donat Rétif. He joined just as loveholidays reached its first peak—it had a proven business model but was struggling with its own success.

The startup technology environment had created a complex and opaque system with serious constraints:

• We deployed changes just once daily (never on Fridays)

• High-traffic periods triggered code freezes to avoid breaking things

• Low-traffic periods also triggered code freezes as we didn't know if something was broken

• In one particularly painful instance, leadership scrapped months of work because trading numbers dipped for a few weeks

For developers joining us today, this approach sounds like a nightmare. And from a technical perspective, it was.

Every spike in traffic was a crisis. Each new feature risked breaking something else. Our startup thinking had become our biggest limitation.

That's when Donat said something that became the catalyst for change:

"We need to stop thinking of ourselves as a travel company that does technology and become a technology company that happens to be in travel."

This wasn't just a clever phrase—it was permission to transform how we approached technology. But transformation requires more than permission—it needs a deliberate approach.

Our transition focused on 3 key areas:

1. Balancing Skills - We brought in specialist technical expertise to complement our business domain knowledge, creating teams that could tackle complex technical challenges

2. Defining Shared Principles - We established 5 technology principles (detailed in a previous article) that would guide our decisions and create alignment across growing teams:

   • Think big, deliver incrementally

   • Invest in simplicity

   • Leave things better than you found them

   • Focus on differentiation

   • Technology is a means to an end

3. Adapting Our Structure - We moved from a centralised, project-based approach to cross-functional teams organised around business problems.

This professionalisation was essential for scale. It gave us the foundation to handle increasing complexity while maintaining our ability to move quickly. These fundamental changes, shift in mindset and approach unlocked the next stage of extraordinary growth.

Phase 2: The Scale-up – Unlocking Exponential Growth

Our next phase was all about leveraging the foundation we'd built to handle scale far beyond what was previously possible.

Success brought unprecedented demands:

• More website visitors - we handled more sessions in a month than we did in the entire year before I joined

• More bookings to manage - last year we sent the equivalent of 1 in 17 people in the UK on holiday

• More teams working in parallel - to keep up with our growth we have had to double the size of our technology team—though we try as hard as we can to limit this

With our new approach in place, we achieved results that would have been unthinkable before:

• Nearly 250,000 system changes annually: developers' output multiplied 7x, non-technical teams contributing independently, and AI automating a significant portion

• Zero outages despite January 2024 traffic exceeding our entire 2019 volume—we now routinely load test to 10x capacity each night

Our most mind-blowing achievement? Our search now pre-computes 6 trillion packages per run. We processed 1.4 quintillion packages last year alone—more than all the grains of sand on every beach in the world.

An important lesson from this transition: it's easy to be critical of the past, especially when you weren't part of it. The approach that seemed limiting now was exactly what the business needed in its earlier stage. Each phase requires its own mindset—and it was this understanding that helped us change again.

The Second Shift: From Scale-up Systems to Platform Thinking

As always, our CEO was one step ahead of the business. For a few years, he had been talking about the need to think again about how we operate.

The crucial lesson keeps emerging as we grow: what got you here won't get you there.

We've now entered our third phase—market expansion—moving beyond UK short-haul beach holidays to:

• Multiple countries: Ireland, Germany and Austria, with more to come

• Diverse product types: City breaks, long-haul and other emerging categories

• Various customer segments: Couples, families, and other distinct groups with unique needs

When you do the maths, we're looking at 160 different customer cohorts to serve effectively.

Scaling our previous approach to cover all these cohorts would be impossible. We'd need to clone our teams 160 times over.

Instead, Donat pushed us toward our second major mindset shift: thinking of loveholidays as a platform business.

This is where things get considerably more abstract, and I've learned that creating a shared vision is essential when dealing with this. As someone with aphantasia (no visual memory) I've come to rely heavily on diagrams to communicate these ideas.

One way I have come to think about our platform is as the "car that gets us to the destination"—in this case, the means of serving 160 customer cohorts across markets. The system is the platform that makes this possible without requiring 160 separate teams. Having a version of this that we all understand is key.

Our approach to this transition focuses on:

• Creating a Unified Platform Vision - Building shared mental models across the organisation so everyone sees how their part contributes to the larger system.

• Continuing to Evolve Our Structure - Shifting from domain-specific teams to capability teams that can support multiple markets and products

• Implementing Configurable Systems - Building core capabilities once that can be adapted for different markets without requiring custom development.

This shift is fundamentally changing how teams think about their work—moving from "How do we solve this specific customer problem?" to "How do we build capabilities that empower others to solve many similar problems across multiple markets?"

The Ripple Effects: Beyond Systems to People and Markets

These technological shifts have had 2 remarkable consequences that extend beyond our systems.

Changing Work

The evolution has changed not just our systems but what work looks like at loveholidays.

Juliana's journey exemplifies this transformation. She started in our contact centre, moved to QA, and in doing so gained a deep understanding of customer needs—but the operational constraints meant implementing solutions was painfully slow.

Eugene and Tom, who were reimagining our Contact Centre platforms, recognised the need for someone with her skills to help build the systems in the most effective way. So they invited her to join the Conversational Design team working on Sandy, our AI assistant.

In her own words:

"In CX, it was often about accepting that something was a problem. In Tech, it's more proactive—identifying the right people, explaining the impact on customers, and pushing for better solutions."

Combining her customer understanding with the new technical capabilities allowed her to have a much bigger impact than before—helping us automate 50%+ of chats and 20%+ of voice contacts. She is also now playing a pivotal role in repeating this success as we scale into new markets and languages.

This pattern is repeating across the business. Everyone at loveholidays is now working in technology, whether directly building it or using it to tailor experiences.

Transforming the Industry

Our approach is also starting to change the travel industry itself.

In late 2023, there was a significant shift in how Ryanair worked with Online Travel Agents (OTAs), implementing changes that made it challenging for many OTAs to continue offering their flights.

While some in the industry pursued adversarial approaches, our "technology company" mindset led us to pursue a different path—focusing on mutually beneficial collaboration. We initiated conversations with Ryanair, centered on how we could work together to best serve our shared customers.

These productive discussions revealed alignment in our goals. Both parties recognised the opportunity to create something that worked better for holidaymakers while respecting each company's business needs.

What followed was a true technical collaboration. Our engineering teams worked alongside Ryanair's technical staff, and together we implemented an integration in just 3 weeks. This partnership approach made us the first OTA to establish a new ‘verified’ booking relationship with Ryanair.

This demonstrated to the wider industry how technical partnerships between airlines and OTAs can create value for everyone involved—especially customers. It reinforced our belief that putting customer needs first while finding technical solutions is almost always more productive than confrontation.

Yet even as we celebrated these successes, past experience ensured that we have never rested on our laurels. Because there is only one certainly in travel: you are always approaching another inflection point.

The Next Mountain is Already Visible

You think you've reached the summit, only to find there's another mountain behind it. That's the reality of scaling a fast-growing technology business—a reality I've now experienced multiple times.

The pattern is clear: yesterday's solutions become tomorrow's problems. The technology approach that works for your startup will fail as you scale. The approach that works for scale-up will collapse as you expand across markets.

In a moment of synchronicity, loveholidays was showcased a second time with Google—this time at their Next conference in Las Vegas. A small part of our story was presented in front of 35,000 people, sandwiched amongst the incredible announcements of the latest advancements in AI.

There couldn't have been a more obvious signal of what lies ahead.

As this AI tornado reshapes how work gets done, the next mountain for loveholidays to climb presents a double challenge: we need to operate at a vastly different scale, while simultaneously adopting entirely new tools and approaches to get there.

I believe that the organisational adaptability we've developed through previous transformations has uniquely prepared us for what comes next. Far from the daunting prospect it was 6 years ago, I see responding to this AI-driven evolution as another opportunity for the company to leap forward, and I’m really excited for what happens next and, yes, Papa John's is still on speed dial.

—

Which Technology Mountain Are You Climbing?

Which mountain is your organisation climbing now? And more importantly, are you still using the gear that got you up the last one?

I'm genuinely curious about your experience with these transitions. Did you struggle to scale agile and abandon stand-ups and showcases? Was it hard to convince the business to break apart the monolith? What was the most unexpected lesson you learned?

Share your thoughts in the comments—I'd love to compare notes on how different businesses navigate these challenging shifts. Or if you'd prefer, reach out directly and we can discuss how our loveholidays journey might provide useful parallels for your situation.

Studio Ghibli recently discovered an AI image generator had mimicked their distinctive artistic style without permission, triggering lawsuits and outrage.

But the problem extends far beyond visual art. AI systems are quietly extracting value from writers, musicians, photographers, filmmakers, and academics—often without attribution or compensation.

Last week, I asked Claude to summarise a business book I'd never purchased.

Within minutes, I had extracted years of someone's intellectual labour, research, frameworks, and insights without paying a penny. I then incorporated these stolen concepts directly into my presentation without even remembering the author's name.

I didn't think twice about it.

These incidents are happening with increasing frequency, and the scale is difficult to comprehend. The pace of AI development and integration is outstripping our ability to adequately address the ethical implications.

This quiet "borrowing" has made me profoundly uncomfortable.

We've Been Here Before (But This Is Worse)

Pondering this, I was reminded of the impact when Google started aggregating news content.

Scott Galloway described in "The Four" how publishers like The New York Times could have acted early and limited Google's access to their material, forcing negotiation for content rights. Instead, they capitulated, allowing unfettered access that fundamentally altered media power dynamics.

As a result, publishers are now begging for views with clickbait headlines while Google's value has soared.

But there's a crucial difference between then and now.

Back then, news publishers faced a classic prisoner's dilemma. If one stand-up outlet blocked Google while others folded and allowed access, they'd lose traffic to competitors.

With AI, something far more insidious happens:

• Value extraction occurs invisibly in private conversations

• No attribution trail exists

• No compensation flows to creators

• The extraction occurs simultaneously across all creative domains

This isn't just Google's news aggregation on steroids—it's intellectual property theft at an unprecedented scale.

The Great Creative Heist: How AI Systems Feed on Original Work

Modern AI systems function essentially as advanced autocomplete tools built on a foundation of others' intellectual property.

I've witnessed the significant impact from both sides:

At loveholidays, we've used AI to summarise research, apply frameworks from books we've never purchased, and deliver projects at 100x the pace. The productivity gains are undeniable and transformative.

On the flip side, my brother, who built an audience of over 100,000 Instagram followers based on his artistic journey, is pulling back from the platform. As he put it, "Meta is scraping all of IG to feed their plagiarism engine."

One more creative voice silenced.

One less person moving society forward, making social media increasingly the sole domain of the mental health crisis-inducing bikini-clad and TRT-augmented influencers.

The issue isn't that AI assistance is problematic—these tools genuinely supercharge productivity. The problem is that creators, the foundation for these systems, receive no recognition, compensation, or protection.

This makes little economic sense. I would have willingly paid for the Project Ghibli version of a picture I took rather than effectively stealing it through ChatGPT.

The Warning Signs: Journalism's Collapse as a Harbinger

We've already seen what happens when content creation becomes economically unsustainable.

The collapse of quality journalism serves as a warning:

• Once-respected news organisations now chase clicks

• Sensationalist headlines replace thoughtful analysis

• Listicles and AI-generated content barely pass for reporting

• Quality investigative journalism becomes increasingly rare

This future awaits every creative domain if we continue down our current path. The same forces that gutted newsrooms will decimate art studios, writers' rooms, recording studios, and photography collectives.

Not through direct competition—but by siphoning away the economic value of original creation without compensation.

Ironically, the gap between idea and creation, facilitated by AI, is shorter than ever. We're undermining the creators who drive innovation.

What will future AI models train on if creating original content becomes economically unsustainable?

Regurgitations of earlier AI outputs?

The logical endpoint is a wasteland of derivative works and diminishing originality.

This creative monoculture benefits no one—not even the AI companies themselves.

Beyond Attribution: The Intellectual Property Crisis

At a recent interview at TED on April 11, 2025, Sam Altman was forced to address the elephant in the room—the massive intellectual property crisis brewing beneath AI's impressive capabilities.

His arsey acknowledgement of the issue ("you can clap for that all you want" 9_9) is essential but doesn't go nearly far enough. The solutions being proposed are insufficient for the scale of the problem.

It's not just visual artists suffering. Authors, researchers, and academics who spend years developing unique frameworks now watch as their life's work is casually summarised and repurposed without a single citation.

The business book I had Claude summarise represented thousands of hours of someone's intellectual labour. Yet within minutes, I had extracted the core value with:

• Zero attribution to the author

• No purchase

• No acknowledgement of intellectual debt

This is intellectual strip-mining at an unprecedented scale. We're extracting valuable minerals while leaving behind an uninhabitable landscape for creators.

When authors can't make a living because AI systems give away their core insights for free, who will invest the time to develop the next generation of groundbreaking ideas?

From Attribution to Fair Compensation

The solution to this existential threat must go beyond Altman's modest proposals and simple attribution. We need comprehensive systems that ensure creators receive fair compensation when their work is accessed through AI systems.

Imagine a system where:

• AI companies are transparent about which creative works are in their training data

• Usage statistics track when AI draws from specific works to answer queries

• A percentage of subscription revenue flows to creators based on how often their work informs AI responses

• Users see attribution and direct links to purchase the works they're accessing

Just as the music industry adapted after Napster decimated traditional revenue models, we need new systems that ensure value flows back to creators.

The critical difference? This time, we need to act before the ecosystem collapses, not after.

From Disruption to Sustainable Innovation

We in technology have a pattern of asking forgiveness rather than permission when disrupting established fields. AI is following the same playbook that search engines used—extracting value from content creators while offering little in return.

As a technologist and born contrarian, I see both sides:

• AI tools offer genuinely transformative capabilities that shouldn't be stifled

• But sustainability requires balance

Without it, the same decline that devastated journalism will spread across artistic endeavours simultaneously, leading to cultural impoverishment that makes everyone poorer—intellectually if not financially.

AI companies, content creators, and regulators must collaborate on a fair compensation framework that maintains innovation and ensures creators can make a living from their work.

Taking Personal Responsibility

While systemic change is necessary, individual actions matter.

I'm still going to use AI, but I am modifying my own behaviour by:

• Buying every book I use or summarise even if I disagree with it

• Reading and subscribing to independent creators

• Purchasing art directly from artists

• Attending more live events and eating locally

These small acts of support won't fix the system alone but represent a personal commitment to valuing original creation.

I'm also maintaining my continued resistance to derivative monoculture by refusing to acknowledge Taylor Swift as a thing, never drinking anything from Starbucks and avoiding the Ikea nesting instinct. Cultural diversity depends on both systemic change and personal choices that push against homogenisation.

How Will You Support the Creators You're "Borrowing" From?

Are you comfortable with the invisible extraction happening through your AI tools?

Have you considered how your creative work might feed these systems without compensation?

The journalism industry's collapse should serve as a warning—once the economic foundation for creation crumbles, it's exceedingly challenging to rebuild.

We have a narrow window to get this right before we witness the same collapse across art, literature, music, and other creative domains.

What specific actions will you take this week to ensure creators are compensated for their work?

Will you buy a book you've had an AI summarise? Support an artist whose style you've admired through an AI image generator? Subscribe to a writer whose ideas you value?

The future of human creativity might depend on it.

And frankly, I'd rather not live in a world where the only new ideas come from algorithms feeding on the ghosts of creators who could no longer afford to create.

(Yes, AI assisted in the writing of this article)

"We're buying uSwitch."

After an exceptional year, Neil, the founder of TrafficBroker, had taken the whole company to Las Vegas for our Christmas party. He then casually dropped this bombshell on us all in a meeting in our hotel suite at the Wynn. The comparison site founded by the Queen's cousin had once sold for £210 million but was now being acquired for just £5 million.

This 97% collapse in value created a unique opportunity for us—but was it a catastrophic mistake or a calculated gamble?

Thinking nothing of it that night, I learned to play blackjack (and how to chip flip, much to the croupier's annoyance). I quickly discovered blackjack rewards those who understand probability and make decisions based on data rather than hunches—a principle fundamental to our success at TrafficBroker. I didn't realise that turning uSwitch around using data-driven decisions would define the next decade of my career.

9 years later, that gamble, assisted by data, became the cornerstone of a £2.2 billion transaction—and revealed a pattern of data maturity I've since repeated across multiple businesses, regardless of their starting point.

The £5 Million Question: What Had We Actually Bought?

A few months after we returned from Vegas, Neil asked my friend Paul Ingles and me to examine what he'd bought. It was a spring day when we turned up at the offices in Victoria that they shared with the Telegraph newspaper.

This spectacular location, where they occupied 2 massive floors, was a nod to the previous glories, but its staggering decline meant the entire company was now huddled into a corner. We met some fantastic and passionate people, many fundamental to uSwitch's subsequent resurgence. But our starting point was the source code, and that was like uncovering a technical horror story—dependency injection so convoluted I immediately vowed never to look at it again.

But the true shocker was the data: no one at uSwitch could tell us how much money they were making, not with any precision, not for yesterday, last week, or even last month.

It contrasted to working at TrafficBroker, where we had optimised billions in marketing spending to the microdollar. This precision helped us buy uSwitch and accumulate enough Amex points to fund extravagant trips to Vegas. uSwitch, on the other hand, operated in what appeared to be a complete data blackout.

The 97% collapse in value suddenly made perfect sense.

A consulting firm's report Neil had provided concluded that "competition had caught up, supplier relationships had deteriorated, and the business was overstretched," but these were just symptoms. The root cause was painfully clear to Paul and me: uSwitch lacked the fundamental data capabilities to understand its performance, make evidence-based decisions, or optimise its operation.

The lack of data focus had the following business challenges:

• No sales visibility - Exact figures on switches were unavailable or unreliable

• Approximate profitability - Calculations were based on a manual estimation process

• System fragility - Basic reports would take hours to run, if they ran at all, and could even take the site down

This was our starting point—and over the years, 3 distinct stages emerged in our transformation journey that any business can learn from.

Stage 1 (2009-2011): Finding Our Footing (and Missing Millions)

Our first priority wasn't implementing cutting-edge analytics or flashy machine learning algorithms but establishing essential control and confidence in our fundamental data.

The existing systems couldn't be instantly replaced—there was a business to run. Building a proper data foundation is time-consuming, so the existing data warehouse that powered the limited insight had to be maintained alongside building something new.

We left a mostly contracted team to manage this data as we moved to more appropriate data stores, implemented proper data governance and created reliable reporting systems.

I remember the frustrating contrast to TrafficBroker, where we invested in structured and segmented data so everyone could own a metric and measure their performance. It created a culture of ownership and responsibility. However, that experience ensured we understood that any technical achievement would be meaningless if we didn't address the data, too.

Objective: Understand the Past

Looking back, we focused on three core principles:

• Control - Standardising processes for data capture, storage and access

• Single source of truth - Establishing authoritative data sources for each part of the system

• Statistical thinking - Ensuring we use basic statistical principles to interpret data generated

This foundation-building phase is unglamorous work, and the impact takes time. You have to trust the process, building sources of data and visibility. Each element released some value, but as we completed parts of the migration, the sum of all the work became substantial.

The Breakthrough: Finding Missing Millions

In an early conversation with the CEO, Steve Weller, he mentioned a cheque from one of the energy companies in his desk drawer. A token payment recognising unpaid invoices, for tens of thousands, that he never cashed in. He was convinced they owed us much more and frustratedly told us, "I know they're not paying what they should be, but we can't prove it."

As we reworked the systems, capturing and improving data collection on the website and implementing proper tracking of conversion data, we also started looking at the quality of data we received from energy suppliers.

Along the way, I never forgot that conversation with Steve. So a few years later, when we could finally undertake the painstaking reconciliation of every switch recorded in our system against what suppliers had reported, I was really interested to see what we would find.

I still remember the look of satisfaction on Steve's face when we showed him the missing millions from our supplier reconciliation. It proved he was right to hold out, as nearly every supplier owed us over £100k. But the biggest offenders (including those who sent the cheque) approached half a million pounds each.

This wasn't just a one-time windfall. By implementing systems to monitor these discrepancies in the future, we:

• Improved the margin per switch

• Enhanced the accuracy of our financial reporting

• Managed credit risk more effectively across suppliers

But perhaps most importantly, we created confidence in our data, which enabled the next stage of our journey—moving from just understanding what happened to seeing and reacting to what was happening right now.

Key takeaway: Before building anything innovative, you need absolute clarity on your baseline metrics. Establishing reliable data foundations creates the confidence necessary for everything that follows - you can't improve what you can't measure.

Stage 2 (2011-2013): From "What Happened?" to "What's Happening Now?"

With reliable base data in place, our focus shifted from understanding the past to creating real-time visibility of current performance. This wasn't a case of making reporting faster; it was a shift in how the business could operate, react and ultimately compete using data.

Objective: React to the Present in Real-Time

Achieving this required rethinking how data served different parts of the business:

• Information creation - Not just providing raw data but creating insights for specific contexts

• Multiple views of the truth - we realised that specialist teams need to view the source of truth we had created in different ways

• Meta-products - Investing in creating our own tools to manage and present data appropriately

A simple example illustrating this shift is visualising geographical data: Consider the difference between the London Tube map and an Ordnance Survey map. Both are built from the same source of truth, but each models it differently for different uses. The Tube map helps navigate the Underground at the cost of accurate distances, while the OS map shows actual distances but might not help you plan the most straightforward journey.

Similarly, we realised that different questions in our business required different data models, all built upon our reliable foundation.

The Breakthrough: Dominating Paid Search Through Superior Data

This flexibility in our data approach gave us a compelling competitive advantage in our most important customer acquisition channel: paid search.

Google's advertising platform operates on a modified second-price auction. For keywords like "energy switching," the higher your bid, the more visibility you receive, and the more customers you acquire.

Back then, most comparison companies managed their paid search like other marketing channels—with fixed monthly budgets. When market conditions fluctuate, they often exhaust their budgets before month-end and disappear from the auctions.

Our approach is powered by our data.

We could attribute our revenue to our marketing spend in real-time by connecting transaction data to supplier commercials and historical trends. It allowed us to build an ROI-based bidding system that:

• Calculated the precise value of each keyword and customer segment

• Adjusted bids automatically based on changing market conditions

• Allowed us to manage our budget based on return, not arbitrary limits

Understanding the auction dynamics with this connected data approach, we learned to "burn" the competition. Instead of maximising our own efficiency, we would strategically sit beneath competitors in auctions, bidding high enough to drive up their costs while staying profitable ourselves. This steadily ate away at their budgets until they would disappear from the auctions entirely.

Adam Kirby, our Head of Marketing, who spearheaded this approach, would be delighted as competitors dropped out of the auctions toward month-end. We'd often find ourselves alone in the auctions, scooping up switches at bargain prices while our competitors sat on the sidelines with depleted budgets.

This approach allowed us to capture more market share at a lower overall cost, dramatically improving our profitability. More importantly, it demonstrated the power of connecting different data sources to create real-time actionable information.

Throughout the organisation, teams began building specialised tools and views of our data, each optimised for their specific needs. We created automated P&Ls for our finance teams, experimented with attribution models for our marketing team, and built supplier performance dashboards for our commercial teams.

The organisation became more data-driven, more autonomous, and notably more effective. But the most transformational was still to come—when we stopped "navel gazing" and turned our focus outside our business.

Key takeaway: Once your data foundation is solid, create flexible views tailored to different teams' needs. This shift represents a move to real-time information that helps you understand what's happening, react faster, and outmanoeuvre the competition.

Stage 3 (2013-2018): Influencing Others with Data

In this final and most impactful stage of our data maturity journey, we shifted from looking inward to looking outward. Using our data to influence customers, suppliers, and even regulators in ways that benefit our company. Our data became a strategic asset that created possibilities beyond just internal optimisation.

Objective: Shape the Future with Your Data

This evolution built upon our previous capabilities but required a fundamentally different approach:

• External sharing - Using our data to influence customers, suppliers, regulators, and partners

• Experimentation and research - Investing in more specialisation within the teams creates new levels of insight

• Market insight development - Positioning ourselves as the authoritative source of market intelligence

This approach helped sustain the business model, opened new possibilities for growth through innovation and, most importantly, gave us a real competitive advantage.

The Breakthrough: From Supplier's Vendor to Strategic Partner

At this point, uSwitch represented approximately 50% of the online energy switching market. This volume allowed us to command market-leading rates for each energy switch, creating tension in our supplier relationships. We received feedback that our commercial terms were too high and that volume alone no longer justified the premium.

As part of one sales process, the buy-side advisors conducted extensive interviews with each of our suppliers. The unanimous feedback: our volumes allowed us to command market-leading rates, but they are unsustainable as we provided no long-term strategic value.

We knew we needed to evolve our relationship with suppliers from purely transactional to strategic partnership. Our bet was that our unique view of the energy market provided the perfect opportunity to create this.

We developed "inSights," a data product for energy suppliers that consisted of:

• Access to raw market data that could feed into their models

• Tools and dashboards that allowed them to access market insights efficiently

• Consultancy services where our data scientists conducted bespoke research

The value became obvious when British Gas, the first supplier to sign up, used our market information to more accurately predict churn and adjust their energy hedging strategy. They recouped our fee almost immediately through improved forecasting.

As the product evolved, we signed up most of the big energy suppliers and started to help the plethora of new start-ups. The strategic impact was profound:

• CEOs became regular users and saw our branded insights

• Teams previously required to accumulate partial marketed insight were no longer needed with access to our tools

• We hired the first power user from British Gas, who joined our team to advance its capabilities

• Suppliers became dependent on our data for their planning and pricing

This strategic partnership transformed our position in the market. We were no longer just a sales channel but embedded in their businesses. And this gave us unprecedented influence.

For example, suppliers began using our data to target specific customer segments in different regions with tailored pricing strategies—something impossible before. This more sophisticated approach to the market improved our platform's conversion rates while helping suppliers achieve their acquisition and margin goals.

All this while generating more than 7 figures in direct revenue from the inSights product.

Beyond Suppliers: Influencing Regulation Through Superior Data

Our data maturity also positioned us to influence regulatory decisions that benefited the market and our business.

When energy regulator Ofgem was tasked with setting a price cap, they initially planned to base the level on consumer research. This approach would provide little insight into actual market dynamics.

Emma Bush, the GM of our Energy vertical, took it upon herself to address this.

Her idea was to draw on our detailed understanding of the correlation between savings and conversion rates that we used daily to understand and optimise our business to ensure the market remained liquid. We provided Ofgem with extensive data and models that simulated the impact of different cap levels on switching activity. Showing them how to implement policy whilst retaining an active energy market.

I'll never forget when Emma returned from Ofgem's offices after walking them through our models. It was the moment I fully grasped how far we had come as a business. We weren't just reacting to the market anymore but helping shape it. The same data approach we used to fix our operations began influencing national policy.

Emma had taken what was possible with data even further than I had imagined.

This work:

• Influenced their decision-making through objective, data-driven evidence

• Established a strong relationship with the regulator

• Enabled us to predict the impact of regulatory changes and plan ahead of time

By this point, data wasn't just helping us understand our business or optimise our operations—it was actively shaping our future and the future of our market.

Key takeaway: At the highest level of data maturity, you can leverage your insights to influence external stakeholders and shape market conditions in your favour. When your data becomes a trusted industry source, you gain influence beyond operational efficiency.

From a £5m Gamble to £2.2 Billion Exit: The Three Stages of Data Maturity

Looking back on that decade-long journey, I feel privileged to have been part of it. Working alongside some incredible people to transform uSwitch often felt like chaos, but in retrospect, a clear pattern emerged—one I've since seen repeated across different businesses.

What strikes me most about this framework is that success doesn't hinge on technology—it's about building capabilities in the correct sequence and applying principles consistently without trying to skip ahead.

The foundation stage gives you control of what actually happened, not what you think happened. With that confidence established, the flexibility stage connects teams with views of the same reliable data to react in real time. Finally, the influence stage—which few companies ever reach—lets you shape external decisions from suppliers to regulators.

The parallels to that night in Vegas, where this journey began, are impossible to miss. In blackjack, amateurs rely on hunches and may enjoy short-term wins, but probability eventually catches up. Professionals understand the system, calculate odds methodically, and make decisions based on data patterns rather than feelings.

Business works the same way. Amateurs make gut decisions that sometimes pay off, creating the illusion of skill. Professionals build systems that understand the past, clarify the present, and ultimately shape the future.

That £5 million gamble in Vegas eventually led to a £2.2 billion transaction—an extraordinary value creation story. But like a winning blackjack strategy, the principles behind this journey aren't magical or mysterious—they're systematic and replicable.

A note of thanks: This transformation was a massive team effort with far more stories than I could share. I've missed acknowledging so many brilliant people who contributed, but special mention to Tom Lyon for teaching us negotiation, Ewan for his Shuffle, Matt Whitaker's labour of love, Paul Lamb and Rachel, who built the first inSights PDF, Tickle Tiger for help at E.ON, and Carl Gaywood and Neil Hutchinson for their guidance and support. These represent just a fraction of the talent behind this journey.

In a world of increasing AI automation, those who use AI to replace their thinking will themselves be replaced—those who use it to enhance their curiosity and knowledge will thrive.

I fell into this trap when crafting a blog post with AI assistance, spending hours refining my prompts to get the best output rather than doing the thinking and writing. When I shared the article with my friend Jo, who has taken on the role of calling me out on my bullsh*t, she told me, "This totally lacks any personality. I can't read it." She was right, the result was an obviously AI crafted sanitised, generic and corporate version of me—technically correct but missing the personality and perspective that made it readable.

AI tools promise to amplify our capabilities, but depending on how you use them, they risk replacing—rather than enhancing—your thinking. Instead of making us smarter, they can make us intellectually lazy and it's getting increasingly obvious.

The consequences are everywhere, but perhaps LinkedIn is worst of all. My feed overflows with the same generic self-aggrandising guff, full of rhetorical questions, each post eerily similar to the last. We risk training ourselves to think in AI-optimised patterns, using it to think for us, rather than developing our own skills and insights.

But it doesn't have to be this way.

I have the privilege of working with some genuinely brilliant people and, from them, I've identified and incorporated 5 approaches that allow me to harness AI as a thinking partner rather than outsourcing my cognitive work. These approaches don't just preserve my voice—they've actually expanded my thinking capabilities and accelerated my learning in unexpected ways.

1. Refactor, Don't Replace Your Thinking

The first and most fundamental approach is about maintaining your role as the primary thinker.

George, our VP of Engineering at loveholidays, coined the term "thought refactoring" when I described how I used AI. The idea is simple but profound: don't use AI to replace your thinking; use it to adapt and improve your output.

What's crucial to understand is that AI works best when refining your thoughts, not generating them. You can add all the context you want, but AI will still hallucinate, make things up, and try to make you sound like a generic LinkedIn marketer if you're not careful.

I've found this refactoring approach particularly valuable when applying concepts I've learned to content I've created. For example, I use AI to apply Nicolas Cole's principles to my blog posts and the Pyramid Principle to help present my ideas at work. Both of these help to structure ideas more effectively whilst maintaining my voice.

This approach ensures I remain the architect of my thinking while AI serves as a valuable collaborator in expressing those thoughts in a way that helps others follow my argument.

2. Show Your Workings (And Encourage Others To Challenge Them)

Having established yourself as the architect of your thinking, the next step is being transparent about how AI is involved in your process. This transparency doesn't just build trust, it actively promotes better thinking practices by inviting critique and collaboration.

At loveholidays, this transparency is becoming part of best practice. When I create documentation or suggest blog content, I attach the prompts I've used as a code block in a tab—I got this from working with Tim, our CDO. This demystifies the process and helps others use AI to improve their own work.

This approach mirrors what we did with our technology principles when scaling the business—being transparent about our methods created alignment and trust across teams.

The prompts we use to get feedback on blog posts are now shared around our team. It's so much quicker to provide feedback that is effective and based on a framework. This should be seen as just a first step, a way of improving your first draft, it's still vital to get feedback from actual humans.

What I love, though, is when people disagree with the output and put their own approach back into it. This creative tension between AI suggestions and human judgement keeps our thinking sharp.

3. Use AI To Expand Your Range (And Connect Unlikely Domains)

Whilst transparency keeps your thinking authentic, deliberately expanding your knowledge range is what makes your output uniquely valuable in a world of AI-generated content.

Years ago I read "Range: Why Generalists Triumph in a Specialised World" by David Epstein, which explores how connecting ideas across different domains leads to innovative problem-solving. Epstein argues that people who develop broad knowledge across multiple fields ("generalists") often have unique advantages in solving complex problems. They can apply analogical thinking, transferring solutions between entirely different domains, which specialists might miss due to their narrower focus.

AI can dramatically accelerate this generalist advantage.

My colleague Joe uses AI to explore multiple domains simultaneously—writing bedtime stories for his kids, developing trading strategies, and creating computer games. He's not just getting AI to do these things; he's using AI as a learning tool that helps him develop knowledge across these fields simultaneously.

I see AI as enabling you to be "an inch deep and a mile wide". The more you understand about different domains, the more connections you can make, and the more effectively you can harness AI's power.

A somewhat unconventional example: I've been using AI to explore Jungian psychology, an interest sparked after reading Jung's biography years ago. Recently, I decided to explore shadow integration work through AI. By providing context from my own experiences and writings, I began a fascinating learning journey. One which is helping me see patterns, identify blind spots, and stop getting in my own way—not only personally, but professionally.

Interestingly, I wasn't just getting answers—I was learning to ask better questions. As my understanding deepened, I found myself reading articles and books by actual humans to take my knowledge to new levels. AI has become a springboard for genuine exploration rather than a replacement for it.

4. Free Up Your Cognitive Bandwidth By Creating Specialised Agents

As your exploration across domains deepens, the cognitive load of managing these diverse interactions can become overwhelming. Plus, there's a practical limitation: continuing AI conversations in the same chat makes them painfully slow and more expensive as the context grows.

Creating specialised AI "agents" for recurring tasks solves both problems—it frees up your mental bandwidth for creative thinking that matters most whilst keeping your AI interactions fast and focused.

I borrowed this approach after seeing how our research team provides instant access to user feedback through a Slack channel connected to an AI agent. This elegant solution inspired me to create my own version when I noticed I was repeatedly using AI for the same types of tasks.

With Claude, I've started using the Projects feature to build purpose-specific agents. For blog writing, I've created an agent with context from my previous writing, examples of my style, and instructions on how I like to collaborate. Each time I need to work on a post, I start a fresh chat within this pre-configured context, allowing for quick iteration while maintaining continuity.

My "blog post" agent now contains examples of my writing style, formatting preferences, and typical structures, making each new draft increasingly aligned with my voice. Over time, I've refined this approach with each post, creating an evolving partnership that improves with every interaction.

With the time you save using these agents, invest in learning to direct AI better—improving your prompting becomes a valuable meta-skill that compounds over time. This isn't just about efficiency; it's about freeing up your brain space for the creative thinking that AI can't replicate.

5. Test Against Multiple Models to Triangulate Deeper Insights

With your cognitive bandwidth freed up, you can elevate your thinking by comparing insights across different AI models—creating a form of cognitive triangulation that reveals your own, often hidden, assumptions.

The current AI systems available each offer differing capabilities, and, after experimentation, I now go to each for different tasks:

• Claude, if I need help or feedback with writing

• ChatGPT for more technical support

• Grok I discovered, can provide more current information

• Gemini, well, it creates podcasts…

However, these AIs evolve rapidly, with models leapfrogging each other in competency with every release. As a consequence, it's worth not just experimenting but also rerunning previous queries after upgrades.

I also find real value in comparing responses across models. The differing approaches or solutions from the AI system can reveal hidden assumptions in your thinking. This cognitive triangulation helps uncover blind spots that wouldn't be apparent from a single perspective (I also find Claud has a better manner as a Jungian Analyst).

Regular experimentation with different models prevents dependence on one AI's particular "style" and pushes you to think more critically about the information you receive.

From Artificial Intelligence To Augmented Insight

In a world where AI threatens to replace human thinking, these 5 techniques share a common goal: ensuring AI enhances rather than replaces your unique perspective—preventing what you produce from feeling artificial.

This distinction reminds me of rewatching 'Aliens' recently. As a kid, I was mesmerised by what seemed like incredible visual effects in the hangar scene. Watching now, I immediately spotted it was just a bloody painting—something I'd completely missed before. What once amazed me now feels transparent and obvious.

This mirrors our rapidly evolving relationship with AI content. What initially impressed us quickly becomes recognisable as artificial once we develop the eye for it. And this recognition is happening faster than ever because we're exposed to so much AI-generated content daily.

Just as I felt slightly embarrassed for not recognising the painting for what it was, we risk looking back at our AI-dependent thinking with similar discomfort if we allow it to replace rather than enhance our cognitive processes.

The goal isn't hiding AI's involvement—it's using these tools so skillfully that they extend your natural thinking abilities rather than act as a substitute for them. In a world increasingly defined by automation, your curiosity and knowledge acquisition become your most valuable currencies. The most successful AI users aren't those who let machines do their thinking, but those who use AI to expand their horizons.

When someone says, "Wait, you used AI for this?" it should be because the insights are so novel and the thinking so advanced that they can't imagine technology being the primary driver. You aren't replacing yourself with AI—you're enriching yourself through it.

Ironically, if you aren't doing this it won't be AI that replaces you, but the humans that are.

Thankfully, both Jo and Claude told me this piece has personality. However, I'm still experimenting with using AI as a thinking partner rather than a replacement, and I'd love to hear your experiences:

• Have you discovered approaches that help you maintain your authentic thinking while leveraging AI's capabilities?

• What unexpected ways has AI expanded your thinking rather than just saving you time?

• Are there dangers to our cognitive capabilities that I've missed?

Please share your thoughts in the comments :-)

Thanks to Tim Bowker, Joe Francis, Georgia Gaskell, Jo Gostling, Luca Grulla and George Malamidis for inspiration and help with this post.

I learned this lesson the hard way.

It was 2017 and we had acquired another business, and as CTO I was responsible for planning the integration. I poured all my knowledge into a dissertation-length proposal, referencing nuanced management frameworks, detailed architecture proposals and suggested operating models. In the review meeting, instead of joint appreciation of just how clever the proposal was, the C-suite met me with blank stares — they hadn’t even read past the first page.

It was a wake-up call.

The higher you climb, the more your success depends on effective communication and increasingly, written communication (thanks Jeff Bezos 9_9).

Communicating your ideas and concepts to inspire action is a specific skill. Like many technical leaders, I was still writing to pass exams rather than to drive change — and it was limiting my career. It wasn’t until I watched Larry McEnerney’s lecture on effective writing that I understood why.

Adapting his insights to business writing, I have developed a 2-step process, which I now use AI to accelerate.

The Problem with Academic-Style Writing

One of McEnery’s key insights is that we learn to write in school, where our teachers are paid to read our work. Their job is to check for the completeness of our understanding. This conditions us to expect our writing to be evaluated rather than truly read. But this explanatory approach isn’t just inappropriate in business — it’s actively harmful.

Your CEO doesn’t need another evening filled with reading your technical exposition. They have not hired you to prove your technical knowledge but to act based on it.

Your engineering team doesn’t want to wade through pages of context they already understand. They need clear direction and compelling reasons to act.

So, if academic writing style is the problem, what’s the solution? It starts with recognising that writing serves 2 distinct purposes: thinking and influencing. By separating these purposes, we can keep the benefits of detailed analysis without subjecting our readers to it.

Step 1: Write to Think

The first phase of technical writing is all about your understanding. Writing isn’t just a way to record pre-existing, fully-formed thoughts and ideas but an active part of the thinking process. It is where we discover and develop what we actually think.

When tackling complex technical concepts, you need space to explore, connect, and validate your understanding. Traditional long-form documentation and structured thinking approaches serve this purpose:

• Create comprehensive descriptions of challenges and rationale

• Document all relevant reference materials

• Link to related concepts you have acquired in your notes system or second brain

• Explore edge cases and technical implications

This initial work is essential. It forces you to understand your ideas, test them and uncover gaps in your reasoning. It’s the foundation of technical clarity — but it’s only half the journey.

Step 2: Writing to Influence

Here’s where most technical writers stumble. We take our exploration and thinking and simply share it, assuming others will invest the same energy in understanding it. They won’t.

I have learned that if you want people to read what you write, and change how they see things, you need to transform your thinking into a format and message tailored to your goal.

The process requires 3 key elements:

1. Know your Audience: Understand their pains, technical depth, and decision-making process because influence comes from addressing real needs, not showcasing knowledge.

2. Craft your Narrative: Frame solutions through the audience’s challenges, in their language, because technical brilliance only creates value when others can see it.

3. Drive Clear Action: Make next steps crystal-clear and define what success looks like because vague recommendations rarely drive change.

The ruthless editing is the hardest part. This transformation from comprehensive explanation to focused influence feels like sacrificing clarity (and killing your darlings) for brevity. It’s a painstaking but essential process.

After struggling with this challenge for years, I started to use AI as an objective editor, helping maintain technical accuracy while reshaping content for maximum impact.

The AI-Enhanced Technical Writing Workflow

The key to leveraging AI in technical writing isn’t to replace your thinking –it’s to accelerate the transformation from understanding to influence. Here’s my process:

• Select a writing framework that is most appropriate for the content, audience and business

• Create a prompt designed for AI to help extract and reshape the content

• Apply the prompt to transform my comprehensive documentation from step 1

Here’s a prompt I use that combines 2 consulting stalwarts frameworks: Barbara Minto’s Pyramid Principle and the SCQA (Situation-Complication-Question-Answer) framework.

You are an expert technical writer tasked with transforming complex technical documentation into a clear, audience-appropriate summary. Follow these steps:

1. First, analyze the audience profile provided and note the key characteristics that should influence the summary's content, complexity, and tone.

2. Use the SQCA framework to structure the initial analysis:
- Situation: Identify the current context and background
- Question: What problem or need does this documentation address?
- Complication: What challenges or complexities are involved?
- Answer: What are the key solutions or recommendations?

3. Then, apply the Pyramid Principle to organize the information:
- Start with the main conclusion/message at the top
- Group related supporting ideas into logical categories
- Ensure each supporting level connects directly to the level above it
- Present ideas in order of importance within each group

4. Adapt the technical complexity based on the audience analysis:
- Adjust terminology and explanations to match audience expertise
- Include or omit technical details based on audience needs
- Use appropriate examples and analogies for the audience's background

5. Provide your summary in this structure:
- Executive Summary (1–2 sentences using the Pyramid Principle's top-level message)
- Key Points (using SQCA framework)
- Detailed Breakdown (organized using the Pyramid Principle)
- Next Steps or Recommendations

Please format the summary accordingly and ensure all technical content is appropriately calibrated for the specified audience.

The output isn’t the final product — it’s a starting point for crafting more impactful technical content. Think of it as an intelligent first draft of your transformation from exploration to influence.

It will still need a lot of work, but that objective reduction can help clarify what’s important. It also dramatically speeds up the process.

Making Technical Writing Matter

Technical expertise only creates value when it drives change. By separating the process of thinking from the act of influencing, we can create technical content that doesn’t generate those blank stares. Leveraging AI can help accelerate this time-consuming but vital process of bridging between the two.

The next time you start a technical document, ask yourself: am I just sharing my thinking, or is what I am sharing specifically created to drive change?

Using this process with our VP of Engineering allowed us to halve the length of a document in just a few minutes while strengthening its message. Afterwards he quipped on LinkedIn: “We’ve deleted a lot of code in our careers with the help of technology, but this was the first time we’ve deleted a lot of words.” He’s right — sometimes, knowing what to remove is the most valuable technical skill.

Remember: The measure of technical writing isn’t in its length or comprehensiveness. It’s the changes it generates.

Thanks David Annez, Hugo Brook, Georgia Gaskell, Jo Gostling, and George Malamidis for helping to review and improve this article 🙏

Keeping on top of the constant deluge of information, tasks, and meetings you experience as a CTO requires adopting and maintaining a structured approach. Having said that, as someone whose organisational style is best described as “contemporary abstract”, I’ve had to work at it. My saving grace? Note-taking.

Over the years, I’ve experimented with countless methodologies. When the Apple Pencil was released for the iPad, I was excited by the benefits of moving my notes to a digital format. After spending an embarrassing amount of money on various productivity apps, I learned that the process matters more than the tool.

By refining how I take notes rather than chasing the perfect app, I’ve developed a sustainable system combining the benefits of handwritten notes, digital organisation and task management. It helps me stay focused and productive in a fast-paced environment.

I’m often asked about this, so I thought it was fitting to write something about how it all works.

Tools and Setup

The iPad and Apple Pencil are the perfect tools for note-taking. I use the 13" iPad Pro because I like more space for drawing and writing, and it means I can easily take advantage of the split screen format to take notes. Attaching a full-size keyboard and using it more like a laptop is handy for everything outside of notes.

I’ve tried and tested a long list of apps with the aim of getting the best out of the iPad as a productivity tool, and I have learned that simplicity often trumps that unique feature. For my purposes, the built-in Apple Notes and Reminders apps are more than sufficient for effective note-taking and task management.

Notes: Simple integrated note-taking

Apple Notes is where I capture and store all my notes from 1–1s, interviews and strategic discussions. Its simplicity kept me coming back, but it also has some extra features that make it really useful:

• Attaching files: PDFs, pictures and scanned documents can be added and scrawled over

• Search: The ability to search across written notes by hand allows you to find that one bit of information you can’t remember

• Corrections: my handwriting is often incomprehensible and my spelling pretty poor so it’s great to have the same blue dotted lines to click on to address both issues

Organisation

I’ve adapted various daily note-taking methods to work in a digital format and this includes how I keep those digital notes organised. I create a simple hierarchical structure that makes it easy to navigate through my notes chronologically while maintaining clear organisation.

My structure has evolved to look like this:

• A main folder for my daily notes

• A readme folder containing a summary of the techniques and tools I use

• A folder for each year that includes a folder for each month

For each month, I create the following files:

• A calendar file that contains a list of the days of the month and all the significant events or deadlines, which I pin to the top of the folder

• A note file for each day in the format 04.12.WED for Wednesday 4th December, making it easy to sort by title in a descending order and the day name is helpful to quickly find the notes for a specific day

Reminders: Streamlined Task Management

Apple Reminders is a surprisingly overlooked app that provides everything I need to track my tasks and what others have committed to doing.

• Contextual Inputs: Quickly add tasks with natural language (e.g., “Remind me tomorrow at 9 am”).

• Integrated Workflow: Attach links, images or meeting notes directly to tasks, ensuring follow-through on action items.

• Customisations: Tags and smart lists help me filter tasks by team, priority or project.

Setup

My setup is a single list, configured using the column view to create a basic Kanban board. I use this list to capture everything related to work and try to incorporate the best practices from ​​David Allen’s “Getting Things Done” (GTD) method by ensuring:

• Each task starts with a verb and describes a specific physical action.

• I break down complex items into subsequent physical actions.

• I include all relevant context, such as links to docs and deadlines.

I use tags to assign tasks to people or meetings to quickly find them when I’m in a 1–1 or checking in to ensure the relevant person follows through.

Why Stick to Defaults?

The real advantage of these apps is their integration with the Apple ecosystem. Features like Siri voice commands, cloud sync and widget support allow me to capture and retrieve information, whether I’m at my desk, commuting or between meetings. While there are apps with some advanced features, I found that chasing the “perfect tool” became a distraction from what matters: staying organised and focusing on outcomes.

Making Your iPad Work-Ready: Profile Setup

Creating a dedicated work profile is essential for using your iPad for note-taking. You can configure it to remove distractions to help when using it as a productivity tool.

Some of the key settings are:

• Work focus is set up to only allow notifications from company communication apps and calendar alerts.

• Adjusting the home screen layout to show only work-essential apps

• Hiding all social media and entertainment apps

• Set up custom widgets to show my upcoming meetings, priority tasks from Reminders, and quick access to my most recent notes.

I set this profile to activate when I arrive at the office and during scheduled business hours, helping me maintain boundaries between work and personal time while ensuring I have the right tools when needed.

The Argument for Handwritten Notes

The foundation of my system is handwritten notes. While typing might seem more efficient, research shows that writing by hand forces your brain to process information differently. When you write notes by hand, you can’t write fast enough to capture everything verbatim — which is why it’s so important. This constraint means you must listen actively, understand what is being said and reframe it in your own words.

I’ve found this particularly valuable in meetings and workshops. Instead of mindlessly transcribing, I’m actively engaging with the content, drawing connections and creating more meaningful notes. This process leads to better retention and understanding.

Whilst it’s beneficial to write by hand, it is more time-consuming, necessitating using a couple of methodologies in different situations to make this approach practical.

Rapid Logging: The Heart of Daily Note-Taking

The core of my note-taking system is rapid logging, a technique borrowed from “The Bullet Journal Method” by Ryder Carroll. Throughout the day, during 1–1s or update meetings, I quickly capture information using a simple set of symbols and short phrases.

This approach lets me stay present in meetings while ensuring nothing important is missed.

My rapid logging syntax includes:

• • for notes and observations

• ○ for tasks and actions

• △ for important ideas or insights

• ★ for priority items

• → for delegated tasks

The beauty of rapid logging is its speed and flexibility. There’s no need to write in complete sentences or worry about organisation. The goal is to capture information quickly and clearly, leaving the processing for later.

Cornell Method: Capturing longer form meetings

When I am attempting to learn something new or following someone else’s thought process, I use the Cornell method. Implementing the Cornell Method divides each page into 3 sections:

• Notes Section (Right, ~70% of page): The main area for rapid logging and note capture

• Cue Column (Left, ~30% of page): Space for questions, key points and action items

• Summary Area (Bottom): A brief synopsis of the page’s key takeaways

• To make this easy in the notes app, I drag a transparent grey box for the left section and resize it as needed while I take notes.

This approach is particularly beneficial in interviews when someone is presenting. It allows me to actively listen and collect questions as I go along to ask at the end rather than interrupting their flow.

Daily Workflow: From Capture to Action

Making this system work requires dedication and routine. To help with this, I set 2 non-negotiable events that bookend each day. They provide the structure that makes the rest of the system work:

Morning Review (5 minutes):

• Check the Reminders app for today’s tasks

• Review yesterday’s notes for any missed items

• Scan upcoming calendar events (so I know what I am going to be late for)

• Set priorities for the day

Evening Processing (10 minutes):

• Review all “rapid logged” items

• Complete any outstanding tasks that are under 2 minutes

• Transfer open tasks (○) to Reminders app

• Add context and due dates to tasks

• Mark completed items

This systematic approach ensures that actions don’t get lost and provides the time to ensure tasks have all the context to make them easy to action. Also, ensuring that I have few priorities at the start of the day ensures I can be intentional about what I work on.

Results and Benefits

Since implementing this system, I’ve experienced:

• Better retention and understanding of information

• Increased meeting productivity

• More reliable action follow-through

• Easier access to past information and ideas

• Reduced stress about missing important details or tasks

The combination of handwritten notes, rapid logging, structured organisation and separate action tracking has created a reliable system that supports both day-to-day productivity and long-term knowledge building.

I also like to browse through all the old notes occasionally. With the pace of change in a technology business, it is fascinating to look back and remember what was happening.

It’s also satisfying to see how an idea has evolved. I recently gave a presentation about the future of our technology platform. Using my notes, I reminded myself of how the ideas and visualisations have evolved over time and some of the conversations we had to get there.

Getting Started

If you want to implement a similar system:

1. Start with daily notes and rapid logging — don’t worry about perfection

2. Establish your morning and evening processing routines

3. Gradually add organisation elements as you see the need

4. Start with the default apps and experiment with others based on different use cases

5. Be patient and keep it simple — it takes time to develop a sustainable system

Remember, the goal isn’t to create a perfect note-taking structure but rather one that supports your work and helps you think effectively. Start small, iterate based on what works for you, and gradually build a complete system that serves your needs.

The most important thing is consistency. Even a simple system you use regularly is better than a complex one you abandon. Start with the basics of rapid logging and daily processing, then build from there.