.png)
Why Engineering Is No Longer the Bottleneck
June 26, 2026
Ask most CTOs and Tech Leaders about their AI journey and you will get one of two responses. Either cautious optimism hedged with caveats, or breathless enthusiasm that does not quite match what is actually happening inside their teams. Julien, CPTO at Lottie, gives you neither. What he gives you instead is something rarer: a genuinely honest, genuinely exciting account of what it actually looks like when a technology team embraces AI not as a strategy, but as a way of working.
Lottie is one of the most compelling businesses in UK tech at the moment. Its marketplace, Lottie.org, helps families find the right residential or home care for loved ones, solving a problem that is deeply human, often urgent, and historically underserved. Alongside the marketplace, Lottie has built a full SaaS platform that care providers use to run their businesses: CRM, occupancy management, billing, or voice agents. That combination is what makes Lottie architecturally interesting. The two sides of the business reinforce each other, giving Lottie access to operational data, such as availability, occupancy trends, and billing complexity, that a pure marketplace could never have. Backed by General Catalyst and Accel, with a team of around 70 and a tech and product organisation of 30, Lottie is at an inflexion point. And Julien, who joined just over a year ago following his time as CTO at Zoe, the nutrition science company, has been at the centre of it.
When Julien arrived, the team's AI usage looked like most companies': a ChatGPT subscription, some experimentation, nothing systematic. Within a year, it looks completely different. The whole engineering team moved onto Cursor, then switched to Claude Code when it became clear that was the right next step. They built their own orchestration layer on top of it, with custom plugins and multi-agent workflows. And rather than keeping that infrastructure inside engineering, they extended it to customer support or marketing.
That last point matters. The most common pattern with AI adoption is that engineering moves quickly and everyone else gets left behind. Lottie is working on avoiding that. The ambition from the start was to make AI a
Be Bold: Innovate Fearlessly
shared capability across the business, not a technical specialism that sits in one team and generates resentment everywhere else.
“It has been a massive discovery process. The teams that win are the ones that find the lever that is right for them and then iterate relentlessly from there.”
One of the most interesting things Julien says is that he is not sure a formal AI strategy is actually the right starting point. The landscape shifts faster than any plan can accommodate. A new model drops, and overnight, a whole category of problems that seemed hard becomes simple. You cannot predict that. What you can do is stay close to the work, move quickly when something promising appears, and build the organisational habits that let you capitalise on step-changes when they arrive.
The practical implication, for Julien, is that you start where momentum already exists. Lottie had a leader in customer support who was genuinely enthusiastic about AI. That became a natural seed outside engineering, and from there, the capability expanded. The insight is not complicated, but it is counterintuitive for organisations used to top-down transformation programmes: let energy lead, and structure follows.
He is also clear that token budgets matter less than people think. Give everyone the tools and the access to experiment freely. Not every interaction will produce something useful, but every interaction builds understanding, and understanding compounds.
The most striking observation Julien makes is about where the actual constraint lies in most AI adoption efforts. It is not resistant engineers. It is not legacy infrastructure. It is the leaders themselves. If you are running a technology organisation but you have not been close to the code for years, you have no reliable way to separate genuine blockers from learned helplessness. You cannot tell whether "it doesn't work" means the technology is genuinely limited or that the team has not yet found the right approach.
Julien's response to this was to go back and write code himself. Not to ship anything on the critical path, but to rebuild his own understanding of what this new world actually feels like from the inside. The returns were significant. He could see, from direct experience, what was genuinely hard and what his team was underestimating. He could challenge assumptions with conviction. And critically, that conviction transferred. Teams who know their leader has first-hand knowledge are more likely to push harder and try things they might otherwise write off.
“The bottleneck to adopting AI is people like me. Until you start doing something yourself, it is really hard to understand what is possible.
Here is where the conversation becomes genuinely energising, because Julien's experience does not fit the pessimistic narrative about AI and the future of engineering careers. It fits a much more interesting one.
The engineers who are thriving at Lottie are not the ones with the deepest narrow specialisms. They are the ones who genuinely care about what they are building, who think about outcomes, who bring curiosity to the product as much as to the code. In a world where AI handles more of the implementation, the premium shifts to judgment, taste, and product instinct. Those are human qualities that become more valuable, not less.
"First-line engineering managers have also adapted brilliantly. They had already built the skill of managing outcomes without needing to control every detail and embracing AI agents turned out to be a natural extension of how they already worked. The muscle they had built managing human teams transferred directly to directing AI."
For engineers who found the transition harder, Julien's approach was thoughtful and generous. Artificial constraints worked well: give someone a project with a timeframe that makes the old approach impossible, and ask them to solve it without writing a line of code themselves. The discomfort is real, but what comes out the other side is real too, a new set of capabilities and, often, a renewed enthusiasm for the work.
Perhaps the most vivid illustration of how much has changed at Lottie is in how the team now thinks about product delivery. The traditional model, specify, design, build, assumed that writing the spec was the essential first step. Julien's view has shifted. At the pace Lottie now moves, the better approach is often to build a working version first, learn from real usage, and refine from there. A written spec and a working prototype are both going to be roughly eighty percent right. The difference is that one of them is easier to test with team members or users.
The result is a much tighter and more alive delivery loop. Product managers are writing code. Designers are contributing to the codebase. Engineers are making product decisions without formal handoffs. The boundaries between roles are blurring, and in the right culture, that is not a source of anxiety. It is a source of energy.
“Engineering is not a bottleneck anymore. We have shifted the constraint to another part of the business entirely. That is a genuinely exciting place to be.”
The practical result: Lottie is shipping features faster than its care provider partners can onboard them. That is a remarkable thing to be able to say. The team is smaller than it was a year ago, but the output is greater. And the constraint has moved from engineering to the wider business's ability to absorb and act on what the technical team can now produce. That is a genuinely new problem, and a much better one to have. There are also new challenges: because everything is an option, or you are a few prompts away from demoing something new, it’s easy to get lost in side quests instead of focusing relentlessly on the big needle movers.
Because Lottie operates in the care sector, working with families in often emotional and urgent circumstances, handling sensitive financial and personal data, Julien is thoughtful about where AI sits in the platform. The fundamentals of good data practice have not changed: opt out of model training, ensure data residency is correct, put appropriate access controls in place. What has changed is the speed at which mistakes can compound, which means the guardrails need to be embedded earlier and more systematically.
The team has invested in automated code review that catches a wider range of issues than manual peer review would. And the models themselves are increasingly helpful here. Even by increasing pull request velocity by 10x, the number of incidents has remained stable over time. Julien recounts a moment where an engineer asked Claude to help them game an internal usage leaderboard. Claude declined immediately. It is a small thing, but it points to something significant: the ethical scaffolding built into these models is genuinely useful in sensitive environments, and it is improving.
For any leader at the start of this journey, Julien's counsel is practical and optimistic. Take a week away from the business and build things. Whatever interests you, whatever problems you want to explore. Do not overthink it. Then take another week and take that energy into the parts of your organisation where people say it cannot be done, and try things there too. No strategy document, no vendor workshop, no leadership offsite will give you what that fortnight gives you: a calibrated, first-hand sense of what is actually possible and where the real edges of the technology lie.
“Take a week and just build things. Then take another week and build inside your business. That will give you more insight than any strategy document.”
The broader message from Julien's year at Lottie is an encouraging one. This is not a story about displacement or disruption in the negative sense. It is a story about a team that found a better way to work, that moved a technical bottleneck that had always held them back, and that is now operating with a level of ambition and pace that simply was not available to them before. The engineers are doing more interesting work. The product people are closer to the product. The business is building things for its partners faster than its partners can use them.
That is what the right kind of AI transformation actually looks like. And Lottie is doing it.
Adam Kinder | Co-Founder Gathered & Found
Join our newsletter and stay ahead!