Six lessons from building Gradient Labs

Rethinking what it means to start a company

I never thought I’d start a company. I even used to joke with a former colleague that we were both too rational for it. In my mind at the time, to be successful, founders needed a level of conviction that bordered on delusion. It felt completely at odds with the way I make decisions.

Of course, the funny part is that both of us eventually did start companies. He founded one of the most successful HR tech companies in London. I went on to start Gradient Labs with two people I had worked alongside for years at Monzo.

Looking back, the decision to start this company did not feel like a leap of faith at all. It felt like the natural outcome of a series of very practical observations about the challenges of customer operations work in Financial Services and how AI had evolved in a way that could solve them. Nothing about it felt reckless. If anything, it felt like irrational not to do it.

Now we’re two years in. Here are 6 lessons from the journey so far:

1. Find something inevitable

Gradient Labs emerged from several factors coming together at once.

The first being that my co-founders and I had seen, day after day, how difficult it was to scale high quality customer operations. We understood the complexity, the edge cases, and the regulatory pressure that shaped every decision.

At the same time, AI models were reaching a point where they could take on parts of that work with real accuracy. The technology was improving quickly, and we could already see how it would map onto the problems we knew well.

And we had something very rare. A founding team that had already worked closely together for years. We knew how to build. We knew how to manage quality. We knew how to operate in an audited and regulated environment.

Once all of these pieces lined up, starting Gradient Labs stopped feeling like a risk and started feeling like an inevitability.

2. Hire people who can tell you what to do

My hiring philosophy changed the day I interviewed Neil Lathia, who is now my co-founder and Gradient Labs’ CTO.

At Monzo, we were looking for our first machine learning engineer. I knew what the organisation needed, but in that first conversation I realised something much more valuable. Neil had expertise that I did not have. He could do things I could not do. I left the meeting feeling that I would learn from him, not the other way around.

That was the moment I understood the kind of people I wanted to hire. Especially in leadership roles. If I find myself telling someone exactly what to do, they are probably not the right hire. The right people take ownership of their domain and bring new insight that shapes the direction of the work.

Gradient Labs is built around this principle. We trust leaders to run entire areas without waiting for instructions. That creates a level of pace and quality that would be impossible any other way.

3. “Boring” experience is an unfair advantage

Many first time founders learn hiring, management, team structure, and quality standards while building their product. It is an exciting process, but it is also demanding. You are figuring out the basics at the same time as you are trying to scale.

My co founders and I had a different starting point. We had spent years building and running data and machine learning teams inside one of the world’s fastest growing banks. We had already learned how to hire well. We had already learned how to create structure and manage teams. None of that felt new.

This is not the kind of experience people talk about in a pitch deck, but it made an enormous difference. It meant we could focus almost all of our energy on the product and the customer problem. We were not spending our time figuring out how to run a team. We were spending our time figuring out how to build something that could handle operational complexity with real accuracy.

This experience may not sound glamorous, but it was one of our strongest advantages.

4. Build what is necessary. Not what’s trending

Years before the term AI agents became common, we were already building something that looked very similar. We did not set out to follow a trend. We were simply building what we believed would be necessary.

We kept asking ourselves a simple question. If a more capable model appears next month, what parts of our system will still matter? That question was a filter. It stopped us from building thin layers on top of someone else’s model that could be replaced overnight. It pushed us toward deeper system architecture and a strong focus on domain expertise.

This is critical in Financial Services. Real value comes from embedding knowledge of fraud, money laundering, onboarding, disputes, and back office processes into the product itself. Traditional software tools support human experts. Modern agentic systems need to act like experts too.

This focus on depth over trend chasing has become one of our key differentiators. And its helped us build something that can adapt to the rapid pace of development in our space.

5. Treat every project like a rocket launch, not a pebble toss

When ideas are abundant, execution is everything. The quality of your execution is one of the few sustainable differentiators in a landscape where technology can become obsolete in months.

This is why we've built our culture at Gradient Labs around a few core pillars:

Deep obsession with quality and domain expertise

Our years at Monzo gave us an invaluable understanding of how to operate in a complex, regulated industry. We didn't have to go through a painful zigzag of deciphering every regulation that applies to our domain. This is critical because modern AI applications aren't just tools for experts; they must be the experts, embodying deep knowledge of finance, fraud, and customer operations to be effective.

A high bar for execution

I have a simple analogy for the team: every project is a rocket launch, not a pebble tossed in a pond. A pebble makes a small splash and sinks; it never takes off. A rocket launch is a high-stakes, meticulously planned event designed to achieve orbit. In AI, a failed launch isn't just a failed project; it means losing a crucial window of opportunity that may never reopen. This mentality fosters an unwavering commitment to excellence.

Pace and nimbleness

A technology that is cutting-edge today might be a commodity in six months. While we need to move incredibly fast, we also need the flexibility to pivot and accept that something we’ve built may need to be re-imagined. This internal culture translates directly into our product.

6. Make your own true north your fiercest competitor

Gradient Labs has performed well in competitive evaluations so far. We have won every proof of concept or bake off we’ve participated in. Yet when I think about quality, I rarely think about competitors. I think about an internal standard that we call the true north.

The true north is the imagined version of the best possible customer experience. It is not what exists today. It is what could exist if every part of the journey worked exactly as it should. When we compare our work to that standard, it becomes clear how far we still want to go, even when the product is performing well.

This mindset helps us avoid complacency. It keeps us focused on continuous improvement rather than small wins. The goal is not to be better than another vendor. The goal is to reach the standard that we believe is possible.

I used to believe that founders needed a kind of productive delusion to start something new. What I have learned is that successful companies often grow out of experience, timing, and a very grounded understanding of a problem.

You need to pay attention to the opportunities your experience has prepared you for. At some point, the rational path becomes clear. And if you are lucky, it may even become inevitable.