In a world buzzing with AI hype, Gradient Labs is quietly building something extraordinary.

Founded by three Monzo alumni, Dmitri Masin, Neal Lathia, and Danai Antoniou, Gradient Labs didn’t start with a flashy launch or bold promises. Instead, it emerged from deep expertise, humility, and a relentless obsession with quality. Now, after raising £2.8M led by LocalGlobe, this early-stage AI startup is tackling one of the toughest problems in tech: how to automate complex, high-risk customer support and actually make it better than human service.

CEO and co-founder Dmitri Masin shares the story behind Gradient Labs, what they’ve learned so far, and how they’re growing fast by building something real, not just another GPT wrapper.

The Unlikely Leap from Operator to Founder

Dmitri’s founder story doesn’t begin with a dorm-room prototype or a pitch deck. In fact, a year before founding Gradient Labs, he wasn’t even thinking about starting something.

“Some people dream about starting a company while still at uni. That wasn’t me,” he admits. It only became obvious after we saw what GPT-4 could actually do.”

By then, Dmitri had spent nearly seven years at Monzo, the UK’s largest digital bank, where he joined as employee #30 and eventually led a 150-person team across data science, ML, and financial crime. That’s where he met his Gradient Labs co-founders, who were both on his team.

Their shared obsession? Operational efficiency.

“We were always deep in the weeds of customer support, ops, fraud, the places where repetitive work eats budgets and humans are prone to error.”

That hands-on knowledge gave them a unique edge when GPT-4 dropped. Suddenly, the scale of automation possible was seismic.

“We went from automating 10% of work to seeing the potential to automate 70–80%. That’s not incremental, that’s transformational.”

But they didn’t just want to build another chatbot.

Building With Intention, Not Just Speed

Gradient Labs didn’t rush to market. In fact, they spent over 15 months in build mode, refining, testing, and obsessing over quality before their first public launch in late 2024.

“YC would probably tell us we did it wrong, but we had this bar in our heads: ‘Would a bank trust this?’ And until we hit that quality bar, we kept building.”

That bar was high. And their early version? Not even close.

“We tested it on live traffic. It felt like just another dumb bot. It was discouraging, but also clarifying. We scrapped it.”

From that moment on, one question became their north star:

“What would a human do? If humans can intuitively ask clarifying questions, navigate ambiguity, and make judgments, why can’t our AI do that?”

That phrase “WWHD” became their product team’s internal mantra. And slowly, over the next six months, they rebuilt everything.

By summer 2024, the results were tangible: Gradient’s AI agents could now understand customers, ask clarifying questions, and operate within the unique mental model of each business.

“It felt like our AI agent wasn’t just responding, it was actually understanding.”

Growing Through Trust, Not Ads

When Gradient launched publicly in November 2024, they weren’t starting from scratch. They had design partners, mostly fintechs, who’d stuck with them through the long build.

“By then, we could say something powerful: our agents perform better than the average human team. Higher CSAT. Fewer mistakes. Lower cost.”

And that proof paid off.

Within just a few months, Gradient signed 11–12 paying customers, all in regulated industries like financial services.

Surprisingly, the team never positioned themselves as a fintech solution. But that’s where the traction happened.

“Turns out, our background gave us an edge. We speak the language of compliance. We know what governance means. That builds trust fast.”

Their best growth channel? word-of-mouth.

“We’d ask customers where they heard about us, and it was always something different. A Monzo engineer. A VC intro. A customer referral. But it was always someone they trusted.”

Outbound didn’t work. Ads didn’t move the needle. But personal networks and real results? That did.

What They’ve Built and Why It Matters

Gradient Labs’ product is more than a support agent. It’s a RAG-powered, company-aware, ambiguity-handling AI trained not just on generic data, but on each company’s internal documentation and processes. Built on top of Anthropic’s Claude 3.5 Sonnet and Google’s Gemini 2.0, their agents don’t hallucinate; they act based on facts and business logic.

And the results?

  • 50% of customer queries resolved out of the box
  • CSAT scores of 80–90%, on par or better than human teams
  • No financial advice. No regulatory missteps. Full audit trails.

That’s not “move fast and break things.” That’s “move thoughtfully and get it right.”

Real Talk: What Didn’t Work

“We thought outbound might work. It didn’t. We thought we’d sell across industries. Turns out, fintech is our wedge. We thought we were building a bot. We were actually building something far more nuanced.”

These kinds of learnings don’t show up in headlines. But they’re gold for first-time founders, and Dimitri’s quick to share them.

“Ultimately, once you build something that delivers real value, growth becomes easier. But the hard part? That’s getting the quality right first.”

Advice to Aspiring Founders

When asked what advice he’d give to someone thinking about starting something in the AI space, Dimitri doesn’t hesitate.

“When I was at Google, mobile was just taking off. I didn’t act. I was too junior, too unsure. But I’ve realized these seismic shifts, mobile, and now AI only come two or three times in your career. This time, I wasn’t going to miss it.”

His advice?

“If you feel the shift, act. You don’t get many of these moments. You don’t need to have it all figured out. But you need to start now.”

Gradient Labs Today

With the product now in the hands of real customers, Gradient Labs has shifted its focus to scaling carefully.

“We’re at about £1 million right now. That’s the trajectory we’ve been tracking since launch.”

Not bad for a company that only launched publicly in late 2024.

The team has grown to 15 people and is continuing to expand.

“We’re looking to grow to around 35-40 by the end of the year, but we still want to keep the quality bar high. That’s always been the most important thing.”

While some companies sprint to scale, Gradient is choosing to move with intention, just as they did in the early days. Every new feature, every new customer, every new hire is another chance to get it right, not just ship fast.

Final Thought

Dimitri didn’t set out to be a founder. He set out to build something excellent. And in doing so, he’s created not just a startup but a model for how to build responsibly in the AI era.

If you’re thinking about taking the leap, he hopes his story helps you realize: the moment is now.

“AI will redefine how companies operate over the next 3-5 years. The only question is, do you want to build that future, or watch it happen?”