Founder Of The Week: Dawid Kotur

• Dawid Kotur is the co-founder of Curvestone AI, a company that uses AI to help legal and financial firms improve compliance checks and reduce risk.
• Before starting Curvestone in 2023, Dawid worked in banking, consumer brands and enterprise AI, including roles at Metro Bank, PwC and GKN.
• He launched Curvestone after seeing how difficult it was for regulated industries to trust AI systems, especially when dealing with complex real-world documents and strict compliance requirements.
• Dawid and his brother Sebastian built Curvestone into a profitable business without raising external funding, focusing on solving a real industry problem and earning customer trust through proven results.

Tell Me About Yourself and Your Company

Initially from Poland, my family came to the UK when I was a teenager and after studying Archaeology and Anthropology at University College London, I built a career inside UK financial services. I was the first Head of Mobile at Metro Bank, then ran a hip-flask brand across 26 countries. From 2017 I embedded AI into programmes at PwC and GKN and spotted that even when organisations automate decisions they still need human oversight.

My brother Sebastian and I founded Curvestone in 2023 when we found that generative AI could give us the technology needed to fully solve the compliance related problems we had been working within these large organisations.

Compliance teams manually spot-check 5–10% of casework because they don’t trust automation, but only checking such a small sample size was leaving firms exposed to error, with an incredibly high cost if you get things wrong. Curvestone checks 100% of cases with a 100% audit trail.

Currently, we are live within the heavily regulated legal and financial sectors where we process thousands of checks a quarter.

What Inspired You To Start Your Company, and What Problem Were You Trying To Solve?

During the years I spent helping big firms deploy AI, I repeatedly saw the same pattern. Systems work beautifully on clean demo data – you feed them perfect documents and they perform perfectly. Then they hit real data and collapse.
I had seen that in the real world, you don’t get pristine documents. You get photographed IDs taken at angles, scanned payslips with coffee stains and email chains that have been forwarded six times. And in regulated industries that stuff doesn’t cut it as official evidence.

To take one example, mortgages are actually the perfect storm for this problem. You’ve got huge document volumes, incredibly strict FCA rules, and the economics are completely broken. You can either audit everything and go bankrupt, or you cut corners and hope nothing breaks. There’s no sustainable middle ground.

That’s what changed when generative AI matured enough. It actually became possible to solve. So instead of consulting on other people’s AI programmes, we decided to build our own.

What Has Been Your Biggest Challenge So Far, and How Did You Overcome Them?

The biggest challenge has been selling AI to compliance teams, the most sceptical of buyers, and quite rightly, you have to be in a regulated industry. So we stopped pitching and started proving. We ran pilots on their own historical files, benchmarked against their own reviewers and showed them the evidence behind every finding. It was slower than selling on promise but we saw the pay off quickly – one sceptical compliance director’s reference beats any campaign.

It’s also how we reached profitability before raising a penny. There’s no shortcut to trust in regulated industries. You build it, case by case.

Can You Describe a Pivotal Moment That Significantly Shaped the Direction of Your Startup?

We spent years building AI for other people and handing back everything we’d learned at the end of every project.

When gen AI hit, document understanding was obviously going to be a commodity. Every vendor would have it but we saw the opportunity was that almost nobody was actually building for regulated spaces and we wanted to build something that could survive FCA scrutiny, where you can’t be 80% right.

That was the moment. We killed the advisory business and bet everything on one product.

The real validation came in production. The first time it actually ran on a real broker file, complete with photographed payslips and paywalled PDFs, not our test data and it held up. It didn’t break. That’s when we knew we’d found something worth building.

How Do You Define Success:

As a Business: It has been about making 100% compliance checking the norm, not 5 to 10% spot-checks. We watch volumes – how many cases run through the platform – not headcount and then revenue follows. It’s increased 7x in twelve months but for us, cases checked are always the north star.

For Myself: It is about things actually running in production, not AI theatre. Demos that never change how a single case gets processed are pointless. True success is also building this with my brother without breaking either the company or the family – and we’re succeeding on that so far!

What Advice Would You Give To Someone Thinking About Launching Their Own Startup?

My first piece of advice is to get profitable before you raise, if you can. It changes every conversation – with investors, with customers and with yourself. You’re not pitching potential but showing evidence.

Aside from that, pick the genuinely hard problem, not the one that demos well. Anyone can build a demo. The moat is the hard bit – the part that takes years and breaks things along the way. That’s why we’re called Curvestone. It’s literally about taking something brittle and shaping it into something stronger.

What’s Next for Your Company – Any Exciting Developments We Should Watch Out For?

We’re expanding across the full compliance stack, including financial promotions, PEP and adverse-media checks, wealth advisory. There is so much potential beyond mortgages, although they will always be a core part of the business. We’re also embedding deeper into the platforms firms already use so there is no need for rip-and-replace theatre.

Governance is something that Curvestone takes very seriously. We’re completing ISO 42001 for AI management this year alongside our existing 27001. We’re also excited to have been selected for the FCA’s Open Finance TechSprint on mortgages and SME finance (Smart Data Accelerator, building and testing synthetic data on explainable decision-making. This kind of experimentation is where we get real value – value that goes straight back into the business. It is a great initiative and means we stay close to real production work with real regulatory constraints. That’s where the honest feedback comes from.

Want to be featured as TechRound’s Founder of the Week? Find out more about this weekly feature and how to get involved here.

Founder’s 5 with Dawid Kotur

We wanted a little more insight into the man behind Curvestone AI, so here’s TechRound’s exclusive Founder’s Five with Dawid Kotur.

Favourite Business Tool

Claude

One Lesson You Learned the Hard Way?

Building B2B SaaS is all about iteration speed. Waiting for customers to use your product and volunteer feedback could kill you as you won’t be improving fast enough. You need to find creative ways to get feedback and get your own team to really use the platform so the feedback is non stop.

One Future Trend You’re Watching?

Cost of intelligence. Our space will transform if the cost of model usage goes down 100 x

One Quote You Live By

“It’s not about 10,000 hours, it’s about 10,000 iterations.”

One Book/Podcast You Recommend

“Huberman Podcast” – great protocols for maintaining energy and focus that is needed to be a founder of a fast-growing company.

Want to be featured as TechRound’s Founder of the Week? Know someone who deserves to be recognised as a founder making waves in the startup landscape? Find out more about this weekly feature and how to get involved here.