Katya Lait and her co-founder Jack officially launched nettle in January – an idea that first took shape in the summer of 2024. Now, just three months in, they are well on their way to tackling a major inefficiency in the insurance industry.

Risk engineers were spending 90% of their time buried in analysis and report writing. nettle is flipping that ratio, significantly cutting the time spent on reports and giving risk engineers back the time to do what they do best.

We spoke with Katya to hear how nettle came to life, the challenges they’ve solved so far, and what’s next on their startup journey.

---

Let’s talk about nettle! What specific problem are you solving?

In short, we’re making the commercial insurance market more efficient. Everything from an oil rig to a fleet of company cars can be insured but, in order for the policies to be written, these assets need to first be assessed. These assessments are carried out by professionals called risk engineers.

Risk engineers are often highly experienced and have extensive industry knowledge which makes them uniquely qualified to assess these assets. They complete the assessments in person which is perhaps the most interesting part of the job because they get to meet people, take observations, and give them suggestions to improve. Unfortunately, it's also the part of the job that takes up the least time – and it's something they'd like to go back to doing more.

Most of their time (about 90% of the job) is spent on analysis and report writing. nettle is reducing that 90% down to 10%, making risk assessments more efficient and giving back the time to do more projects and assessments. So we’re looking at an 80% reduction in the time spent per project. In effect, that creates four times value per risk engineer in an organisation.

 

How did the idea come about? 

Jack and I started the business together last year; the nettle of today is the combination of work we were both doing in our respective projects. The idea for the underlying technology originated from a research project I was working on at Quantum Black, which involved researching and building their pioneering multi-agent system now referred to as AgentFlow. Our focus was on automating the writing of credit memos for big banks – a document not too dissimilar to a risk engineering report.

Today's use case and the problem we tackle at nettle came from Jack, who spent six years in McKinsey working across insurance. Throughout that time, this opportunity kept coming up in strategic advisory projects for insurance enterprise clients. They wanted to grow commercial underwriting but the biggest bottleneck was this risk assessment step, which is unavoidable.

That’s where the idea for nettle was born, and my technical skill set complemented Jack’s vision well. We got to ideating, talking to potential customers, and went from there!

[FOUNDER IMAGE]

Do you find there to be any common misunderstandings about AI when talking to customers?

We typically have two kinds of conversations.

The first is always the best. They get excited and start asking us, “Can it do this? Can it do that?”  We’ll get a list of 10–20 things they want nettle to do. It’s great and exciting, but building the kind of features they want takes time, and we have to temper expectations a little bit in the short term. We put our users at the heart of every decision, so we prioritise the AI features that create real value and avoid adding unnecessary features just for the ‘wow’ factor.

The second is more difficult. They may misunderstand what's happening, so they'll think that we’re reading all of their data, all of their personal information. They don’t trust the system and want us to list every single source we use.

For us, it’s a balance. Obviously, we want to create trust within the product, but multi-agent systems are very complicated and dense in their workflows. We pull data from many sources and process them in complex ways, so we have to make tradeoffs between creating trust and not completely overwhelming them with what's going on in the background.

What’s the most exciting technical challenge that your team has solved so far?

This is super specific to a current client we're working with, but it's deploying an open-source model and building an entirely secure on-prem solution. 

We're working in a country right now that has strict data sovereignty laws, so no customer data can leave the region. This ruled out using OpenAI or other services that would’ve made our jobs easier. Instead, we built this multi-agent system using in-house solutions, open source models, and internal GPU and cloud servers. It was definitely intense, but we got there in the end. 

Wow! How long does it take to do something like that?

Thankfully, the way we built our agent system means replacing the underlying model is pretty quick. The challenge was working with a model that's not as smart and making it more robust with limited resources. It involved more sophisticated prompting and optimisation of the available context. But honestly, finding a GPU took the most time.

We couldn’t rely on AWS, Azure, or Google, so it was all on-prem and fully in-house. But we’ve learned a lot about how to build a safe and robust AI platform. When we work with clients in regions with more open data laws, we will be able to leverage the latest AI models while also delivering a highly secure solution. 

What is the biggest challenge in bringing AI products like yours to market at scale?

Scalability isn't top of mind yet. We’re still in our first year, building bespoke versions of nettle for clients. Year Two is when we’ll step back, consider the ecosystem of solutions we’ve built, and focus on scalability. 

That said, some obvious steps we’re taking in the short-term are to distribute requests across our specialised agents, reducing unnecessary calls to APIs and building generic components we know we will reuse across all our clients. As much as we can, we cache everything and have the system learn from itself – a crucial feature for our mission to close the knowledge gap in risk engineering.

That makes sense. What has been the biggest challenge in your startup so far?

Bringing AI to new markets! Beyond data sovereignty and creative use of open-source solutions as I’ve mentioned, the biggest challenge has been translation. 

We’re working in a country where English isn’t spoken, so we have a pretty sophisticated AI system that reasons in English but delivers results to the user in their language. We have a very complex workflow around this, but it's pretty bulletproof at this stage. It also means that we can support this in virtually any country moving forward.

[TEAM IMAGE]

You’re only three months in and already doing great things! How do you see your technology evolving over the next few years?

Right now, we’re positioned as an in-house solution for a specific function in insurance companies. For Series A and beyond, we’re considering two directions.

One is expanding into risk assessments: becoming not just a tech company, but also a service company. It’d involve getting the most advanced hardware to do our assessments and becoming a centre of excellence on risk assessment. The other option is producing software and hardware together, bringing in drones and 3D visualisations that feed directly into our platform.

There's a lot still to be uncovered. When we come into an organisation – one of the big three insurers, for example – it's challenging to get them to migrate from their existing solutions. Often, it's not ideal and it'll never be able to support the AI things that we can do. Instead of forcing migration, we could build an AI agent that runs alongside their current system, and controls their machines to fill in reports automatically with their existing software.

When it comes to AI trends, what do you find exciting?

For me, it’s all the amazing developer tools that are coming out right now! 

I am just so much faster. What we’ve achieved in the past three months is fuelled by the amount and quality of the AI tools I have at my disposal. I’ve got an AI agent reviewing all my code, catching errors I wouldn’t have spotted. It’s accelerated everything and made our platform so much more secure.

We're testing everything on the market right now. Anytime I hear about it's a new dev tool, I test it out immediately. It’s exciting because the four of us can now do the work of 10 engineers.

Which are your favourites? Any recommendations?

Oh yeah! I like Qodo – that’s what I use to review my code. We use Granola for all our meetings: it's so helpful for us when we're doing tech overviews and onboarding people via Zoom. The obvious one is GitHub Co-Pilot… everyone's using that, it's fantastic. Cursor is also really good.

My all-star favourite – it's not quite a dev tool, but it's one that I use about 20 times a day – is Claude. It's my thought partner during the day; I constantly have it open to ask quick questions.

And is there anything in the AI world that you think is overhyped?

Not overhyped per se, but I’ve been underwhelmed by some built-in AI features in tools like Word or Notion. They’re pushing them a lot, but I still see a bit more room for improvement there.

I’m also not sold on AI personal assistants that respond to emails and manage your calendar. I've heard about people who end up having their inbox full of AI talking to each other - it kind of misses the point of an email!

[KATYA PIC]

You must be incredibly busy. How do you relax outside of the work you do?

Not a whole lot these days! I’m obviously working a lot, but I recently got an Apple Watch, so I’m trying to be more active when I can. The way they’ve gamified it helps me to stay motivated.

Finally, if you weren't working on Nettle, what other problems would you be trying to solve with AI?

Well, this is a good opportunity to shout out another startup on the Faces of AI list – MAIHEIM. What they’re doing is super exciting.

If I wasn't working on something industry-specific, I would definitely be in the realm of dev tools and AI for tech developers. That’s an area that’s going to balloon like crazy and I can’t wait to see what comes out of the space.

---

[FACES OF AI BANNER]

📥 Katya is one of the 54 innovative technical founders recently featured in our 'Faces of AI 2025' report, which is available to download for free now.

At Generative, we work with AI-driven startups to scale innovation, refine strategy, and accelerate growth. If you're building in AI and looking for the right talent to take your company to the next level, let's talk.