We run forward-deployed AI teams inside regulated businesses. The patterns we see become the products we build. Services fund the studio. The studio sharpens the services. Neither works alone.
AI is easy in unregulated industries. Ship fast, iterate, apologise later. That model doesn't survive iGaming, fintech, or healthcare. These industries reward companies that ship fast AND survive audit. That's a very small club. We want to be at the top of it.
Every decade, a technology shift redraws the map of who wins in regulated markets. Cloud in the 2010s. Mobile before that. AI is the current shift, and regulated industries are responding the same way they always do: with suspicion, slow procurement, and compliance reviews that outlast the vendors they're reviewing.
The winners in this shift won't be the firms with the cleverest models. They'll be the ones who can build production AI inside regulated operators, alongside their compliance teams, without cutting corners. That's a discipline, not a pitch.
Regulated industries sit at an unusual intersection. They have real commercial pain that AI can solve. They have the revenue to pay for it properly. And they have enough complexity that anyone who builds a reliable playbook gets a moat that's difficult to copy. That's where we want to live.
Most AI vendors can't or won't do the compliance work. The ones that do, earn long memory from their customers.
Regulated operators aren't experimenting for the sake of it. They have commercial pressure and the budget to solve it.
The switching cost of a vendor that's already passed audit is genuinely high. That's a commercial asset, not a risk.
Big Four is too slow and too generic. Silicon Valley doesn't want to read the regulations. The middle is wide open.
Most venture studios burn capital until a winner emerges. Most services firms stay stuck at the ceiling of billable hours. We run both, on purpose, because each one makes the other stronger.
We place forward-deployed engineers inside regulated operators. Real problems, real constraints, real data.
The second and third time we solve the same problem in a new industry, we know it's not an engagement. It's a product.
We turn the pattern into a product. Spin it into its own entity. Find its market. OneHazel was the first.
The product is a tool our engineers use in the next engagement. Our clients benefit from IP we've already built.
Fintech, iGaming, healthcare, digital media, software testing. You want production AI shipped inside your business without risking your licence or your quarter.
Work with usYou've got a prototype, some traction, and a category forming. You want a home that understands your sector, not a generic accelerator that treats you like a SaaS template.
Join the studioWe're raising a pre-seed round. Services-funded. One venture already live. Investors who want to back studios building in regulated industries should get in early.
Talk to the teamSenior AI and platform engineers who are bored of either pure consulting (no ownership) or pure product (narrow scope). You want both. We're hiring and we're hiring well.
See open rolesWe don't run a cohort. We take an equity stake in a small number of ventures a year and back them the way we built OneHazel: an embedded engineering team plus hands-on direction on product and go-to-market. No demo day, no fixed intake — pitch us whenever you're ready.
Senior AI and platform engineers join your build — the same team that ships our client work and OneHazel.
Product and go-to-market input from operators who've built inside regulated industries, not passive observers.
We take a stake instead of a fee. We only win if the venture does.
Fintech, iGaming, healthcare, digital media, software testing. The unifying thread is regulation, not vertical.
We're not consultants who discovered AI. We're operators and engineers who've shipped inside the industries we now build for. That's who we want working on client engagements. That's who we want building ventures.
Meet the full teamOperator, founder, investor, or engineer, there's a conversation worth having. Thirty minutes, no deck.