KAY2 Studios builds and runs applied-AI products — starting with FindIt, the directory that only trusts what it can verify. We take businesses that already exist, already trade, already have customers, and we put AI where it actually saves hours.
Everyone in AI is hunting the next unicorn. We think it's already out there — and it's been open since 7am. It's the supplier directory that's been running since 2017. The workshop that builds things out of timber. The café, the florist, the family firm on the high street.
They don't need disrupting. They need what the giants have, at a size that fits them. That's the levelling we're here for — and it's the fastest way we know to make AI mean something outside of a demo.
A business that already trades has the two things no startup can buy: customers who turn up, and years of hard-won knowledge about what they want.
Point AI at that and you're not gambling on product-market fit. You're compounding something that already works.
Enterprise-grade search, instant answers, always-on monitoring, automated intake — none of it is technically out of reach for a ten-person firm. It's just never been packaged for them or priced for them.
Closing that gap doesn't need new science. It needs someone to actually do it.
We measure "changing the world for the better" in something unglamorous and real: whose evening got shorter, whose backlog cleared, who stopped dreading a Monday admin pile.
Do that a thousand times over and you've moved something that a leaderboard never will.
We're not trying to replace the businesses that hold a high street together. We're trying to make them the most capable version of themselves — and then do it again, and again, until "small business" stops meaning "outgunned".
The AI space is drowning in directories that rank whatever pays. We built the one that ranks what it can prove.
Every listing is tied to a live repository and scored on public activity — stars weighted for credibility, contributors, commit velocity, whether the project is still breathing this month. Paid placement exists, and it is always labelled.
Our own products, plus client work. Client projects stay unnamed on purpose — their business is theirs to announce, not ours.
Point it at a URL and ten specialist agents try to break it in a real browser — console, layout, forms, accessibility, speed, links, mobile, journeys, security.
Every finding ships with a screenshot, reproduction steps, a failing test and a paste-ready fix. A verification pass throws away anything that won't reproduce, so the report stays short and believable.
It found a genuine exposure on our own infrastructure on its second ever run. We fixed it and wrote the guard the same night — which is exactly the point of pointing it at yourself first.
A directory trading since 2017, moved off WordPress onto the edge without dropping a single live listing — supplier accounts, moderated reviews, renewals, enquiries and claims.
Search is photo-led and instant, with an AI intent parser sitting quietly behind ordinary search copy. If the model is slow or down, search degrades to local results instead of breaking.
A parametric drawing that redraws the product live as the customer changes the options — around twenty choices across structure, openings and finish, with a scale figure so proportions read true.
Only genuinely published extras carry a price. Everything else says "ask us" rather than an invented number.
A chat widget on a public page, fenced in tightly enough that it cannot invent an answer about the business it represents.
The hard part isn't the chat. It's the refusal — knowing when to say "I don't have that" instead of guessing on a real customer.
Around 67 supervised services across two machines. Everything binds locally and publishes through a tunnel — no raw public ports anywhere.
Cross-machine failover flips DNS after three consecutive health failures and only returns once the primary is genuinely stable. We get paged, so you don't.
Give it any business-directory URL and it works out the structure itself, then flags the businesses whose web presence is broken, missing or quietly costing them work.
This is how we find the everyday businesses AI can genuinely help — evidence first, pitch second.
KAY2 Studios is deliberately small and unusually staffed: one human who owns every decision, and a crew of AI agents that never sleeps.
Founder · engineer · the one who gets paged
Works end to end — schema and API through to CSS, DNS, process supervision, monitoring and the on-call that follows. Every line of client work, every deploy, every number on this page goes through one person.
Anything that moves money, touches DNS, or reaches a customer is a human decision. The agents propose; Kane approves. That line never moves.
An always-on edge machine that watches the live estate — what's slow, what's broken, what's costing a client work — and writes the spec for fixing it.
The heavy box. Builds, tests, renders and deploys — the work that would otherwise queue behind one person's afternoon.
Agents that ship code and review each other's work, running against the same hard rules a senior engineer would enforce on a pull request.
Ten specialists that attack every release in a real browser before a customer ever sees it. We sell this one — because we needed it first.
This isn't a slide. The studio's own infrastructure, QA, monitoring and research all run on the same agent stack we build into client work — which means every weakness in it costs us first.
It's also the honest answer to "how does one person run 67 services and still take on client work?" We didn't hire a department. We built one, and we keep it on a short leash.
The idea is only worth anything once it turns into work. This is the shape that work takes on a real job.
A directory that's been trading since 2017. A workshop that builds physical products. A café. These aren't startups waiting to be disrupted — they're working assets with real revenue and real customers.
We put AI inside them: better search, better intake, better follow-up, fewer dropped enquiries. The business stays theirs. It just stops leaking.
The AI conversation is currently held by people who already speak it. The plumber, the florist, the family supplier directory — they hear the noise, can't tell the tools apart, and correctly suspect most of it is sold to them rather than built for them.
FindIt exists because that gap is a trust problem before it's a technology problem. Verify what's real, label what's paid, and an ordinary business can choose for itself.
Not "AI replaces your team". The opposite. Every hour someone spends re-typing a listing, chasing a renewal or answering the same question for the fortieth time is an hour not spent on a customer, a quote, or the next job.
We automate the dull, repeated, error-prone middle — and give that time back to the people who own the business.
The work we automate is rarely glamorous — imports, renewals, checks, replies, tests. That's exactly why it's worth automating. Nobody built a business to do data entry.
Direction, not dated promises. We ship in public — this page changes when the work does.
Small, dependency-light, portable. We reach for a framework when it earns its weight — and not a moment earlier.
Deliberately absent: heavyweight build chains on pages that don't need them, and any AI call without a declared fallback. Every model path names what it falls back to when the provider is slow, rate-limited or simply wrong.
These aren't values on a wall. They're the rules that stopped us shipping something wrong.
If you run something real — a directory, a workshop, a shop, a service — and you're losing time to work a machine should be doing, that's the conversation we want.