Most people stop after step two and that's fine. Nobody gets a three-month discovery phase, a slide deck, or an account manager who's never seen their business. You deal with the person doing the work, start to finish.
An hour at your place. We watch how the work actually flows, ask daft questions, and look at the screens you're clicking round all day. You get a one-page list: three or four ayes (worth doing, roughly what it'd cost and save), and the nays (not worth it, and why). No pitch. If none of it stacks up, you've lost an hour and a brew.
Four hours with you and whoever does the admin, using Claude or ChatGPT on your own data. By the end you've built the first thing yourself, you know what to ask for, and you've got a one-page rule sheet for what never goes into a public AI. Most businesses get the biggest single win here, and plenty don't need anything else for a year.
When the half-day runs out of road, one job gets built properly, into the tools you already use. Fixed price agreed before we start. You see it working every week, not at the end, so if it's going the wrong way you say so in week one. At the end you own it: the code, the accounts, the data. No lock-in.
Software changes, and the thing that worked in March can break in September when your booking system updates. This keeps it running: updates when something upstream moves, a monthly check that it's still doing what it should, and a person on the phone when something looks odd. Cancel whenever. Nothing stops working if you do.
Honest version of what happens afterwards: once the first thing works, you'll spot the second. The morning brief leads to the Friday chase, which leads to one place to ask across everything. For a bigger firm that gets to the point of wanting someone senior to own it, that's the fractional AI lead: a day or two a month, a roadmap you own, and someone to say no to the wrong tools.
Almost never. It sits between the systems you've got and removes the manual steps: the copying, the re-typing, the checking. Your booking system, your accounts package and your shop stay exactly where they are. If something genuinely needs replacing, you'll be told that on the look-round, not sold it later.
The look-round is an hour. The half-day is four hours and you leave with something working. A build is typically four weeks from agreeing the price, and you see progress every week. Nothing takes "a quarter".
You do. The code, the accounts it runs under, the data, the prompts. If you stop working with us, it keeps running and you can hand it to anyone. We keep the right to reuse general methods, never anything specific to your business.
It stays in your systems. The AI reads what it's given permission to read, under business terms that exclude training, with everything logged. Nothing is copied into a consumer AI app, and nothing leaves the UK where the provider lets us choose. The customer data post has the long version.
It will, occasionally, which is why every build has a point where a person checks before anything reaches a customer or touches money. The AI drafts, flags and summarises; a human approves. When it does get something wrong, the log shows exactly what it read and why, so it gets fixed rather than guessed at.
Nay. It replaces the part of their day that's typing the same thing into three systems. Every business we've done this in has the same people afterwards, doing more of the job they were hired for. If a tool's whole case is "then you can let someone go", we'd rather not build it.
Usually a business-tier AI account at £20–30 a month per person who uses it, and that's often it. If a build needs a server or a service, it goes on an account you own and we tell you the running cost before starting. No "AI platform" subscriptions.
Anywhere within an hour's drive for the look-round and the half-day, which covers most of North and West Yorkshire. Builds and support work anywhere, since most of it happens at a desk. For the bespoke and MCP work, location doesn't matter at all.