How we scope a custom AI system: from discovery to roadmap
Short answer: we scope a custom AI system by listening before we build. A few recorded discovery calls let us map how the business really runs, then we turn that into a phased roadmap with the most urgent piece first. Full programmes start from £15,000.
Key points
- We don’t write a proposal until we understand the whole workflow.
- The roadmap is phased, so the most urgent piece gets built first.
- Payments follow milestones, not a big upfront bill.
- We roll out one office at a time, so problems show up before everyone is live.
- After launch there’s ongoing support, with hours for changes included.
Why we don’t start with a proposal
Plenty of AI projects go wrong in the first week. Somebody writes a proposal from a short conversation, and the thing that gets built solves the problem as it was described, not the problem as it really is.
Most businesses run on habits nobody wrote down. The person who always checks a document before a booking goes out. The spreadsheet that gets updated on Fridays. The phone call that sorts out the exceptions. If we don’t find those, we build something that works on paper and falls over on Monday.
So we start by listening.
Step 1: discovery over recorded calls
We run a few calls with prepared questions, and we record and transcribe them. The questions are different for every business, but they tend to follow the work: how does an enquiry arrive, what happens next, who checks it, what goes wrong, what do you do then?
Then we analyse the transcripts and map each process end to end. That map shows what the system has to do, where the awkward exceptions are and which jobs take the most time.
It’s slower than guessing, and it’s the part that saves the most trouble later.
The kind of questions we ask
- Walk me through the last booking, order or job from start to finish.
- Where does someone have to stop and check something by hand?
- What goes wrong most often, and who fixes it?
- Which jobs are done by one person that nobody else understands?
- If you could switch one thing off tomorrow, what would it be?
The answers are usually more useful than any document the company already has.
Step 2: a phased roadmap
Once we understand the workflow, we write the plan. It’s phased, with an honest estimate for each stage.
The order matters. We put the most urgent piece first. On one project, that was a portal for a trust, because they had a board meeting coming. Other parts of the system could wait a few weeks, so they did.
A phased plan also means you see something working early. You’re not waiting months for one big reveal, and you can change course if what you see isn’t quite right.
Step 3: milestone payments
We don’t ask for the whole bill upfront. A typical structure is a deposit to start, a payment at a milestone and the balance on completion.
That keeps things fair. You’re paying as the work lands, and we’re paid for work we’ve delivered. Our full programmes start from £15,000, and we’ll give you a written scope and price before anything begins. All prices exclude VAT.
Step 4: a phased rollout
Building it is only half the job. Getting people to use it is the other half.
We roll out one office at a time, rather than switching everyone on at once. A small group finds the problems, we fix them, and the next office gets a better version. By the time the whole company is live, the rough edges are gone.
Step 5: ongoing support
Businesses change, and a system that’s perfect today will need adjusting in six months. After launch we support it for a monthly fee, and that fee includes allocated hours for changes.
That means you’re never stuck waiting for a quote just to add a filter or change a rule.
What you own
A custom build is made for your business. That matters more than it sounds. One of our clients came to us partly because a past software vendor had changed direction, and they couldn’t change their own system. With a bespoke build, you’re not at the mercy of someone else’s roadmap. We agree the ownership terms when we scope the project.
What this looks like in practice
We’re building a full AI integration into a UK recruitment agency, and it began the same way: recorded discovery calls, a map of the whole process from booking to timesheet, and a build in phases. The full story is on our project page.
Is it right for you?
A custom system makes sense when you’re rebuilding how a team works, not when you want to save a few hours a week. If that’s closer to your situation, a single AI teammate is a cheaper and quicker place to start, and plenty of our larger projects began that way.
Want to see how this works for your company? Read more on our custom AI systems page, or book a call and we’ll talk it through.