They fail because the wrong decision was targeted, or because the data never supported the claim. Both are discoverable in the first fortnight if you look, and expensive to discover in month five. So we look first, every time, and we tell you what we find even when it shrinks the project.
Phases one to three are where a project should change shape. If they do not, we have probably not looked hard enough.
Week 1
Before a model, the decision. Who acts on the output, how often, on what information, and what a wrong call actually costs.
Weeks 1–2
What data exists, what is missing, what is biased and what cannot lawfully be used. Most projects narrow here, and that is the point.
Weeks 2–3
A simple, cheap approach gets built and measured before anything sophisticated. If the baseline is good enough, we tell you and stop.
Weeks 3+
Design and engineering in reviewable increments, with the explanation built alongside the model rather than bolted on at the end.
Before launch
Access control, audit logging, monitoring, accessibility, security review of every endpoint, and a defined route for a person to challenge a decision.
After launch
Into production with monitoring, drift checks and a support window. Handover means your team can operate it without calling us.
If you are not sure which fits, describe the problem and we will propose one, including the smallest option.
Two to three weeks. We work through the data you hold, what a model could realistically add, and what it would cost to run. You end with a decision, an architecture and a cost model — including the option to stop.
We design and build the system, then hand it over with the documentation and tests to run it yourself. Delivered in reviewable increments, with production access and a support window at the end.
Our engineers work inside your team on your roadmap, for the period you need the capability rather than the headcount. Suited to organisations building a first AI function.
These are the things clients tell us made the difference, so we have written them down.
Working software, not a status document. If there is nothing to show, we say why.
You talk to the engineers. There is no account manager translating between you and the build.
Every milestone is a point where a project can be paused or ended without losing what came before.
Documentation, tests and sessions until your team can run it. Then a thirty-day support window.
A discovery sprint takes two to three weeks and ends with a written recommendation you can act on, whoever ends up building it.