How we work
One job at a time,with a decision at every stage
Every project moves through four stages. Between each one is a decision point, so you never commit to more than the next stage, and you can stop with something useful in hand.
Four stages, three decision points
Each stage ends with something you can use on its own, and a choice about what happens next.
- 01Discover and Prove answer one question: is this job worth doing with AI?
- 02Each decision point is yours. Stopping is always an option.
- 03Launch and Grow widen use only as the results support it.
What happens at each stage
- 01
Discover
What happens
We work with the people who own the work to choose one job, and agree how success is measured: time saved, errors avoided or answers given faster.
What you get
A short brief: the job, the measure, the data it needs and the risks.
Decision pointBuild a pilot, or stop and keep the brief. - 02
Prove
What happens
We build a working pilot on your real data and test it against your own examples.
What you get
A working pilot, an evaluation sheet and a go or no-go recommendation.
Decision pointLaunch, adjust, or stop. - 03
Launch
What happens
A small group uses it first, with monitoring and human review in place.
What you get
A live system for one team, with monitoring, logs and a review screen.
Decision pointWiden use, adjust, or pause. - 04
Grow
What happens
When the results support it, use widens to more people, or the next job starts at Discover.
What you get
A support plan, regular reports and a list of next candidates.
Decision pointExpand, start the next job, or keep things as they are.
What stays the same at every stage
- One job at a time
- Each project is about one job, so its value is easy to see and easy to judge.
- Measured
- The measure is agreed before work starts, and results are reported against it.
- Tested on your examples
- Every system is checked against real cases from your work before anyone relies on it.
- People in control
- People approve the actions that matter, access is limited to what each task needs, and there is a record of what the AI did.
- Yours to keep
- You keep the code, prompts, configuration and documentation, at every stage.
- Honest advice
- If a simpler tool or a plain automation would do the job, we'll say so, even if it means less work for us.
Who's involved on your side
Three roles, often fewer people. We keep meetings short and ask for examples rather than documents.
A sponsor
Owns the result and makes the call at each decision point.
The owner of the work
Knows how the job is done today and supplies the examples we test against.
IT and security
Review access, design and code, and agree before anything connects to your systems.
Questions
Can we start at Prove?
Yes, if the job, the measure and the data are already clear. We do a short check first to make sure.
Can we stop after Discover?
Yes. You keep the brief, and you can take it forward with us, another team or not at all.
How much of our team's time does it take?
Usually a few hours a week from the owner of the work, plus reviews from IT. We'll agree it at the start.
What happens if the pilot is a no-go?
You get a written explanation of why, and what would need to change. That's a useful result too, and a cheaper one than finding out later.
Tell us one job you'd like AI to do
We'll tell you honestly whether AI is the right fit, and what a sensible first step looks like.
Thanks. We'll reply within one business day.
Your message has been received.