Discover · Prioritise · Prototype · Roll out · Measure · Scale
Six steps, with an honest stopping point after the pilot and a metric agreed before anything starts.
The stopping point
If the pilot does not prove its value, the project stops and I explain why.
Dropping a case with data behind it after four weeks costs far less than finding out in production six months later.
Understand how the work happens today
I look at processes, teams, tools and opportunities.
Interviews with the people doing the work, not only those managing it. The output is a process map and a list of where the time goes.
Deliverables
- Process map
- Interviews by area
- Tooling inventory
Choose where to start
I rank the cases by impact, effort, feasibility and risk.
Not every case is worth it. I score each one against the same criteria and we decide together which ones go ahead.
Deliverables
- Use case catalogue
- Prioritisation matrix
- Roadmap
Build something testable
I build a scoped pilot and measure whether it actually helps.
One specific case, with real users and a metric agreed before starting. If it does not prove its value, it stops.
Deliverables
- Working pilot
- Agreed metric
- Evaluation on real cases
Put it into daily work
I integrate the solution and train the people using it.
The technical integration is half of it. The other half is the team knowing how to use it, when to doubt it and who to ask.
Deliverables
- System integration
- Training by role
- Responsible use guidance
Check that it helps
I instrument adoption, quality, cost and latency.
Everything I hand over comes with its dashboard: what gets used, what fails and what it costs. Without that there is no way to improve anything.
Deliverables
- Adoption dashboard
- Quality evaluation
- Cost control
Expand what works
I add new cases and maintain the ones already running.
With one case working and measured, the next costs considerably less. This is where the initial investment starts paying off.
Deliverables
- New prioritised cases
- Usage best practice
- Ongoing support
Four criteria I always apply
Process first, tool second
If the process is broken, automating it only produces errors faster. I start by understanding it with the people who run it.
A pilot rather than a long proposal
I would rather show something working in four weeks than discuss an architecture for three months.
Measurement from the start
Every delivery comes with its metric: usage, quality and cost. What is not measured gets quietly abandoned.
Human review where there are consequences
AI proposes and prepares; a person approves anything with real effect. That is defined when designing the flow, not afterwards.
The first step is usually the same
A two to three week assessment that ends in a roadmap you can make decisions with.
Shall we talk about your case?
Tell me which process takes the most time. In half an hour I will tell you whether AI solves it, how I would approach it and what it would take.