Okawai Consulting's proprietary method
P.R.E.U.V.E.: from AI decision to production results.
Six steps. Measured proof at every step. Most approaches stop at a deck approved by the executive committee. We take it further, through to production and run.
The six steps
P.R.E.U.V.E. is French for "proof". Each of the six letters names one step of the method, described here in English.
P — Photograph.
Establish the real state of play: declared and hidden uses, data quality, contracts and vendor clauses already signed.
What comes out: a one-page snapshot per department, a register of GDPR and AI Act exposure, a first list of champions.
R — Connect.
Quantify what AI changes by function, in euros and in risk, and bring the executive committee up to speed using its own cases.
What comes out: a quantified impact map, the business cases, the skills to develop or hire.
E — Commit.
Decide on four criteria: value, feasibility, regulatory risk, the teams' capacity to absorb the change.
What comes out: a written, dated committee decision, with named owners, budgets and stop criteria.
U — Build.
Put the first use case into real production, on your own tools, with governance set up at the same time.
What comes out: one use case in production with before-and-after measurement, a trained network of champions, secured vendor contracts.
V — Verify.
Measure what was promised, six indicators at most, and decide on corrections.
What comes out: a short dashboard, dated corrective decisions, a reasoned go/no-go on scaling up.
E — Embed.
Extend adoption across the whole company and make the setup survive the departure of those who drove it.
What comes out: an onboarding plan by population, an active network of champions, practices embedded in processes.
What your committee signs off before launch
- Who is accountable when the AI gets it wrong?
- Do we build, buy, or connect to what already exists?
- What error rate do we accept, and who signs off?
- When do we stop?
- When does the CSE enter the loop?
A question without a written answer is a debt the project will pay later.