"The board wants an AI strategy": where to start in 90 days

The phone call this story opens with now reaches us every month, with minimal variations: the board has asked for an AI strategy by next quarter, the assignment has landed on a capable and suddenly lonely executive, and the instinctive first move, already on the agenda, is a round of vendor demos. It is the move this series has spent six months advising against, and the piece that closes it wants to offer the alternative in executable form: the ninety-day sequence, with the weeks counted and the deliverables the board expects to see.

The premise that orders everything sits in a distinction: the board is not asking for a list of technologies, it is asking where AI serves the business, what it costs, what it risks and who answers for it. That question is not answered by starting from tools, it is answered by starting from a photograph.

Weeks 1-4: the photograph

The first month produces the factual base that makes every subsequent decision honest, and it runs two parallel workstreams. The first is the maturity measurement on the four dimensions readers know, Readiness, Delivery, Risk and Confidence: where we are prepared, where we know how to execute, what we are permitted to do, who believes in it. The second is the census of what already exists, the ongoing projects with their true status, the systems in use including the never-authorised ones, because in over 90% of companies, MIT counted (2025), shadow AI is the de facto strategy awaiting the official one.

The end-of-month deliverable is the declared starting point: the rating with the gaps named, the inventory, and the first uncomfortable truth to bring to the top, which almost always concerns data or governance, not models.

Weeks 5-8: the choices

The second month turns the photograph into a portfolio. The use case discovery collects candidates from every function, the prioritisation grid orders them with weights decided beforehand, the Canvas forces the top three to five to answer the six questions on one page, opportunity, risks, priority, stakeholders, capability, dependencies, and the go/no-go criteria fix the thresholds that will decide in place of opinions. In parallel starts the minimal governance that cannot wait for the perfect plan: the systems register, the named responsibilities, the everyday-usage policy that turns shadow AI into governed adoption.

The deliverable at the end of month two is the motivated portfolio: what we do, what we do not do and why, with the nos written next to the yeses, because a strategy that discards nothing is not a strategy, it is a wish list.

Weeks 9-13: the test and the plan

The third month does the thing that separates a strategy from a document: it starts verifying it. The use case at the top of the portfolio enters validation with the protocol we know, golden set, signed thresholds, stopping conditions, three weeks; the outcome, whatever it is, becomes the strategy's first real data point, and a no at this stage is worth as much as a yes, because it costs weeks instead of quarters. Around the test the rest closes: the twelve-to-eighteen-month roadmap with the workstreams sequenced, the measurement model that will say every quarter whether the plan is working, the compliance chapter with the real deadlines, August's and 2027's, inside the plan instead of beside it.

On day ninety, in front of the board, the deliverables are four: the photograph with the rating, the portfolio with its nos, the outcome of the first validation, the roadmap with governance and measures. No transformation promises, no tools recommended on faith: a measured position and a verifiable plan, which is exactly what a governing body can approve without pretending to understand what was never explained to it.

Closing the circle

Whoever has read this series from the beginning will recognise in the ninety days the miniature of the whole journey: measure first, decide next, always validate, govern while doing. We opened it in January with the stuck pilots and we close it with their antidote, and with the one number from the Politecnico Observatory (February 2026) worth repeating a final time: 71% of large companies have AI projects, 9% have structured governance. The ninety-day strategy is, at bottom, the bridge between those two numbers.

It is also, plainly said, the work we do: the whole sequence, from the rating to the approved roadmap, is the ZeroFive journey, and the first conversation costs half an hour of calendar, calendly.com/fabiolalli/zerofive, or hello@zerofive.ai. If the phone call we opened with has reached you too, the right question to start with is not which tool to choose. It is: what do we know, today, about where we really are?