Ninety days has become a fashionable unit of measurement. It is nearly always used either to promise an entire transformation, which is impossible, or to justify a plan that changes nothing. But three months is enough for some very concrete things, if you choose them well.
What fits
- An honest picture of how your operation works today, with times and costs measured rather than estimated.
- Two or three chosen fronts and, above all, a list of what stops. It is common for half of what is currently running not to survive the review.
- One critical process redesigned and running with a real team, with comparable before and after numbers.
- Between twenty and a hundred people using AI in their daily work, with good examples saved and shared.
- Clear rules: what data can be used, who validates what, and how much can be spent without asking.
- Named internal owners for every piece.
What does not fit
Changing the culture of a seven-hundred-person company does not fit. Rebuilding five processes at once does not fit. Replacing your ERP does not fit. And proving consolidated financial return does not fit either: at ninety days you have pilot indicators, which point somewhere, but they are not the P&L yet.
Promising any of those is the fastest way to lose credibility in month four, exactly when you need it most.
The split that works for us
The first three weeks are for understanding and focusing. It is the part people most want to shorten and the part least worth shortening, because it is where you decide what stops. Weeks four to eight are for building and testing with real people. Weeks eight to twelve are for extending, training and writing down how it is sustained.
At day ninety, what matters is not how much you have saved. It is whether the team can carry on without you.
The signal we look for at the end
There is a question we always ask in the final meeting: how many things can you change without calling us? If the answer is “almost everything”, the project worked even if the savings are still modest. If the answer is “none”, we have built a dependency, and that is a bad outcome no matter how good the pilot numbers look.
That is why the last phase is not a report. It is training, named owners and documentation somewhere your people will actually find it.