A sequencing plan that turns strategy into a shipped pilot within a quarter.
Key takeaways
The first phase is about narrowing, not building. Take your highest-priority use case from an opportunity mapping exercise and scope it down further than feels comfortable — a single workflow, a single team, a single measurable outcome. The instinct to scope broadly in order to “prove it really works” is exactly what turns a 30-day pilot into a six-month project with no clear finish line.
By the end of day 30, you should have a written scope document, a named owner, and the specific metric that will determine success, agreed on by everyone with a stake in the outcome before any building begins.
The second phase is where the pilot actually gets built and tested against real work, with real staff using it on real tasks rather than a clean demo environment. This is also where the human review step gets designed and tested, not bolted on as an afterthought once the pilot is technically working.
Expect friction in this phase — that's the point of it. A pilot that surfaces no friction in days 31-60 either wasn't scoped ambitiously enough to teach you anything, or the friction is being quietly absorbed by staff instead of surfaced, which is worth investigating directly.
The final phase measures the pilot against the specific metric agreed on in day 30, not against a new, more favourable metric chosen after seeing how the pilot actually performed. This discipline is uncomfortable when results are mixed, but it's exactly what protects the business case for whatever comes next from being built on shifting goalposts.
By day 90, the roadmap should produce one of three clear outcomes: expand the pilot with a scoped follow-on business case, adjust and re-pilot with specific changes, or stop and redirect the budget — all three are legitimate outcomes of a well-run 90 days.
The 90-day window itself is less important than the discipline of building explicit decision points into the plan from the start. A roadmap without decision points tends to drift indefinitely in whatever direction feels comfortable, regardless of what the actual results are showing.
Firms that adopt this rhythm for their first pilot tend to keep using it for every subsequent AI initiative, because it consistently produces a clear answer — yes, no, or adjust — rather than the ambiguous, stalled-out pilots that plague so many first attempts at AI adoption.
A 30-minute call is enough to tell you whether AI pays for itself here.
