Why the lighthouse build belongs in the first 90 days.

There is a fair objection to capability-first. If you spend the first months building skills, when does the business actually see value? The answer is the lighthouse, and it belongs in the first 90 days.

A lighthouse is one use case, chosen deliberately, that is both high in value and genuinely winnable. It is not the whole transformation. It is the single, visible proof that the approach works, delivered early, while the broader capability is still being built underneath it. Capability-first does not mean value-later. It means value-early, in one place, done properly.

The sequencing is backed by the evidence. MIT's 2025 research found that the AI projects most likely to pay off were not the glamorous customer-facing ones. They were back-office tasks, the slow, repetitive, internal work, because that is where the return is highest and the risk is lowest. The same research found adoption succeeds when line managers drive it, not a central lab, and when organisations buy or partner rather than try to build everything themselves. A good lighthouse honours all three. It picks a high-return internal task, it is owned by the team that does the work, and it is built with the simplest thing that helps.

Choose it well. A good lighthouse is a task that happens often, takes too long, and matters to someone senior. You time it before, you build the smallest thing that moves the number, and you time it after. A report that took two hours now takes thirty minutes. A quote that took half a day now takes an hour. That is not a promise in a slide. It is a number a sponsor can take to the board, and it is what funds and protects everything that follows.

The order, then, is simple. Prove, scale, embed. The lighthouse is the proof, and it comes first, inside the first 90 days, because a programme that has shown one real win has earned the right to build the rest.

Source: MIT NANDA, The GenAI Divide: State of AI in Business (2025).