Your Most Valuable Problem Is The Wrong Place To Start

Ask a room of executives whether their organization has started an AI project in the last two years and most hands go up. Ask them to keep their hands up if that project is finished, measured, and still running, and almost every hand comes down.

That gap is not a technology problem. The technology works and has for a while. The gap is a selection problem.

The instinct that causes it

Every leadership team wants to start with the biggest number. The eight-million-dollar inefficiency. The thing that would transform the business. It is the right instinct about value and the wrong instinct about sequence.

That project touches six departments, three systems and a compliance question nobody wants to open. By the time it half works, the sponsor has been promoted, the budget cycle has turned, and the organization has learned a lesson it will repeat for years: AI does not work here.

Four characteristics of a first project that finishes

High volume. It happens hundreds of times a week, not twice a quarter. Volume is what turns a small per-instance saving into a number your CFO can see.

Clear success criteria. You can state the number before you start, and you know the baseline. Teams skip the baseline because measuring is boring, then spend a year arguing about whether the project worked.

Tolerant of imperfection. An eighty-five percent answer is still useful. AI is probabilistic and will be wrong sometimes; choose a first project where wrong costs an apology rather than a filing.

Owned by someone who wants it. This is the dimension teams score most dishonestly and the one that predicts failure best. If you had to assign it, you have already lost. Find the person who is frustrated by this problem today.

Three projects to refuse

I build custom AI for a living and I decline work regularly, for these three reasons.

Zero tolerance for error. If one wrong answer is a regulatory event or a safety incident, a human stays in the loop. AI can draft, sort and retrieve inside that workflow. It does not get the final decision.

No clean source of truth. If your own experts disagree about the correct answer, there is nothing to learn. That is a process problem wearing a technology costume.

Nobody owns the outcome. A project with no named beneficiary ends the day the enthusiasm does.

The reframe

Your first AI project has exactly one job, and it is not to transform the business. It is to finish, show a number, and earn you the right to do the second one. Boring projects that finish beat ambitious projects that do not.