Your independent director waits until the last ten minutes of the meeting. Then she asks it: "What is our AI plan?"
The room does what rooms do. The CFO mentions a tool finance is trying. Operations says IT is looking into it. You say the company is being careful. She writes something down and says she will ask again next quarter.
That moment is coming to most owner-led companies, from a board, a lender, a buyer or a new operating partner. The answer that fails is a list of tools. A tool list is a shopping list. It does not say who decided, who is accountable, or what the company has ruled out.
The answer that holds up fits on two pages.
Page one: what is already happening, and who is doing it well.
AI is probably in use in your company right now, and nobody decided it. A vendor switched on a new feature in software you already pay for. Someone in scheduling pasted a customer list into a public tool to draft emails. Someone in accounts built a small automation that runs on every inbound invoice. Page one lists these in plain words: who is doing what, and which company information it touches. It is not pleasant to write. It is the most useful page, because a plan that ignores what is already running is a plan for a company you do not have.
Page one also names two people. In most companies this size there are two or three employees who already use AI well without being asked, and colleagues who walk over to them when they are stuck. Name them. Write one paragraph each: what that person may build, what they may not, and who checks the result.
Those two are your plan. Nufar Gaspar, who has set up programs like this inside large companies, calls the right person "peer plus one": inside the team, doing the team's work, one step ahead. The nominations that fail are the familiar ones. The CFO nominates himself, but he belongs on the steering side. The hardest-working supervisor has no spare hours. The new hire from a software company does not yet know how your work really happens.
Page two: what the company has decided.
Three short lists.
What it must never do. Write this first, because it settles who checks, how you test, and who it is for. For example: it never sends a quote below your price floor without a named sign-off. It never sends customer data outside the company except through a route you approved. It never decides credit or a termination without a person.
Who owns the day after. One name for the plan. One name for month two, when the excitement is gone and someone has to carry what the two people built to the rest of the team. This is where most plans quietly die, because it is nobody's job.
How you will know it worked. One measure, read on a date: how many people are using the shared method ninety days from now.
You keep every decision on those pages. The paper only makes them visible before a board member asks.
A note on what this is. It is my judgment from running a company through a period of heavy growth in a regulated industry, and from building AI systems every day since. It is a working hypothesis about what a board will accept, not a measured result across companies. If it is wrong for yours, the two pages will show that quickly, which is also useful.
If you want page one written for your company within a week, with the two names and their written scopes, that is the AI Board Briefing, from $2,500. Book thirty minutes and tell me the question your board is asking: https://calendar.app.google/ZRFppsyHiGMCeozTA
Darren
Want to see where you stand? Take the 3-minute AI readiness scorecard.