Keep your AI strategy to one page: seven rules that force the decision

Most AI strategies are a wish list photographed in the vendor aisle. Seven rules for getting yours down to a single page your team can actually act on, with a fill-in template and the one-page test.

Ask ten businesses for their AI strategy and most will hand you a slide deck of intentions: become AI-first, upskill everyone, build a copilot for the whole company. None of that is wrong, but it is not a strategy, it is a wish list someone photographed in the vendor aisle. A strategy has to say what you will do first, what you will not do yet, and how you will know it worked.

A single page is what forces those answers out. It strips away the logos and the aspiration and leaves you with the handful of bets you are actually willing to be measured on. Nicholas Kelly makes the same case in his new book, The AI-Driven Data Team: the constraint is rarely the technology, it is the discipline to choose. Here are seven rules for getting your AI strategy down to something a busy team can read before the coffee goes cold, and then go and do.

Start with the decision, not the model

The point of AI in your business is not to be impressive, it is to make specific decisions faster or safer: which quotes to prioritise, which invoices to chase, where to hold stock, who to put on tomorrow’s rota. Name two or three of those before you name a single tool. Everything else on the page exists to serve them, and anything that does not trace back to a decision you have written down comes off the page.

Put the skills gap on the page, honestly

Kelly is blunt that capability, not software, is what actually limits most teams, and a one-page strategy should say so out loud. One line is enough: what your people can do today, what they cannot, and how you are closing the difference. This is what stops you buying a capable platform that then sits idle because nobody was equipped or expected to use it.

Extend the stack you already have

You almost certainly do not need to rip anything out. The higher-return move is to put AI on top of the model you already govern in Power BI or Fabric, so the new capability inherits the data, the definitions and the permissions you have already paid for. Your page should say what you are extending, not what you are replacing. Rip and replace is rarely a strategy, it is usually a delay wearing a strategy’s clothes.

Ship a few pilots on a ninety-day clock

Ambition without a deadline drifts. Pick two or three pilots, weighted toward clear revenue or cost, and give each one ninety days to prove or disprove itself. The short clock is the feature, not the compromise: it forces scope down to something you can actually ship, and it gives you permission to stop the ones that are not working without it feeling like a failure.

Write governance as guardrails, not gates

Governance earns its line on the page when it lets people move faster with confidence, not when it becomes a queue. Anchor it to one recognised standard, ISO or NIST, and design it so the safe path is also the obvious one. The test is simple: a good decision should be makeable in a day, not held for a fortnight while it waits for three signatures.

Put a number next to everything

If an initiative cannot carry a number, it is a hope, not a bet. Every line on the page should have a figure attached: the cost it removes, the revenue it supports, the hours it hands back, the error rate it lowers. The number does two jobs at once, it forces you to be specific about value, and it gives you an honest basis for killing what underperforms before it quietly becomes permanent.

If it does not fit on a page, you have not decided yet

Treat the page limit as the discipline, not the inconvenience. When you cannot make it fit, the problem is almost never the margins, it is that you are still trying to do everything at once. Cut until what remains is what your team can act on this quarter. A one-page strategy that gets executed beats a thirty-page one that gets admired in a drawer.

What the page actually looks like

If it helps, this is the shape we would put in front of a mid-market leadership team. Fill in the blanks and you have your page:

• The bet. In one sentence, what are we backing AI to change this year.
• Three decisions we are improving. Named, with the person who makes each one.
• What we are extending, not replacing. The existing platform the work rides on.
• Two or three ninety-day pilots. Each with an owner and a number.
• The guardrail. The one standard we hold to, in a line.
• What we are not doing yet. The tempting things we are deliberately parking.

The one-page test

Hand your draft to someone who was not in the room. If they can tell you, in their own words, what you are betting on, what you are not doing, and how you will know it worked, you have a strategy. If they cannot, you have a document. This is exactly the kind of plan our Strategy and Leadership work exists to write alongside your team, and it starts the same way our data strategy work does, with the decisions, not the tools.

If you would rather not start from a blank page, that is exactly what we do. Get in touch and we will help you turn your decisions into a one-page AI strategy your team can act on.

An AI strategy is not a document you finish. It is the shortest list of bets you are willing to be measured on.

Written with reference to Nicholas Kelly’s The AI-Driven Data Team: Improve Your Analytics with AI (Kogan Page, 2026).