How to build an AI strategy that isn't a list of pilots
By Erick Truong · · 3 min read
The short answer
A good AI strategy fits on one page. It ties AI to two or three business outcomes that matter this year, names one accountable owner, sets a baseline, ranks the first problems by impact, readiness, effort and risk, and puts guardrails in before anything is built. If it doesn't fit on a page, it's activity, not strategy.
What is an AI strategy, really?
An AI strategy is a mandate: a short, specific statement of what AI will change in the business, who owns that change, and how you'll know it worked. It is not a list of pilots, a tool roadmap or a slide of use cases.
The test is simple. Hand the page to someone who wasn't in the room. If they can tell you what you're doing, who owns it and how you'll know it worked, you have a strategy. If they can't, you have a list of pilots.
What goes on a one-page AI strategy?
Six sections, filled in by the leadership team in a single sitting.
1. The AI ambition
One paragraph. Two or three business outcomes that matter this year, in language a board would accept. If it doesn't fit in a paragraph, it's activity.
2. The accountable owner
One executive who owns the outcome, by name. Not a committee. Ownership without a name is how initiatives stall between departments.
3. The baseline
The metric, where it is today, and the target with a date. No baseline means no honest "after," and no way to prove value later.
4. The first three problems, ranked
Map the real workflows, then score candidate problems from 1 to 5 on impact, readiness, effort and risk. Disagreement surfaced now is cheap; disagreement discovered mid-build is not.
5. Guardrails before you build
Acceptable use, a named risk owner, vendor criteria, human review where output meets a customer, monitoring after launch, and a way to measure. Every one you can't tick is a place a rollout stops. Start with a one-page acceptable use policy.
6. The first 90 days
What ships, who owns it, and what proves it worked. Concrete enough that someone outside the room could check it.
My free one-page AI strategy template lays out all six sections. It covers the first three stages of THE MASTERS FRAMEWORK: Map, Assess and Shape.
What are the most common AI strategy mistakes?
- Starting with the tool. Choosing a platform before mapping the work leaves you hunting for uses.
- Committee ownership. A steering group can advise. One named executive has to own the outcome.
- Chasing impressive demos. Pick problems that move a metric someone is accountable for, even if they're less exciting.
- Skipping the baseline. Without today's number, every result is an anecdote.
- Treating governance as the last step. Guardrails added at the end are what stall rollouts. Why AI pilots fail covers this in detail.
How do you measure whether an AI strategy is working?
Against the baseline you wrote down before building. Track the business metric itself, such as cycle time, cost per transaction, error rate or revenue per rep, not activity counts like logins or prompts. A hundred logins that produce work people redo isn't adoption.
Review the page quarterly. Retire problems that didn't move the number, scale the ones that did, and rank the next three.
Frequently asked questions
- Who should write the AI strategy?
- The leadership team, together, in one working session, with the accountable owner holding the pen. A strategy written by one function and handed to the others rarely survives first contact with the business.
- How long should an AI strategy be?
- One page. The discipline is the point: it forces you to choose outcomes, an owner and the first problems, instead of listing every possible use case.
- How often should we update it?
- Review it quarterly against the baseline. Keep the ambition steady for the year, and refresh the ranked problems and the next 90 days as results come in.