A strategy has a no list.
The cheapest way to look like you have an AI strategy is to fund a portfolio of pilots and rebrand it. The signal that you have one is a written no list: the use cases, AI products, and teams you have decided not to bet on this year. McKinsey's 2026 State of AI survey found that only about 6% of respondents qualify as AI high performers. Those high performers are twice as likely as other respondents to have defined processes for measuring the impact of their AI work.
Without that, every team runs every experiment and no single investment gets enough depth to compound.
The eight steps.
Step 01
Name the business outcomes.
Three to five outcomes AI is meant to move in the next 12 months. Stated in business language: revenue, retention, cost, time-to-market. If you cannot say it without the word 'AI', it is not an outcome yet.
Step 02
Pick the bets.
Decide which use cases, teams, and AI products you are betting on, and which you are explicitly not. Write the no list. Without it, every team runs every pilot.
Step 03
Connect to customer value.
At least one bet must show up for the customer. Internal-only AI strategies underperform because no one outside the team feels the change.
Step 04
Map the people work.
Per role: what changes, what they need to learn, how their work is measured differently, and who their model user is. This step costs far more than generic AI training.
Step 05
Redesign the workflow.
Pick the workflows touched by each bet and rebuild them with AI in the loop. Layering AI on an unchanged workflow returns time-saved that disappears into more of the same output.
Step 06
Choose the platform.
Decide the AI stack here, once the workflows are clear. Build vs buy, sanctioned vs tolerated, who owns each. A strategy that leads with this step can stall before it reaches the workflow step.
Step 07
Set the measurement.
One adoption metric, one capability metric, one outcome metric per bet. Quarterly. Public to the team.
Step 08
Schedule the re-read.
AI moves quarterly. A strategy with no re-read date will be a year out of date in nine months.
Common failure modes.
- - Platform-first. Strategy starts with the platform choice. People and process get a paragraph at the end.
- - All-internal. Every metric is efficiency, so customers feel no difference and board interest fades by Q3.
- - Leadership exempted. The CEO and CMO talk about AI without using it. Adoption stalls at the manager layer.
- - No no list. Every team is funded, so no investment gets enough depth to compound.
- - One-and-done. Strategy shipped, never re-read. Six months later the AI model landscape has moved and the plan has not.
Where to start.
Before you write the strategy, get a read on where you are. The AI readiness assessment guide covers the six dimensions to score, and the free AI Profit Ready Assessment is the self-serve version of that read. The longer argument behind it sits in our book, The Elephant in the Algorithm.
Questions people ask.
What is an AI adoption strategy?
A written plan for how an organization will get value from AI through the people who use it. It covers which AI bets the organization is making, which it is not, how those bets connect to business strategy and customer value, and how the workforce will be brought along.
Is AI adoption the same as AI implementation?
No. Implementation is getting the AI stood up. Adoption is people using it on the right work, often enough to move an outcome. A failed AI program can implement cleanly and still stall at adoption.
How long should an AI adoption strategy take to build?
Four to eight weeks, depending on how much groundwork already exists. Under four weeks usually means the listening was skipped, and past eight the strategy starts to date before it ships.
What is the most common mistake?
Treating it as a technology strategy. The hard parts of AI adoption are how leaders model use, how workflows change, and how value is measured. Choosing which AI model to license is the smaller decision. A good strategy spends most of its pages on people and process.