An AI strategy people actually use is a short list of decisions the leadership team has already made for everyone else: what AI is for, where effort goes first, what staff may do today, who owns it, and how you will know it works. Written plainly enough to act on without asking permission, it changes what people do on Monday. A document that does not has not been used.
Most AI strategies are written for the people who approve them, not for the people who have to live with them. That is why so many are read once, praised in a board meeting, filed in a shared drive and never opened again.
I have sat in enough strategy sessions with senior leadership teams to know how this happens. Nobody sets out to write a document no one will use. The team is thoughtful, the language is careful, the ambitions are sincere. Then a teacher or a team leader opens it on a Tuesday afternoon with a real decision in front of them, finds nothing that helps, and does what they were going to do anyway.
An AI strategy that people actually use is a different kind of document. It is shorter, more specific, and much braver about saying no. This article is about how to build one.
What an AI Strategy That People Use Actually Is
An AI strategy people use is a small set of decisions the leadership team has already made on everyone's behalf, written down so clearly that a member of staff can act on them without asking permission. It is not a vision statement, and it is not a policy. A vision says where you hope to end up. A policy says what people must not do. A strategy says what you have chosen, in what order, and why, so that a hundred smaller decisions across the organization start pointing the same way. The line between the last two is worth holding onto, because a document that only prevents decisions is a policy whatever the cover says.
That distinction matters because "use" is a precise word. A strategy is being used when it changes a decision somebody would otherwise have made differently. A head of department decides to pilot AI for feedback rather than for lesson planning, because the strategy says feedback comes first. A manager declines a vendor demo because the strategy says the organization is consolidating tools this year, not adding them. A new starter knows which tools are approved, and what for, by the end of the first week. Nothing in that list requires anyone to reread the strategy. It has already done its work. If a strategy does not produce moments like that, it is not being used, however elegant it is.
Why Most AI Strategies Are Never Used
Most AI strategies go unused because they answer the leadership team's question rather than the staff member's question. The leadership question is "what is our position on AI?" The staff question is "what should I do differently on Monday?" A document can answer the first beautifully and the second not at all.
The evidence on this is uncomfortable. BCG's AI at Work survey, published on June 3, 2026, asked close to 12,000 frontline employees, managers and leaders in more than a dozen markets. Only a third of frontline employees said leadership's communications about AI were clear. Just 28 percent of respondents saw a strong connection between what leaders say and what the organization actually does, and 66 percent still receive limited or no guidance on what to do with the time AI saves them. Those are not adoption problems. They are strategy problems wearing an adoption costume.
Gallup's earlier work points the same way. In its December 2025 workplace report, based on 23,068 US employees surveyed in August 2025, 23 percent did not know whether their organization had implemented AI technology at all. The further people sat from decision making, the less they knew: 26 percent of individual contributors said they did not know, against 16 percent of managers and 7 percent of leaders. The strategy existed. It never reached the people it was about.
There is a second reason, and it is more awkward. Many leadership teams write the strategy before they have agreed a shared position. The document then becomes a way of avoiding the argument rather than settling it. Every hard choice is softened into a phrase everyone can live with, and the result offends nobody and directs nobody. When I run strategy sessions with executive teams and school leadership teams, this is usually where the real work is: not the writing, but the agreeing.
The Five Decisions a Usable AI Strategy Makes
A usable AI strategy makes five decisions explicitly: purpose, priorities, permissions, people and proof. Everything else is supporting material. If your current document is long, find each of the five in it; if you cannot find one in a sentence, it has not been made.
In the first Insights article I defined what it actually means to be AI ready this way: "Being AI ready means your organization can make, communicate and hold good decisions about AI, consistently, without depending on any one person." The strategy is where those decisions live.
1. Purpose: what AI is for here
The purpose decision states, in one or two sentences, what AI is for in this organization and what it is not for. Not "to embrace the opportunities of AI" but something a skeptical colleague could disagree with. "AI is for giving teachers time back for the parts of teaching only they can do, and for improving the quality of feedback students receive. It is not for replacing the judgment of a qualified teacher about a child." A school could argue with that. That is the point. A purpose nobody could argue with is not a decision.
This is also the first question in the Readiness Test: "The Readiness Test is three questions an organization must be able to answer consistently, from the top to the front line, before it can call itself AI ready: What is AI for here? What may I do with it? Who decides when it goes wrong?" If your strategy cannot answer the first of those in plain language, stop and answer it before writing anything else. If more than one of the three is shaky, the seven dimensions of AI readiness will tell you which part to work on first.
2. Priorities: where first, and where not yet
The priorities decision names the two or three areas where AI effort goes first this year, and names at least one area where it deliberately does not. Leaders find the second half of that sentence far harder than the first. Nobody enjoys telling an enthusiastic team their idea is a good one for next year. But a strategy with twelve priorities has none, and staff know it. The organizations I see making real progress have usually chosen fewer things than they wanted to, and then finished them.
A practical rule I give leadership teams: anchor the first priorities to friction that already exists. The best first uses of AI are attached to things people already find slow, repetitive or frustrating, because the pull is already there. Novelty needs pushing. Friction pulls on its own. Once the priorities are chosen, putting them in order is the job of a 12-month AI roadmap.
3. Permissions: what people may do, today
The permissions decision tells every member of staff what they may do with AI now, with which tools, and with what kinds of information, without asking anyone. This is the decision most often missing, and its absence does the most damage. Without it, two things happen at once. Cautious people do nothing, because they assume the answer is no. Confident people do whatever they like, because nobody has said otherwise, and that is how shadow AI starts. Both groups then believe they are following the strategy.
Permissions belong in the strategy, not only in the policy, because a policy is written to limit and a strategy is written to enable. The permissions section should read as an invitation with edges: here is what we want you to try, here is where the line is, here is who to ask when you are not sure. Keep the edges short enough to remember, which is the case for a short AI policy over a long one.
4. People: who leads, who supports, how capability grows
The people decision names who is accountable for the strategy, who supports its use day to day, and how the organization will build capability beyond a one-off training session. BCG's AI Radar survey of 2,360 executives, including 640 CEOs, published on January 15, 2026, found that 72 percent of CEOs now describe themselves as the main decision maker on AI, twice the share of the year before. That is a sign of seriousness. It is also a risk. A strategy that lives in the chief executive's head, or the headteacher's, fails the moment they are in a different room. Ownership has to be visible, and it has to be shared, which is easier once you separate the decision, the delivery and the expertise rather than handing all three to whoever volunteered. A steering group is where that sharing becomes concrete, and who sits on it decides which decisions can be made in one room.
The capability half of this decision is where most strategies become vague, with a line about "ongoing professional development" and nothing more. Be specific: who gets what, in what order, and what they should be able to do afterwards. And treat managers and middle leaders as the priority group, not an afterthought. Gallup's July 2026 engagement report found that employees who say their manager actively supports their team's use of AI had an engagement rate of 48 percent, against 30 percent for those who do not. Managers are the channel through which a strategy reaches people, or fails to.
5. Proof: how you will know it is working
The proof decision states what the leadership team will look at, and when, to judge whether the strategy is working, and what would cause them to change course. Usage numbers are the easiest thing to count and the least useful. Licenses activated and prompts sent tell you people are touching the tools, not whether feedback improved, time was given back, or anyone's judgment got better. The quickest way to sort a real measure from a flattering one is to ask whether the number could have risen while the work itself stayed exactly the same. Choose two or three measures that connect to the purpose in decision one, review them on a fixed rhythm, and write the date of the first review into the strategy itself.
What the Evidence Says About Strategy and Use
The research points one way: clarity about the plan goes with people who are more engaged, and the gap is largest where managers do the explaining. Gallup's July 21, 2026 report, drawing on 43,262 US workers surveyed in February and May 2026, found engagement 15 points higher among employees whose organizations have a clear plan for integrating AI than among those without one. Gallup does not say in the text I could open which employees that comparison covers, and a survey like this shows association, not cause. Read it as a gap worth closing, not a promise.
For school leaders in England, the Department for Education's AI in education roadmap (page last updated June 2, 2026) names six digital standards schools are expected to meet by 2030: broadband, wireless networks, network switching, digital leadership and governance, filtering and monitoring, and cyber security. They are infrastructure standards rather than AI standards, and in my reading the fourth is the one an AI strategy answers to. The same page records the international summit on generative AI in education that the department hosted in January 2026, and says education benchmarks for AI tools will be developed across 2026. None of that writes your strategy for you. All of it raises the bar for what "we are still thinking about it" will sound like to governors, trustees and inspectors. Across the government and system-level bodies I have advised, the worry is rarely whether schools should have a strategy. It is how to make one land across hundreds of them, which is a translation problem. The next section is about exactly that.
The Mistake I See Most Often
The mistake I see most often is a leadership team treating the strategy document as the finish line, when it is the starting gun. The document is signed off, the announcement goes out, and everyone quietly assumes the work is done. A strategy is used when people have heard it explained by someone they trust, in language that connects to their own work, more than once.
When I work with leadership teams on this, I normally push for a translation step before publication: take the five decisions and rewrite them, briefly, for each main group in the organization. What does this mean for a classroom teacher? For a finance team? For a head of year? For a new starter? If a decision cannot be translated for a group, it usually means it was never really a decision, only an aspiration in a suit.
The second most common mistake is close behind. Leaders write the strategy as if the skeptics were the problem. In my experience the skeptics are the most useful readers the strategy will ever have, because they will find every sentence that means nothing. Give the draft to your most doubtful senior colleague before you give it to the board, and read how to bring the skeptics with you before you do. If they cannot find a decision they disagree with, the document is not finished.
The Monday Test
The Monday Test is a single question to ask of any AI strategy before it is signed off: could a member of staff read it and know what to do differently on Monday morning? Run it with a real person in mind, not a category: a specific colleague, in a specific role, opening the document with a specific decision in front of them. Then read it as they would.
Most drafts fail the test in the same places. The purpose is there but the permissions are not. The priorities are listed but nothing has been declined. The ownership is implied but no name appears. Each failure is easy to fix once it is visible, which is the whole value of the test. It turns "is this a good strategy?", which invites debate, into "would Sarah know what to do?" (Sarah is a hypothetical colleague), which invites an answer.
The Monday Test scoring sheet
To run the test on a draft, give each of the five decisions a pass or a fail. Score it with a real colleague's name beside each row, not with your own opinion.
| Decision | Ask of the draft | It fails when | The fix |
|---|---|---|---|
| Purpose | Could a skeptical colleague disagree with it? | Nobody could object to it | Add what AI is not for here |
| Priorities | Is at least one good idea declined, by name? | Every idea made the list | Name what waits until next year |
| Permissions | Could a new starter say what they may do today, with which tools and which data? | The answer is "it depends" or "ask someone" | Write the yes list, the no list and who to ask |
| People | Does one named person own it, and does a manager explain it to each team? | Ownership belongs to "the leadership team" | Name the owner and book the first manager briefing |
| Proof | Is the first review date written in, with two or three measures tied to the purpose? | The measures are licenses or prompts sent | Choose a measure that cannot rise while the work stays the same |
Four or five passes and the draft is ready to publish. Two or fewer and you have a vision statement. If purpose or permissions fails, stop whatever the total, because those are the two that staff meet first. That scoring is my suggested rule of thumb, not a validated instrument.
What Good Looks Like
A good AI strategy is short enough to be remembered, specific enough to be disagreed with, and owned visibly enough to survive a change of leader. In the strongest organizations I have worked with, you can ask three people at three levels what AI is for here and get three versions of the same answer. Not the same words. The same decision.
There are other observable signs. Middle leaders can explain the priorities without looking anything up. Staff know which tools are approved and what for, and the approved list is short. Somebody has said no to a good idea recently, and explained why in terms of the strategy. The leadership team has a date in the diary to review the proof measures, and the last review changed something. New starters hear about AI in their first week from their line manager, not from a document.
None of these signs mention the document itself. That is deliberate. A strategy that is working shows up in behavior, not in a file.
How to Build It: A Practical Sequence
Build the strategy in this order: agree the position, make the five decisions, translate them, publish the shortest possible version, and put the first review date in the diary before anyone announces anything. The sequence matters more than the format, and if you are new to the job the same logic runs across your first ninety days.
Start with the leadership team's shared position, and do not skip this because it feels slow. Half a day in a room, with the honest disagreements on the table, saves months of a strategy nobody believes in. This is the part of the process I am most often asked to lead, and it is the part that leadership teams most often try to do by email.
Then make the five decisions in writing, each in a sentence or two, and run the Monday Test on every one. Translate for each group. Cut everything that is not a decision or the reason for one; background reading can live in an appendix nobody has to open. Publish, explain it in person through managers and middle leaders, and hold the first review when you said you would.
Finally, expect to revise it. Curiosity over fear, evolution over revolution: a strategy for a technology that changes every quarter should be a living document with a version number, not a monument.
The Next Step
If your leadership team has a strategy that reads well but is not changing what anyone does, or is trying to write one and keeps stalling at the hard decisions, this is the kind of work I support through AI strategy sessions with school leadership teams. I write each week about what makes a strategy survive contact with a real term, in the newsletter.
Sources and further reading
Sources last checked October 24, 2026.
- BCG, "AI at Work: Why Strategy Matters More Than Tools," June 3, 2026. bcg.com
- BCG, "As AI Investments Surge, CEOs Take the Lead" (AI Radar 2026), January 15, 2026. bcg.com
- Gallup, "AI Use at Work Rises," December 14, 2025. gallup.com
- Gallup, "Employee Engagement Remains Flat as AI Adoption Accelerates," July 21, 2026. gallup.com
- UK Government, "AI in education," Roadmap for modern digital government, page last updated June 2, 2026. gov.uk
Dan Fitzpatrick is the founder of The AI Educator and works with leadership teams, boards and government bodies on AI strategy and readiness. More about Dan.


