Seven in 10 provosts use AI every week. One in 10 has a strategy.

The people who run academic life at American colleges have made AI part of their week. Their institutions mostly have not made it part of a plan.

That is the plainest reading of Inside Higher Ed’s 2026 survey of chief academic officers, published on September 23. It drew answers from 376 provosts, mostly at public and private nonprofit institutions.

What the survey found

Personal use is common. Roughly seven in 10 provosts use AI in their own work at least once a week. The most common uses are ordinary office work: 71 percent use it to summarize documents and reports, 66 percent for presentations and meetings, and 65 percent to draft communications to faculty, staff or students.

Institutional planning is thin. Only one in 10 provosts reports a centralized AI strategy. One in five agrees (17 percent) or strongly agrees (3 percent) that their institution has a coherent vision for how AI will change what it teaches and how to teach it. Fifteen percent say their institution has not invested in AI at all.

The returns so far are about efficiency. More provosts see clear gains in individual employee productivity (39 percent) than in teaching and learning (29 percent) or research (7 percent). Only 12 percent say AI investment has optimized department workflows, and 8 percent report institutionwide operational transformation.

The obstacles are people, not software. Staff capacity and skills (51 percent) and faculty or staff resistance (50 percent) top the list, followed by concerns about how AI affects students (39 percent). And fewer than half of provosts (47 percent) say they have what they need to implement initiatives.

Why the gap matters

When the leader’s own use runs ahead of the institution’s plan, it invites three problems. Each department writes its own rules, so students meet different expectations from one course to the next. Purchasing happens tool by tool, so nobody can say which contracts cover student data. And the one place AI could matter most, teaching, gets the least deliberate attention, because it is the hardest to change and the easiest to postpone.

The survey’s barrier numbers point to the fix. If the limits are capacity and resistance, a strategy that starts with software will stall. One that starts with a few named people, a small budget of their time, and a clear question about teaching has a chance.

A one-page strategy in five lines

A strategy does not have to be a 40-page document. For most institutions, one page that answers five questions is enough to start, and it is more likely to be read.

1. The teaching question. One sentence on what you want AI to change, or protect, in how students learn. Example: “Every gateway course states, per assignment, what AI use is allowed and how students disclose it.”
2. The owner. One named person who can convene deans and IT, with a few hours a week protected for it.
3. The data rule. Which kinds of student and staff information may go into which approved tools, and who reviews new contracts.
4. The capacity plan. Who gets trained this term, on what, and whose workload drops to make room. If no one’s load drops, the plan is a wish.
5. The evidence. Two or three measures you will check at the end of the term, set now, before anything changes.

A decision rule for moving from personal to institutional

Many provosts already have a use that works for them. Here is a simple rule for when one should become an institutional practice:

• Promote it when it saves time on recurring work, uses no student records, and a colleague can repeat it from a short written guide.
• Pilot it when it touches student records or grades. Run it with one department, under a reviewed contract, with a person checking the output.
• Hold it when you cannot say who checks the output or where the data goes.

The rule is deliberately cautious about student information. Privacy and records questions carry legal duties, so bring your counsel and registrar in early. This post is general education, not legal advice.

What to do this week

• Write your own five lines. Do it in an hour, alone, before any committee sees it.
• Pick one owner and ask what would have to come off their desk to make room.
• List every AI tool your institution pays for, and mark which ones have a contract that covers student data.
• Choose one gateway course and ask its instructors to state the AI rule on each assignment this term.
• Set two measures for December and write them down now.

The survey shows provosts are not waiting to learn the tools. The work now is to make what already helps one leader useful to the whole institution, in a way that faculty trust and students understand.