Use is outrunning governance.
Six in ten teachers used AI last school year. Fewer than one in five work in a school with a policy for it. The gap is an opportunity as much as a risk.
In a 2025 survey of more than 2,000 U.S. public school teachers, Gallup and the Walton Family Foundation found that six in ten had used AI tools in their work during the 2024–25 school year. Only 19 percent said their school had a policy on AI.1
The pattern looks similar in higher education and workforce training. Faculty, instructional designers and staff are adopting AI on their own, faster than their institutions can decide how it should be used. That isn’t a failure of anyone’s diligence. It’s what happens when a powerful, general-purpose tool arrives overnight and immediately saves people time.
Why a policy isn’t enough
The natural institutional response is to write a policy: what’s allowed, what isn’t and what must be disclosed. Policies matter. But a policy on its own doesn’t change what the tools know.
When a faculty member asks an AI assistant to draft a unit, the assistant doesn’t know the program’s outcomes, the accreditor’s standards, the prerequisites students arrive with or the institution’s expectations for accessibility and voice. Whatever the policy says, the draft starts from generic assumptions, and someone has to check it against all of those by hand. That checking is exactly the work that gets skipped when time is short.
Governance as shared context
We think the more durable approach is to treat governance less as a set of rules about tools and more as shared context that every tool can use. What an institution intends to teach, the requirements it has to meet and the decisions it has already made should live in one place that people and AI tools both draw from.
When that context exists, a few things change:
- Drafts start aligned. An assistant working from a program’s outcomes and standards produces material that already reflects them.
- Review gets faster. Faculty and designers spend their time judging and improving, not re-checking the basics.
- Decisions are traceable. When someone asks why a course says what it says, the answer is a record, not a recollection.
- Changes reach every course. When a standard is revised, the affected courses can be found and updated, instead of discovered at the next review.
Where to start
Institutions don’t need to solve this everywhere at once. A practical first step is a single program or department:
- Write down what the program is for: its outcomes, its audience and the requirements it answers to.
- Connect its courses to that record, so each piece of content can be traced to the outcome it serves.
- Give the AI tools faculty already use access to that context, and ask them to work within it.
- Keep people in charge of what’s published. Automation should speed up the work. It shouldn’t decide what’s taught.
The gap between use and governance won’t close by slowing adoption down. It closes when the tools people already rely on know what the institution is trying to do.
- Gallup and the Walton Family Foundation, Teaching for Tomorrow: Unlocking Six Weeks a Year With AI, June 2025. Survey of 2,232 U.S. public school teachers. ↩