Building an AI council on campus — governance that survives the first hard call.

A standing AI council is the single highest-leverage governance investment a campus can make this year. Not a task force. Not a working group. A permanent body, chartered by the cabinet, with explicit decision authority and a published cadence.

We have helped institutions stand them up. Here is what works.

What the council is for

An AI council exists to answer questions that no single executive can answer alone, in the time frame the technology demands. Its purpose is decision velocity at acceptable risk, not consensus building.

The council exists to:

• Approve, deny, or escalate AI vendor procurements above a threshold.
• Issue institutional positions on emerging AI-policy questions (academic integrity, research integrity, brand voice, equity).
• Maintain the institutional risk register for AI-related exposures.
• Brief the cabinet and the board quarterly.
• Be the named owner when an external party (accreditor, journalist, attorney) asks "who at this institution owns AI."

What it is not for:

• Reviewing every individual faculty syllabus.
• Approving every staff AI use case.
• Running the procurement process itself.

Membership

Eight to ten members. Larger councils don't decide. Smaller councils miss perspective.

The chair should be the provost or a designated executive sponsor — not the CIO. The reasoning: AI is fundamentally an academic and institutional question first, an infrastructure question second. The CIO is essential at the table; they should not chair.

Charter

A one-page charter, signed by the president, that says:

• The council exists.
• Its scope is institution-wide AI policy, procurement above a threshold, and risk.
• It meets monthly, with a 5-business-day emergency convening capability.
• Its decisions are binding on the institution, with appeal to the president only.
• It publishes a quarterly summary internally and an annual public summary.
• Membership rotates on a fixed cadence — most positions are 3-year terms.

If the charter is more than a page, the council is being set up to fail. The legalese can live in an addendum.

Cadence

Monthly meetings, 90 minutes, structured agenda.

Every meeting produces three artifacts: an updated risk register, a decision log, and a one-page communication for the rest of the institution.

The four early decisions

Every AI council we have stood up faced the same four questions in its first 90 days. The councils that handled them cleanly earned credibility for years. The ones that hesitated lost ground they didn't recover.

Decision 1 — The institutional position on student AI use

Not a detailed policy. A one-paragraph position that every faculty syllabus can reference. The position should be adoptable, defendable, and humane. The bad versions are "AI is prohibited" (unenforceable, dishonest) or "AI is at faculty discretion with no guidance" (chaos, inequity across courses).

The good versions name the four positions a syllabus can take and require faculty to declare one.

Decision 2 — The institutional position on AI in admissions

This is the question every counsel will field from journalists in 2026 and 2027. The position must be specific. Examples that hold up:

• "We use AI to prepare application materials for human readers. We do not use AI to make admission decisions. Every decision is made by a human reader operating against a documented rubric, with an inspectable audit log."

• "We use AI to triage application questions and respond to common process inquiries. We do not use AI to make admission decisions. Application content is not used to train any model."

If the institution cannot make a specific written statement, the answer is: don't deploy in admissions yet.

Decision 3 — The institutional data-classification and vendor-approval policy

Tied to the security framework. The council ratifies the four-tier data classification, the vendor checklist, and the procurement fast lane. Once ratified, individual decisions move much faster — the framework does the work, the council only has to handle the edge cases.

Decision 4 — The communications posture

Who at the institution speaks publicly about AI. What is on the public website. What is in the FAQ. How the institution responds when a journalist asks for comment.

The council does not need to be the spokesperson. It needs to designate the spokesperson, equip them with current talking points, and approve revisions quarterly.

What separates councils that work from councils that don't

We have seen both. The councils that work share four characteristics.

1. They actually decide. They do not punt. They take votes, log them, and live with the consequences.
2. They publish internally. Faculty, staff, and students know the council exists and can see (at the right level of detail) what it has decided.
3. They escalate cleanly. When something is above their level, they bring it to the president or the board with a clear recommendation and the data to support it.
4. They are honest about uncertainty. When a decision is provisional, they say so, set a review date, and stick to it.

The councils that fail share two characteristics. They meet to discuss without deciding. And they avoid the controversial calls by referring them to committees that don't have decision authority either.

Bottom line

The institutions that look strongest in five years are not the ones with the best AI policies on paper. They are the ones whose AI council has been making consistent, defensible, well-documented decisions for thirty months.

Stand one up. Charter it cleanly. Staff it with the right eight people. Meet monthly. Publish quarterly.

Most of the AI risk a campus faces over the next decade gets handled at that table, or it doesn't get handled at all.