AI change management is the hardest unbudgeted line item on every cabinet's agenda.

Every president we talk with knows two things at once. AI is the most consequential technology to hit higher education since the internet. And almost no campus is ready to absorb it — not because the tools are too hard, but because the change management is harder than anyone budgeted for.

The hard part is not picking a vendor. The hard part is getting 1,800 faculty, 600 staff, 12,000 students, three unions, two accreditors, and a board of trustees to a shared understanding of what the institution is going to do with AI — and what it is not going to do.

This is the playbook we wish every campus had on day one.

Why traditional change management breaks on AI

Higher education has a mature playbook for change. It is consultative, deliberate, committee-driven, and slow. That playbook works for curriculum reform, capital projects, and most ERP migrations. It breaks on AI for four specific reasons.

1. The technology moves faster than the committee cycle. A working group that meets monthly will be debating last quarter's model capabilities.
2. The locus of decision is unclear. Is generative AI an academic affairs question, an IT question, a legal question, a student-life question, a research-integrity question, or a marketing question? Yes. All of them. Simultaneously.
3. Bottom-up adoption is already happening. Surveys we ran in spring 2026 show 71% of undergraduates and 58% of faculty report using a frontier model weekly. Your policy is competing with established habits.
4. Risk is asymmetric and visible. A single FERPA leak, a single hallucinated transcript, a single biased admissions output makes the local news. The downside is loud; the upside is diffuse.

You cannot manage AI change with a single task force and a 90-day timeline. You need a permanent operating posture.

A four-track operating model

We recommend cabinets stand up four parallel workstreams, owned by named executives, with explicit decision rights. Not one task force. Four.

Track 1 — Policy and guardrails (Owner: General Counsel + CIO)

This track answers what is allowed, what is banned, and what requires approval. Outputs include:

• An acceptable use policy for generative AI by population (faculty, staff, students, contractors, researchers handling regulated data).
• A data-classification map that tells everyone, in plain language, what information can be pasted into which tools. Most campuses end up with a four-tier model: Public, Internal, Confidential (FERPA / HIPAA / GLBA-adjacent), and Restricted (research, IRB-governed, export-controlled).
• A vendor review pathway that is faster than the typical 9-month procurement cycle. Aim for a 30-day fast lane for tools that pass a pre-cleared security profile.

Track 2 — Teaching and learning (Owner: Provost + CTL Director)

This track answers how AI changes what and how we teach. It does not answer "is AI good or bad for learning" — that framing is a dead end. Outputs include:

• A syllabus addendum library with three to five canonical positions faculty can adopt: AI-prohibited, AI-permitted-with-attribution, AI-required, AI-as-tutor, AI-in-assessment.
• A faculty development arc — not a one-time workshop. A three-semester progression from awareness to fluency to redesign.
• A revised academic integrity policy that survives contact with the reality that detection tools do not work reliably.

Track 3 — Operations and workflow (Owner: COO / VPF + functional VPs)

This track answers how we use AI to do our administrative work better. Outputs include:

• A prioritized process map of where AI can take the next 10% of friction out of admissions reading, financial aid packaging, advising notes, IT helpdesk, HR screening, alumni outreach, and grant administration.
• A shared services AI platform so units are not each procuring their own copilot license at retail.
• A measurement framework: time saved, errors reduced, response-time improved. Not "AI projects completed."

Track 4 — Brand, recruitment, and AI search (Owner: CMO + VPEM)

This is the track most campuses forget exists, and the one that compounds fastest. Outputs include:

• An audit of how the institution appears in AI answers (ChatGPT, Gemini, Perplexity, Claude, AI Overviews). For most institutions in spring 2026, the answer is "incorrectly, incompletely, or invisibly."
• A structured-data and content strategy designed for AI retrieval (AEO/GEO), not just classical SEO.
• A brand narrative refresh that survives being compressed into a three-sentence AI answer.

Sequencing — the first 18 months

A common failure mode is trying to do all four tracks simultaneously, at full intensity, in semester one. Cabinets burn out, faculty senate rebels, the initiative stalls. Better sequencing.

What we see go wrong

A short, honest list of the failure modes we watch campuses walk into.

1. The single committee. One task force, monthly meetings, 18-month timeline, no decision rights. The technology laps them twice before they publish a report.
2. The vendor-led strategy. A cabinet lets a vendor's pitch deck become its AI strategy.
3. The ban-and-hope. A blanket prohibition on student AI use, unenforceable by design, that destroys trust without changing behavior.
4. The IT-only frame. Treating AI as an infrastructure project. The infrastructure matters, but the change is academic, cultural, and brand-facing.
5. The "we'll wait for the field to mature" stance. The field is maturing toward more capability, not stability. Waiting is a decision to fall further behind.
6. The faculty senate side-step. Decisions made without governance buy-in get rolled back the moment something goes wrong. Bring governance in early, in writing.

The cabinet conversation we recommend this month

Block 90 minutes. Walk the cabinet through five questions, in order.

1. Where is AI already being used on our campus, by whom, and with what data? Be ready to be surprised.
2. What is our institutional position on AI in teaching? Not what individual faculty will do — what the institution stands for. One paragraph.
3. Who owns each of the four tracks above? What is their budget and decision authority?
4. What is our 18-month definition of success? Faculty adoption rates, operational hours saved, AI-search visibility lift, student outcomes on AI-permitted assessments.
5. What is our communication plan to students, parents, faculty, alumni, and the board? Silence reads as either fear or neglect.

If you cannot answer all five at the end of the meeting, you have your agenda for the next three.

Bottom line

The institutions that look strongest in five years won't be the ones that picked the right vendor or wrote the best policy. They will be the ones that built the muscle to absorb continuous technological change without losing their academic identity.

AI is the first stress test of that muscle. It will not be the last.