For cabinet & board

Leadership essentials

The decisions only leadership can make, the two constituencies that can actually stop you, faculty and students, and the honest parts most AI material skips: who owns it, what it costs, what not to do, and how it goes wrong.

Benchmarks and cost tiers here are pattern-based, not survey data. They describe what comparable institutions typically do, to calibrate expectations, not to assert measured outcomes for any specific campus.

The one slide you'll actually get

A one-page brief for the board

You rarely get thirty minutes. You get one slide and a hallway conversation. The whole model and the single 90-day ask, on one printable page you can forward before a meeting.

Open the one-page brief

What tends to matter, by institution type

Maturity varies enormously within every category, a community college can be well ahead of a research university. What's more predictable is where a given mission pulls its AI priorities first. Read this for the concerns and pitfalls typical of your type, not as a ranking of how far along you are.

Community & two-year college

Mission centers on access, affordability, and workforce relevance, with close ties to local employers and transfer pathways.

First energy: Access and workforce relevance, meeting students where they are, and employer-facing skills.
The trap: Letting AI access divide by who can pay, premium tools for some students, nothing for others, widening the very gaps the access mission exists to close.

High-leverage moveAnchor every AI decision to the access and workforce mission, and make equitable student access an explicit requirement, not an afterthought.

Regional public university

A broad-access mission across several colleges, with initiatives often running in parallel and a recurring need to connect them and clarify who decides.

First energy: Coordination, connecting scattered pilots and giving someone real decision rights.
The trap: Letting initiatives multiply across colleges with no shared decision path, so the same questions get re-litigated unit by unit.

High-leverage moveRun coordination through one cross-college body with real decision rights and fast lanes for low-stakes uses, so alignment scales without a bottleneck.

Large public research (R1 / R2)

Research-intensive and decentralized, with strong units and authority spread across colleges, which makes a single, uniform policy hard regardless of how advanced any one unit is.

First energy: Research integrity, funder compliance, and data governance across decentralized units.
The trap: Trying to centralize everything, a single mandate collides with college and faculty autonomy.

High-leverage moveGovern by data, behavior, and risk (not by unit), and give the senate a standing role in academic AI policy.

Private liberal arts college

Small and teaching-centered, with dense faculty governance and a strong values identity. Lean staffing shapes how work gets done more than it dictates how far along they are.

First energy: Teaching, academic integrity, and mission/values alignment, faculty voice is central early.
The trap: Letting the conversation become a referendum on banning AI rather than a values-led design choice.

High-leverage moveLead with mission and faculty co-design: adapt syllabus AI statements and ratify principles through governance.

Private research university

Research-intensive and complex, multiple schools, often a medical center, and central IT, where the recurring challenge is a shared operating model across them.

First energy: Risk, reputation, and a defensible portfolio across high-stakes systems.
The trap: Documentation theater, heavy process on low-risk tools while the genuinely high-stakes systems go untracked.

High-leverage moveBuild a registry and monitoring for high-stakes systems only; keep light paths for everything else.

Multi-campus system or district

Campuses sit at genuinely different readiness levels, and the system office is asked to set a floor without flattening those differences.

First energy: A shared floor, common principles and minimums, without overriding campus autonomy.
The trap: A one-size mandate that the most-advanced campus has outgrown and the least-ready can't meet.

High-leverage moveSet system-wide principles and a minimum bar; let each campus choose its own pace above the floor.

Who owns it, and what it costs

The first cabinet question is never the ideal end state. It's "who's accountable, and who pays?" Three places AI governance can live, and three honest effort tiers.

Where it can live

Provost-led

Best whenAcademic-mission-first institutions where teaching, integrity, and faculty governance are the center of gravity.

Keeps AI anchored to mission and academic authority; senate trusts the venue.

Can under-weight infrastructure, security, and operational AI. Pair with a strong CIO partnership.

CIO / IT-led

Best whenInstitutions where the pressing risks are data, security, procurement, and vendor-embedded AI.

Strong on tooling, data classification, and documentation; close to where AI actually enters.

Faculty may read it as a technology decision rather than an academic one. Give the senate a real seat.

Cabinet-chartered cross-functional body

Best whenMost institutions past the earliest stage, a standing group with delegated decision rights.

Balances academic, IT, legal, and student voices; survives leadership turnover better.

Becomes theater without a named owner, a charter, and the authority to actually decide.

What it takes, effort, not just dollars

Start-light

People: Part of one person's job
Spend: Essentially no new budget, time, not dollars
You get: A front door, an adopted one-pager, and acceptable-use guidance. Enough to prevent the obvious harms.

Nascent campuses, or anyone proving the lightest version works before asking for more.

Coordinated

People: A small cross-functional group, part-time; a named lead
Spend: Modest, reallocation, a little training, maybe an enterprise license already in hand
You get: Intake, risk-tiering, a baseline maturity read, and 1–2 visible initiatives run through the Compass.

Emerging campuses connecting scattered pilots into one coordinated effort.

Funded program

People: A dedicated owner or small office, with budget authority
Spend: A real line item, staff time, tooling, and monitoring for high-stakes systems
You get: A maintained registry, portfolio tracking, monitoring and incident response, and an annual reflect-and-reset cycle.

Established programs deepening and hardening for durability, not expanding for its own sake.

What not to do in year one

Everything looks important. That's the problem. The most useful thing a framework can do is tell you what to defer.

Stand up a big new standing committee first

Authority without anything real to govern becomes theater. Fold AI into a meeting that already exists, then formalize once there's a portfolio to manage.

Write a comprehensive AI policy before you've run anything

Policy written ahead of practice ages badly and invites litigation over hypotheticals. Start with guidance for the questions you actually face; let policy follow evidence.

Ban student AI use as the headline move

Bans are unenforceable, push use underground, and widen the gap between students who navigate the rules and those who don't. Set course-level expectations instead.

Document every system at once

A full registry on day one stalls under its own weight. Inventory the highest-risk handful first; let coverage grow with the program.

Chase maturity in pillars that don't match your risks

Effort should follow your mission's real exposure, not a desire for an even radar chart. Deliberately leave low-stakes areas light.

Buy a flagship platform to signal seriousness

Tools don't substitute for governance. A purchase made before there's an owner and acceptable-use guidance usually creates the very shadow-AI problem it was meant to solve.

How you'll know it's working, and how it goes wrong

Real signals over vanity metrics, and an honest catalog of the failure modes, because every campus AI story is told as a success until it isn't.

Signals it's working

Initiatives arrive through one intake instead of routing around governance.
Routine, low-risk uses get a decision in days, not weeks of review.
The senate and student government can name their role in AI decisions.
Your highest-stakes systems have a named owner and an appeal path.
People surface shadow AI to you instead of hiding it.
The program survives a leadership change without losing its cadence.

Where it goes wrong

Committee theater

A council meets, minutes accumulate, but no decision gets faster and nothing ships.

FixGive it delegated decision rights and a named owner, or disband it and fold the work into an existing body.

Ban-and-whack-a-mole

Each new tool triggers a new prohibition; students and staff move their AI use off the books.

FixGovern by disclosure, data, and human-in-the-loop, not by chasing product names.

Documentation theater

Heavy paperwork on harmless tools while a high-stakes model runs unregistered.

FixRight-size review to risk: real scrutiny for high-stakes systems, light paths for the rest.

The equity gap widening

Students who can pay for premium tools pull ahead; accommodations lag.

FixMake access and accessibility explicit success criteria, not afterthoughts.

Consultation-after-the-fact

Faculty are 'consulted' once the decision is effectively made; trust erodes.

FixDefine which decisions require senate input before, not after, they're settled.

Taking it to faculty senate

The senate can stop you. Here are the objections you'll hear, each steelmanned, with a response that respects faculty authority and what to bring to the room.

"This infringes on academic freedom and my classroom."

The real concernA central AI policy could dictate how a professor teaches or assesses.

The framework is explicit that decisions about teaching, assessment, and academic integrity belong to faculty through shared governance. Institutional policy sets the floor, data protection, disclosure, accountability, and leaves classroom choices to the instructor.

Bring: The shared-governance authority map and adaptable (not mandated) syllabus statements.

"An AI policy means surveilling students and faculty."

The real concernDetection tools and monitoring could be punitive and inaccurate.

Govern by disclosure and design, not surveillance. Lead with clear expectations and human judgment on any consequential decision; treat detection tools as unreliable and never as sole evidence.

Bring: The red/yellow/green acceptable-use grid and the human-in-the-loop requirement for high-stakes uses.

"You're moving too fast. This is consultation theater."

The real concernDecisions arrive as faits accomplis with token faculty input.

Name, in advance, which decisions require senate review and at what point. Academic AI policy goes through governance before adoption, not after. Put the consultation points in the charter so they're not optional.

Bring: A draft charter naming the decisions that require senate consultation.

"This is an unfunded mandate on already-stretched faculty."

The real concernMore committees, more service, no resources.

The recommended start is light by design, adapt one syllabus statement, join one existing conversation. Heavy process is reserved for genuinely high-stakes initiatives, and faculty effort is matched to real risk, not spread evenly.

Bring: The 'what not to do in year one' list and the start-light path.

"Banning AI is simpler and safer."

The real concernAllowing AI invites cheating and erodes rigor.

Bans are unenforceable and push use underground, which is less safe, not more. Clear, course-level expectations with disclosure preserve rigor while preparing students for a world that uses these tools.

Bring: Examples of tiered course-level AI expectations by assignment type.

"Why should faculty trust IT or administration to lead this?"

The real concernA technology office making academic decisions.

It shouldn't lead the academic decisions. The framework separates who governs the infrastructure (IT) from who governs academic use (faculty). The senate holds primary authority over the academic core; IT vets tools and documents systems.

Bring: The spheres-of-authority map showing faculty-led, shared, and administrative zones.

See the full shared-governance authority map

Students: where you'll take the most heat

Integrity, access, and accommodation are where the parent emails and the senate questions land. Four decisions to make on purpose.

Academic integrity, without an arms race

Set course-level expectations and lead with disclosure. Treat detection tools as unreliable signals, never as proof. Make the honest path the easy path.

Policies & Guidelines

Equity of access

Free and paid tools create a quiet advantage gap. Decide deliberately what the institution provides so capability isn't rationed by who can pay.

AI Principles

Accessibility & accommodations

AI can expand access for disabled students, or quietly exclude them. Build accommodations and accessibility review into adoption, not after.

Governance Guide

Clarity in the syllabus

Students mostly want to know what's allowed in this course. A clear, consistent AI statement removes most of the anxiety and most of the disputes.

Syllabus statements

AI you didn't choose to buy

Most campus AI didn't arrive through a decision. It was switched on inside tools you already license, the LMS, the CRM, the productivity suite, the ERP, or carried in by people using their own accounts. You govern it whether or not you chose it.

Place your campusFind your starting point Then sequence itThe first 90 days