AI Literacy / AI policy
Does your community college need an AI policy?
Yes, but probably not the one you are about to write. The sequence matters more than the document.
Most colleges need an interim set of AI guidelines within a term, and a full policy within a year. The mistake is writing the policy first. A policy drafted by people who do not yet understand what the technology does will either ban too much, permit too much, or be so vague that nobody can apply it.
The order that works: build literacy across the groups the policy will govern, publish short interim guidance so people are not operating in a vacuum, then write the durable policy with faculty, staff, and students who now understand what they are approving.
Why the policy-first approach fails
The pattern is consistent. A president asks for an AI policy. A committee forms, usually drawn from IT, academic affairs, and whoever volunteers. Within three meetings the group discovers that its members hold irreconcilable assumptions about what AI is, and the conversation stalls on definitions rather than decisions.
What eventually gets published is one of three documents. A prohibition, which faculty ignore and students route around. A permission, which creates FERPA and academic integrity exposure nobody assessed. Or a statement so general that a department chair facing an actual case cannot tell what it requires.
None of these are failures of drafting. They are failures of sequence. A group cannot govern a technology it has not been taught, and no amount of careful wordsmithing substitutes for that.
An institution cannot write a credible AI policy until the people affected by it understand what they are being asked to approve.
The sequence that works
Four steps over roughly nine months. The first is the one institutions skip.
| Step | What happens | Why it comes here |
|---|---|---|
| Month 1-2 | Shared literacy | Cabinet, senate leadership, and the eventual policy group work through the same foundational training. The goal is a common vocabulary, not expertise. Without it, every subsequent meeting relitigates definitions. |
| Month 2-3 | Interim guidance | Two pages, published fast, covering the questions people are already facing: student use in coursework, staff use with student data, and what must not be entered into a public tool. Explicitly labeled interim. |
| Month 3-7 | Consultation and drafting | Academic senate, classified senate, student government, and the bargaining units. This is where a policy earns legitimacy, and it cannot be compressed without costing you the legitimacy. |
| Month 7-9 | Adoption and review cycle | Board adoption, with a scheduled revision date. Any AI policy without a review date is obsolete before it is approved. |
The interim guidance step is the one that buys you the time to do the rest properly. Without it, pressure to publish something forces the full policy through in six weeks.
What belongs in the policy, and what does not
- Principles and decision rights - who approves what
- Data classification: what may never enter a public tool
- Academic integrity expectations and syllabus requirements
- Disclosure obligations for staff and administrative use
- Accessibility and equity requirements
- Procurement and vendor review triggers
- A named owner and a scheduled review date
- Named products, which change faster than policy can
- Specific prompts or usage instructions
- Course-level rules, which belong to faculty
- Detection tooling, which is unreliable and contested
- Technical configuration details
The distinction matters practically. Anything naming a vendor or a tool version will require board action to amend. Keep those in guidance a dean can revise.
Five mistakes we see most often
- Writing it in ITAn AI policy is an academic and personnel policy that happens to involve technology. Drafted in IT, it reads as a systems document and gets no faculty buy-in.
- Skipping the bargaining conversationAnything touching evaluation, workload, or monitoring has bargaining implications. Discovering that after adoption means starting over.
- Banning detection tools, or mandating themBoth positions are currently indefensible. Detection accuracy is contested enough that policy should not depend on it either way.
- No student voiceStudents are the most affected group and the most experienced users. A policy written without them is both less legitimate and less accurate.
- Treating it as doneA policy adopted without a review date will be quietly ignored within eighteen months as the technology moves past it.
The California context
For California community colleges specifically, three things shape the work. Participatory governance under Title 5 means the academic senate has a defined role in anything touching curriculum and academic standards, and skipping that step invalidates the process regardless of the policy's merits. The Brown Act governs how the board discusses and adopts it. And collective bargaining agreements often speak to evaluation and workload in ways that constrain what the policy can require of faculty and staff.
None of this makes the work harder in substance. It makes the sequence less negotiable. Institutions that treat participatory governance as a step to satisfy after drafting tend to spend more total time than those that build the policy inside it.
Who does this work
Former chancellors, presidents, and system executives who have taken policy through participatory governance and elected boards.
How long does it take to write an AI policy for a college?
Nine months is realistic for a durable policy that has been through participatory governance, with interim guidance published in the first two or three months so people are not operating without direction while the full policy is developed. Compressing the full policy below about four months usually means shortcutting consultation, which costs you legitimacy later.
Should the academic senate or administration own the AI policy?
Both, in defined roles. Anything touching curriculum, academic standards, or student academic conduct falls within the senate’s purview under participatory governance. Data handling, procurement, and administrative use are management responsibilities. Most failed policies tried to put the whole thing on one side of that line.
What should an interim AI guidance document cover?
Three things, in two pages: what data must never be entered into a public AI tool, the default expectation for student use in coursework along with the fact that individual faculty may set stricter terms, and who to ask when a situation is not covered. Label it interim and give it an expiry date.
Do we need a separate policy for student use and employee use?
One policy with distinct sections is usually cleaner than two documents, because the underlying principles and data rules are shared. The obligations differ enough that the sections should be written separately, but splitting them into separate policies tends to produce contradictions.
Can we adapt another college’s AI policy?
You can and probably should read several, but adopting one wholesale fails for a specific reason: the value of the policy is largely in the consultation that produced it. A borrowed document has the words without the agreement, which is why it does not change behavior.
Start with the literacy, not the document.
The first AI Foundations module is free and takes about ten minutes. It is the fastest way to see whether this would give your policy group the common vocabulary it needs.
Related