Same questions, smaller answers
For departments, offices, and small teams: the eight pillars at the size a unit can carry.
The eight pillars name what must be considered. They don't set how big the response has to be. A unit of five answers the same questions as a whole campus, with smaller answers.
Inherit first
- 1Start with what the campus provides
Approved tools, data classification, acceptable-use guidance, security review. Use it as-is.
- 2Fill only local gaps
Write down what's specific to your work, nothing the campus already covers.
- 3Hand up what's bigger than you
Sensitive data, student-facing decisions, new vendors, and anything the escalation questions flag go to campus governance.
Eight minimal artifacts
Three escalation questions
Most unit-level work belongs in the Compass low-risk lane. Hand it up if any answer is yes:
- 1
Does it touch sensitive data (student records, health, personnel, research)?
- 2
Does it inform a decision about a person (admission, grade, aid, hiring, conduct)?
- 3
Will people outside the unit see its output or rely on it?
All no: proceed in the Strategic Compass low-risk lane. Any yes: hand it up to campus governance.
Quick self-check by pillar
Three questions per pillar. Answers stay in this browser. Any "Partly" or "Not yet" points to that pillar's minimal artifact.
Can you say in one sentence what AI is for in your unit?
Does that sentence tie to your unit's mission?
Does your team know it?
Do you know where the campus AI principles are?
Have you named 2–3 local principles for your work?
Has a recent AI decision been checked against them?
Can everyone find campus AI guidance in under five minutes?
Do you have one page of local guidance, or a pointer plus your additions?
Does it say what's off-limits?
Does your team know which of its data is sensitive?
Do you keep a list of the AI tools in use, and whether each is approved?
Do you know who to ask before a new tool touches unit data?
Is someone connected to a campus community of practice or a peer unit?
Does the team have a regular place to share what's working?
Have the people affected by an AI use been asked what they think?
Do you know who in the unit uses AI, and for what?
Do you know who needs help to get started?
Has anyone had basic AI training in the past year?
Is one person named to decide on AI uses in the unit?
Is someone named to check AI output before it's used?
Does everyone know who those people are?
Does each AI use have a stated reason to start?
Does each have a signal you watch to see if it's helping?
Does each have a condition that would make you stop?
Compass check for one AI use
Before starting a specific use, answer six questions. The first three test the value case; the last three are the escalation questions.
Does it solve a problem the unit has now?
Can you tell within 30 days whether it helped?
Can you stop or undo it easily?
Does it touch sensitive data (student records, health, personnel, research)?
Does it inform a decision about a person (admission, grade, aid, hiring, conduct)?
Will people outside the unit see its output or rely on it?
What you can set aside
- A formal maturity assessment
- Portfolio review across initiatives
- A standing AI committee
- Algorithmic impact assessment, unless an escalation question triggers a hand-up
The first 30 days
Find the campus guidance, approved tools, and data classification. Adopt them.
Write the purpose sentence and the one-page local guidance.
List the tools in use, what data each touches, and which are approved.
Run one actual task with AI, with a named checker reviewing the output.
Connect, don't isolate
Use what the campus already provides instead of rebuilding it.
Send anything the escalation questions flag to campus governance.
Report what worked and what didn't so the next unit starts further ahead.