A school that acquires a library does not just acquire books. AI is the new library — almost every school has the books and very little of the rest. Two questions follow, and every engagement answers both: is the library working, and can the institution carry what having one implies?
Is the library working?
Do the teachers know how to use it. Do the students. Is there someone who helps them. Can everyone reach it, or only some. Is it actually making the teaching better.
Can you carry what having one implies?
Who is the librarian. Who decides what goes on the shelves. What happens when a book goes missing, or a child reaches something they should not have. Who answers the parent when something goes wrong at eleven o'clock on a Tuesday.
Both are measured in every package that measures, and they are never sold
apart. The finding is not either score — it is the quadrant they make together, and
a quadrant needs both axes. A school measured on readiness alone cannot be told whether it is
ready but ungoverned or running it properly, and those need opposite responses.
A school can be strong on one and weak on the other, and the two failures look nothing alike. Ready but ungoverned — capable staff using AI well, no
policy, no incident route, no named owner; the first serious incident is institutional, not
pedagogical. Governed but not ready — a signed policy, a decision matrix, a
register, and staff who cannot use the tools and students who were never taught; the paperwork is
defensible and nothing has improved. Neither is a partial version of the other, and fixing one
does not move the other.
| Quadrant | Readiness | Governance | Meaning |
|---|---|---|---|
| Early on both | below 3.25 | below 3.25 | Neither governed nor deliberately used for learning |
| Ready but ungoverned | 3.25 or more | below 3.25 | People use it well; nobody owns the risk |
| Governed but not ready | below 3.25 | 3.25 or more | Rules exist; people cannot yet use the tools well |
| Running it properly | 3.25 or more | 3.25 or more | Capability, with mechanisms behind it |
An agent runs on your own server, next to your governance site. It prepares every step; the person who owns your AI policy approves it, and nothing goes live without that approval. Twenty-four school weeks, about two hours a week of the owner’s time. No training sessions, no board meetings, no visits: the agent prepares, the school decides and presents.
| Phase | The agent prepares | Your policy owner |
|---|---|---|
| 0 · Preparationweeks 1–2 | The calendar of school weeks, the privacy notice, the evidence request | Names the contacts, books the legal review window, approves the notice |
| 1 · Baselineweeks 3–8 | Survey invitations, the ten-day window, the Exposure Scan instructions for IT, both scores and the quadrant, red flags for confirmation, the baseline report and registers | Sends the surveys, supplies evidence; leadership confirms every red flag before anything is written about it |
| 2 · Policyweeks 9–14 | The Living AI Policy, 14 sections, drafted from what the baseline found; decision rights; approval and DPIA checklists | Fills the names; sends the draft to your legal counsel; leadership approves v1.0 |
| 3 · Role materialsweeks 15–17 | A kit for each role, each line citing the policy clause it rests on; student kits by band; a family kit | Approves them and shares them through the school’s own channels |
| 4 · The policy meets realityweeks 18–22 | The approval workflow on two tools you already use; the incident register; an equity check on what students touch | The panel decides. At least one tool must be refused or modified, or the workflow was not tested |
| 5 · Steady stateweeks 23–24 | The year-one calendar, a succession note, a board briefing for the school to present | Names a deputy; the head or the owner presents the briefing |