The approval chain is the hard part — not the visit
The school visit is the visible work. The difficulty lives afterwards: getting a draft report from a lead inspector, through quality assurance, back to the school for response, through senior sign-off, and out to publication — with a defensible record of who decided what, and when.
A platform earns its place here or nowhere, and the test is whether it handles the unhappy path:
- A report rejected at QA and sent back for revision, then resubmitted — without losing the original judgements or the audit trail.
- A school's formal response to findings, captured against the specific judgement it contests — not bolted on as a loose comment.
- A status that is always unambiguous: any inspector can see exactly where any report sits, and what is blocking it.
Every judgement must trace back to its evidence
A judgement that cannot be traced to the observation behind it is indefensible the moment it is challenged — by a school, a ministry, or an auditor. This is the single most important property of an inspection record, and the easiest to lose when a paper-and-spreadsheet process is digitised carelessly.
- Each judgement links to the specific evidence — lesson observations, document review, interviews — that produced it.
- The link survives editing, re-grading and the passage of years, so a report defended long after publication still holds together.
- Evidence is captured on site, at the point of observation — not reconstructed from memory days later.
AI can draft. It must never decide.
AI is genuinely useful in inspection — compiling school context, structuring collected evidence, turning an inspector's notes into well-formed draft prose. It removes administrative weight, and that weight is considerable. But its role has a hard boundary, and a credible platform enforces it in the architecture, not in a policy document:
- Every AI-generated output requires named human approval before it affects any published outcome. Nothing reaches a school automatically.
- The judgement is always the inspector's. AI structures and drafts; it does not grade.
- Every AI action is written to an append-only audit log — so the inspectorate can show, at any later date, what was generated, by what, and who signed it off.
This matters more each year. A government inspectorate must be able to answer, to a minister or an auditor: where was AI used, and who was accountable for the result? A platform that cannot answer that cleanly is a liability, however capable it appears.
Data must be sovereign — and provably segregated
Inspection data is among the most sensitive an education system holds: named pupils, identifiable staff, judgements with real consequences. Three things are non-negotiable:
- Segregation is architectural, not procedural. Where a platform serves more than one client, one jurisdiction's data must be structurally incapable of informing another's — not merely a rule an administrator is trusted to remember.
- Residency is the inspectorate's decision — where the data is hosted, and under whose jurisdiction, is the client's call, not the supplier's.
- Exit is defined before entry. Full export in open formats and a documented path out, agreed at the start. Data a supplier can hold hostage was never sovereign.
The framework is yours — the platform must bend to it
Every inspectorate is different: its evaluation areas, its grading scale, its inspection categories, its approval routes, its report structure. A platform hardcoded to one country's model forces every other client to adopt that model — which no serious inspectorate will, nor should.
The right question is not "does it match our framework today?" but "when our framework changes — and it will — is that a configuration change, or a redevelopment project?" A framework revision should not stall published output for a year while the system catches up.
What good looks like, in one line
A defensible judgement, traceable to its evidence, moved through an approval chain that survives challenge, assisted by AI that never decides, on data the inspectorate wholly controls, within a framework that is genuinely its own.
None of this is exotic. But it is the part that is hard to build and easy to get wrong — and it is worth testing for directly before committing to any system, this one included.
Written from experience, not theory
Simon Bell developed the national school inspection system for Jamaica's National Education Inspectorate, and earlier worked on Cambridge Education's Ofsted inspection systems. EDUINSPECT360 is the product of both.