What happened in Paris
UNESCO reported on 9 September 2026 that more than 25 education ministers and designated representatives had adopted a joint statement on education as a human right and a common good in the age of AI. The statement was presented in Paris on 8 September during Digital Learning Week. Its priorities include public oversight, inclusive design, teacher agency and attention to the full cost of adopting AI. UNESCO also reports a call for providers to demonstrate educational benefit and accept accountability. These are international priorities, not a new binding rule for every school.
Start with the learning problem
A polished demonstration can begin with a tool and work backwards to a use. A more useful discussion starts with the difficulty a school actually faces. Is the task preparing alternative explanations, giving timely practice feedback, adapting reading material, or reducing repetitive administration? Each problem needs a different test. “We want to use AI” is too broad to tell whether a trial succeeds.
Write one sentence naming the intended benefit and one observable sign of improvement. For example: teachers want a faster way to draft three explanations of a science concept, while retaining responsibility for checking vocabulary and accuracy. A small trial can record preparation time and the corrections required. It should not announce improved learning merely because staff produced more material.
Ask who makes the consequential decisions
Map a proposed activity from the first input to the final decision. Who selects the information? Who sees the output? Who can correct it? Who explains a disputed result? A draft worksheet and a judgement about a child’s future have very different consequences. The more consequential the decision, the more clearly responsibility needs to remain visible to the people affected.
Consider a fictional system that recommends support groups. An attractive dashboard might conceal weak assumptions about language, attendance or previous attainment. Before discussing automation, ask whether a teacher can inspect the basis of the recommendation, hear a learner’s explanation and change the decision. Convenience does not answer those questions.
| Discussion area | Ask for something concrete |
|---|---|
| Purpose | One defined task, its current difficulty and a realistic trial outcome. |
| Quality | Examples of errors, how they are caught and who checks the final material. |
| Access | A workable route for learners who cannot use the proposed device or account. |
| Responsibility | A named person who handles corrections, concerns and decisions. |
| Exit | A practical way to stop the trial and retain the work the school needs. |
Test the awkward cases
Prepare examples that resemble the school’s real work, including cases likely to expose a weakness. A question in a local language, a scanned page with imperfect text, a diagram whose labels matter, and a request for a source can reveal different limitations. Use invented or properly cleared examples in demonstrations; a vendor does not need a real child’s private record to show how a feature works.
Include teachers who will have to use the system on an ordinary busy day. Ask them to correct an answer, find an explanation, export useful work and recover from a mistake. Record the steps and time involved. A feature that works only when a presenter supplies an ideal prompt may not fit the classroom routine.
Count the work around the software
A trial has costs beyond its advertised price. Someone must set it up, explain it, check material, handle problems and decide whether to continue. Create a simple record of these tasks instead of treating staff attention as unlimited. Ask what happens when a free offer ends, a device fails, an account is lost or the school wants to move its work elsewhere.
Agree on a stopping point before enthusiasm makes every problem a reason for a larger rollout. A useful trial can end with “this helps with one task”, “this needs changes”, or “this does not fit our setting”. Stopping after a careful review is a legitimate result, not an admission that the school opposes technology.
A short agenda for the next staff discussion
- Name one learning or administrative problem before naming a product.
- Choose a limited trial and identify who will review the results.
- Prepare realistic examples, including errors and access difficulties.
- Decide what information is necessary and who can see it.
- Write down the human checking, training and support work.
- Set a review date and describe how the school can stop or change course.
The useful contribution of the international debate is a wider set of questions. A school can welcome experimentation while asking for a clear purpose, visible responsibility and a fair opportunity for everyone to participate. Those questions belong in the discussion before the demonstration becomes the decision.
Sources & further reading
Reviewed 28 Oct 2026. Sources were checked for this article. Policies and developing stories may change after this date.
