A school proposes using AI to give pupils feedback on oral presentations. The demonstration is impressive: a recording becomes a transcript, then a set of comments. A parent asks a quieter question: what happens to the recording after the feedback arrives?
The school needs an answer before asking pupils to participate.
Responsible use of student data and AI starts with understanding a specific activity: what information enters the system, what new information it produces, who can use it and what happens afterwards. Trust depends on being able to explain the decision honestly, including its limits. A reassuring product description is a starting point for that work.
This article focuses on school leadership, drawing on current guidance for England and findings from the UK regulator. Schools elsewhere need their own applicable requirements and specialist advice. The discussion below is a planning approach, not a legal assessment of a product.
Begin with the pupil’s experience
From the pupil’s perspective, an assignment may arrive as a simple instruction: record your presentation and submit it through the school’s chosen service.
Adults see several decisions behind that instruction. Someone selected the activity, approved the service, arranged access and decided how the feedback would be used. The pupil may have little influence over any of them.
That makes the school’s explanation especially important. It should connect the educational purpose with the actual process, rather than expect families to infer everything from a vendor’s privacy page.
The useful question is not whether a school can promise that nothing will ever go wrong. It is whether it can explain what it has decided, why, what remains uncertain and where a pupil or family can seek an answer.
What current guidance and regulatory findings show
The Department for Education’s current guidance on generative AI and data protection asks schools to involve their data protection officer or IT lead, check approved uses and understand how personal data is handled. It also stresses transparency and notes that services may collect information beyond the content a user enters.
Its separate EdTech procurement chapter, added in July 2026, places data protection considerations early in the process. It addresses necessary data, processing purposes, storage, retention and changes to a tool. These are matters for a school’s responsible decision-makers, not questions to leave to an individual teacher at the point of upload.
The Information Commissioner’s Office supplies a reason to examine the details. In its 24 June 2026 statement on Edtech examined, the ICO described consensual audits of 28 providers during 2024–2025. It reported positive security practices alongside gaps involving roles, data-flow mapping, minimisation and privacy information.
Those findings concern the audited providers and a range of educational technologies. They are not a representative estimate of all AI products or proof that a particular service is unsuitable.
The practical implication is modest: broad reassurance should lead to a specific inquiry, rather than end it.
Follow one proposed recording
Consider a fictional English department exploring AI feedback on short oral presentations. The learning objective is to help pupils organise an explanation clearly and connect their points.
A teacher can observe that structure directly. A recording might offer another opportunity to review it. An AI service might produce feedback more quickly. These are possibilities to examine, rather than benefits established by the demonstration.
Before choosing the route, the team describes the proposed sequence in ordinary language.
A pupil records a presentation. The file is transferred to a service. A transcript is produced. Comments are generated. The teacher examines them. Some information may then appear in the school’s learning platform.
Each stage can raise a different question. A person might hear another pupil’s voice in the background. A transcript might misrepresent a sentence. A generated comment might imply a weakness that the performance does not demonstrate.
This is an illustrative planning scenario, not a report about a named school, product or Learning Coach engagement.
Separate the learning need from the information available
The department wants to examine the organisation of an explanation. That does not automatically justify every kind of information a service could collect or generate.
Ask colleagues to complete two sentences: “We need to understand whether the pupil can…” and “To examine that, we propose using…”
If the first answer concerns structure but the second includes a wide range of personal information, the difference deserves attention. The team may have included information because it was convenient, rather than because it served the objective.
It can also compare other approaches. A teacher observation, a pupil’s own outline or a rehearsal using an adult-created fictional recording might answer part of the question. None is automatically the best option for every learner.
The purpose of this comparison is to preserve a real choice. The school should be able to explain why the proposed route is appropriate, including why a less extensive approach would or would not meet the need.
Remember that AI creates records as well as receiving them
The recording is only the beginning of the fictional process.
A transcript and an AI-generated comment can become part of the account adults use to understand a pupil’s work. Their appearance may give them an authority they have not earned.
Suppose the transcript misses a linking phrase and the generated feedback says the presentation lacks a clear connection between ideas. A teacher who examines the original performance may reach a different conclusion.
The school therefore needs to decide how the generated material will be interpreted. Is it a draft for a teacher to examine? Will pupils see it directly? Could it be copied into another record? How will an error be questioned?
These are educational decisions as well as information-handling questions. The learning objective should guide which observations matter. A polished comment should not quietly become a general judgment about a child.
Make the review concrete
A productive meeting needs more than the question “Is this tool safe?”
Bring the proposed activity, a plain description of the information involved and the specific decisions that remain open. Let the relevant school owners identify what evidence and specialist review are required.
In the fictional department, the curriculum lead can explain the presentation objective. The teacher can demonstrate the intended sequence using wholly fictional material. Relevant data, technical and safeguarding colleagues can examine the proposal within their remits.
That division helps avoid a familiar problem: a teacher is asked to answer technical questions they cannot verify, while a technical colleague is expected to judge the educational need.
Unanswered questions should remain visible. A well-organised meeting does not make an incomplete answer sufficient.
Our guide to turning AI policy into classroom practice explores the wider arrangements that connect these responsibilities.
Compare the explanations different people receive
Imagine asking the department leader, the teacher and a pupil to describe the proposed activity.
The leader says AI will provide draft feedback for staff. The teacher expects pupils to receive comments immediately. The pupil believes the tool will give a final grade.
The inconsistency is useful evidence. Even if the formal review is thorough, the activity is not yet understood in the same way.
Resolve the difference before presenting a confident message to families. Explain the actual role of the recording and feedback, what the teacher decides and which questions remain open.
An explanation should be understandable without becoming misleadingly simple. “The teacher stays in control” is incomplete if nobody can say when the teacher sees the output or what they can change.
Likewise, “We remove names” cannot, by itself, explain the treatment of a voice recording or a detailed personal account.
Treat questions as part of the decision
A family may ask why recording is needed, whether the feedback is reliable or what happens if a child finds the activity difficult.
These are not necessarily objections to technology. They may reveal a gap between the school’s purpose and the experience it is proposing.
Give the question to someone who can answer it. A teacher can discuss the learning objective; they should not have to improvise an answer about an external service’s arrangements. Where the school does not yet know, say what must be established before a decision can be made.
Avoid treating a completed communication as proof that trust has been achieved. People can receive information without understanding it, or understand the proposal and still question it.
The aim is an explainable decision and a usable route for challenge.
Preserve a worthwhile learning experience
The school should also examine how the activity works for pupils with different communication, language or accessibility needs.
In the fictional presentation task, one pupil may communicate effectively through an agreed supported format. Another may need a different way to demonstrate the target capability. The team should avoid turning conformity to one recording method into an unintended learning requirement.
Appropriate arrangements depend on the pupil, the task and the school’s responsibilities. A general article cannot decide which alternative is required in an individual case.
The educational question remains consistent: what opportunity does the pupil have to develop and demonstrate the intended skill?
A change to the technology should not leave that question unanswered.
Rehearse before the pressure of a real lesson
A staff rehearsal can reveal confusion without using real pupil information.
Walk through the proposed process with fictional adult-created material. Ask where a teacher would stop if the output were wrong, a setting appeared different or a pupil raised an unexpected concern.
The rehearsal may reveal that a help route exists on paper but is difficult to find. It may expose a mismatch between the agreed activity and the instructions staff received.
That is a reason to improve the arrangements. It is not a reason to collect more pupil information to make the rehearsal realistic.
A successful rehearsal also has limits. It cannot establish the full reliability, suitability or legal status of the service. It prepares staff for an activity that still needs the appropriate institutional decision.
Keep the explanation current
The school’s account should describe the activity people will actually encounter.
If the intended use changes, the earlier explanation may no longer fit. A department that began with teacher-reviewed draft comments might later propose direct pupil access or a different assessment purpose.
Those changes deserve attention because the decision has changed, even if the product name has not.
Ask who will notice a difference and bring it back to the responsible people. Staff should know where to raise a concern without needing to solve it alone. Families should have a clear contact point rather than be passed between the school and supplier.
The purpose is continuity of understanding: the explanation, the approved activity and everyday practice should remain connected.
Make privacy a leadership conversation
For a school-network event, use a fictional activity like the presentation example.
Ask participants to explain it from three positions: the person proposing it, the teacher carrying it out and the pupil expected to participate. Identify where those accounts differ and what decision would resolve the difference.
A keynote can establish why trust requires a clear account of the work. A facilitated session can examine a concrete proposal. Neither replaces specialist data protection, safeguarding or technical review.
Our guide to choosing an AI in education keynote speaker helps organisers give that discussion a practical purpose.
The research for this article is selective: two current DfE guidance chapters and an ICO statement summarising audits. The full audit report was not reviewed. These sources do not establish that one communication method improves trust or that the illustrative activity improves learning.
Learning Coach’s position is that a school should be able to connect its educational purpose with an honest explanation of the proposed experience. Before deploying AI with pupil data, the question is concrete: can the people responsible describe the same activity, answer the important questions and recognise what remains unresolved?
Planning a school leadership event about AI, pupil data and responsible educational decisions? Explore Learning Coach’s international keynotes and request a tailored proposal.
References
Department for Education, Generative artificial intelligence and data protection in schools. Current guidance chapter; the parent manual shows an update on 14 October 2026.
Department for Education, Procuring educational technology. Chapter added 9 July 2026, according to the manual’s update history.
Information Commissioner’s Office, Statement on Edtech examined. 24 June 2026; regulator’s summary of consensual provider audits. Sources consulted 29 October 2026.
