Schools Don't Need Every Teacher to Become an AI Expert

A school needs more AI capability than any individual teacher should be expected to hold.

Every teacher needs enough AI literacy to judge educational fit, check outputs, protect information and know when to stop or seek specialist advice. That does not make every teacher responsible for policy, procurement, security, privacy or technical assurance. Schools need broad teacher literacy, specialist depth, leadership accountability and cross-functional governance, while preserving professional judgement.

The wrong question is whether every teacher should become an AI expert

Schools are right to invest in teachers’ understanding of AI. The mistake is to assume that every problem created by AI can be solved through more teacher training.

A teacher may need to decide whether an AI-generated explanation is accurate, appropriate for the age group and useful for the learning objective. That is professional judgement.

The same teacher should not also be expected to assess supplier security, interpret data-protection law, review contractual terms, define whole-school policy and approve an unfamiliar pupil-facing use. Those are institutional responsibilities.

The distinction matters because schools are introducing AI into systems that are already under pressure. Across participating education systems, the OECD’s TALIS 2024 findings show that administrative workload is a significant source of teacher stress and that some teachers are being asked to implement change without the resources they need.

These cross-system, self-reported findings do not show that a particular AI operating model will reduce workload. They do show why leaders should be cautious about placing another undefined set of responsibilities on individual teachers.

AI capability cannot be treated only as a teacher-training problem. More training will not resolve unclear ownership of approved tools, safeguards, procurement, escalation or institutional decisions.

The better leadership question is not, “How do we make every teacher an AI expert?”

It is, “What does every teacher need to judge well, and what must the school organise around them?”

Every teacher still needs a professional AI baseline

Distributing responsibility is not an argument for reducing teacher capability. Every teacher needs a professional baseline because teachers remain responsible for the educational decisions they make.

The baseline is judgement, not technical mastery

The UNESCO AI Competency Framework for Teachers describes teacher capability across human-centred mindset, ethics of AI, AI foundations and applications, AI pedagogy and AI for professional learning.

The framework recognises progression rather than presenting every educator as an identical technical expert. While it does not prescribe a staffing model for individual schools, it provides a useful global reference for the breadth of teacher competency.

A useful baseline should help teachers understand:

• What the school permits or restricts, and which concerns teachers must escalate
• What information should not be entered into a tool
• How AI outputs can be inaccurate, biased or inappropriate
• Whether a use supports the curriculum and the intended learning objective
• When AI assistance may interfere with assessment expectations
• When to reject an output or choose not to use AI
• Where to ask for specialist advice

This article does not replace a fuller account of AI literacy for teachers. Its concern is what the institution must provide around that literacy.

The Department for Education’s guidance for England keeps responsibility for the quality and appropriateness of educational use with professionals and education organisations.

Schools elsewhere need to check their own legal, regulatory and policy requirements. The underlying educational principle still travels well: a tool cannot take responsibility for whether its use is appropriate for a particular learner, subject or purpose.

Teachers therefore need enough understanding to question an output, adapt it, reject it and explain the decision they have made.

They do not need to know everything about the system behind it.

Teacher agency is part of the safeguard

A professional baseline should expand teacher judgement, not narrow it into compliance with a new tool.

Teachers should be able to decide that AI is not appropriate for a task, that an output does not meet the required standard or that a proposed use creates more educational risk than value. They should also know when and how to escalate a concern that requires specialist advice.

School boundaries still apply, but those boundaries should not turn teachers into passive operators.

The OECD’s TALIS 2024 analysis of teacher leadership and autonomy reports associations between greater instructional autonomy and outcomes including adaptability, job satisfaction and lower stress across many participating systems.

The evidence is correlational and does not establish that autonomy causes those outcomes. It nevertheless supports treating teacher agency as something to preserve, not an inconvenience to remove.

Distribute specialist depth, but keep accountability visible

The following four-part capability model is a TopSchool synthesis and practical starting point. It is not an externally validated universal framework, and no single authority endorses this exact arrangement. Each school needs to adapt it to its size, structure, jurisdiction and local legal and policy requirements.

• Broad teacher literacy: A common professional baseline for making educational judgements about AI.
• Specialist depth: Additional expertise for questions that extend beyond classroom practice.
• Leadership accountability: Visible ownership of institutional direction, resources and boundaries.
• Cross-functional governance: A connection between educational judgement and the school’s wider responsibilities.

Leadership owns the institutional conditions

Leadership must decide what the school is trying to achieve, what it will permit and who is responsible when a question crosses classroom boundaries.

That includes setting institutional direction, providing resources and support, establishing clear expectations and ensuring that staff can reach the appropriate expertise. It also includes deciding when a proposed use should pause or stop.

This is different from asking leaders to check every AI output or personally deliver every support function. Accountability means ensuring that the work has an owner, not performing all of it alone.

Where schools need to formalise who approves particular uses, that is a separate governance question. Our guide to AI governance roles and decision rights addresses that process in more detail.

Academic and professional-development leads connect capability to practice

Academic and professional-development leads connect AI capability to curriculum, assessment, teaching quality and staff learning. They can help define what role-appropriate capability means in practice and ensure that professional learning relates to the decisions staff actually need to make.

EEF’s guidance supports professional development that is well designed, carefully selected and implemented with attention to teachers’ time and competing demands.

This is general education evidence rather than proof for a particular AI model. Its relevance is that staff learning should reflect what people are actually responsible for, rather than giving everyone the same generic session.

Specialist review should follow the risk, not the job title

IT, privacy, security, safeguarding and procurement roles may need to assess access, data flows, supplier claims, contracts, age appropriateness, integrations and incident handling.

The specialist involvement required should follow the intended use, the data involved, the users affected, the educational stakes and the supplier conditions. It should not depend on every teacher developing specialist depth or on one fixed job title being expected to answer every institutional question.

For example, the Department for Education’s data-protection guidance describes defined organisational responsibilities and specialist involvement within the English context.

It is not universal legal advice, but it illustrates why questions about personal data should not be left to individual teachers to resolve alone.

Cross-functional governance connects educational and institutional judgement

AI decisions rarely belong to one function.

A pupil-facing use may involve pedagogy, assessment, privacy, security and safeguarding at the same time. A staff workflow may be technically permitted but educationally weak. A promising classroom idea may create data or contractual questions that the teacher is not equipped or authorised to answer.

Educational fit cannot be decided by technical roles alone. Institutional risk cannot be left to individual teachers. Schools need a way for teacher and academic judgement to connect with leadership and specialist review.

Ofsted’s qualitative study of 21 early-adopter settings in England found examples of strategic leadership, champions and collaboration across functions, including data, IT and curriculum roles.

The study involved a selected group of early adopters, so it is not representative proof that one structure works everywhere. It does show that emerging school practice is already broader than teacher training alone.

The Education Endowment Foundation’s implementation guidance similarly says change should be contextual and supported by a structured process.

This is not AI-specific evidence, but it supports the principle that a new practice cannot be embedded through information transfer alone.

Cross-functional governance does not require every decision to pass through a large committee. It requires the relevant perspectives to be available and the escalation route to be visible.

Champions can help, but they cannot carry the institution

An AI or digital champion can be valuable because colleagues often need help translating a general rule into an everyday teaching situation.

A champion might:

• Demonstrate an approved workflow
• Connect a use to local curriculum practice
• Help colleagues recognise common output problems
• Gather questions and identify recurring friction
• Convene peer learning
• Direct an issue to the appropriate specialist

The important boundary is that a champion supports practice. They should not quietly become the school’s policy owner, procurement lead, data-protection specialist, safeguarding authority and technical support service.

Enthusiasm is not the same as accountability. Informal influence is not a substitute for authority, time or specialist expertise.

Schools should also avoid creating dependence on one capable individual. If the rules, approved uses and escalation routes exist only in one person’s knowledge, the capability is fragile and the responsibility is not genuinely institutional.

A champion should make the support system easier to use. They should not be the support system.

The strongest objection is that distributed responsibility can become blurred responsibility

The strongest objection is not simply that distributed responsibility may be difficult to organise. It is that distribution can fragment accountability.

Teachers may not know where to escalate. Specialists may become detached from classroom reality. Leaders may use distribution as a reason not to develop their own understanding. Smaller schools may lack every named specialist role.

These are substantive risks. If everyone is involved but no one is clearly accountable, the school has diffused responsibility rather than distributed it.

Distribution is not diffusion. A workable model still needs:

• A common professional baseline for teachers
• Specialist depth that is visible and accessible
• Explicit leadership accountability
• Connected educational and specialist judgement
• Flexibility for one person to hold more than one responsibility
• Teacher participation in decisions that affect classroom practice

These conditions do not prescribe one staffing model. They protect the distinction between sharing responsibility and allowing responsibility to become unclear.

Clarity should support teachers, not deskill them

The purpose of distributing responsibility is not to tell teachers that AI is someone else’s concern.

Teachers still need to understand the school’s boundaries, protect learning goals, question outputs and escalate concerns. Specialists still need teacher and academic input because a technically safe use may be educationally inappropriate.

The aim is clarity about where one role’s judgement ends and another role’s responsibility begins.

That clarity should make specialist support easier to reach without taking educational judgement away from teachers. Teachers remain active participants in deciding whether and how AI belongs in their practice.

The model should flex with school size, not disappear

A large school network and a standalone school will not distribute capability in the same way.

Distribution does not require a separate person for every responsibility. One person may hold several responsibilities, but each responsibility, decision and escalation route should remain explicit.

The following are illustrations, not universal recommendations:

• A standalone school may combine responsibilities across existing leaders or use external expertise where internal depth is unavailable
• A school network may centralise supplier, data or security review while preserving local academic judgement

Both arrangements can keep responsibility visible while reflecting the institution’s actual scale and structure.

Singapore’s Ministry of Education describes MOE-provided educational AI tools with built-in safety guardrails, used under teacher supervision.

This Singapore-specific example cannot be transferred unchanged to every international or private school. It illustrates the distinction between teacher judgement in practice and safeguards supplied at system level. It is not evidence for combining roles in smaller schools.

Leaders should ask what teachers are being expected to carry

The practical question for leaders is what teachers are currently expected to carry, and which responsibilities should sit elsewhere.

A leadership review should ask:

1. What does every teacher genuinely need to understand and judge?
2. Which specialist responsibilities are teachers currently carrying by default?
3. Where does deeper capability already exist?
4. Do teachers know how to reach it?
5. Does leadership accountability remain visible?
6. Can teacher judgement still influence institutional decisions?

For readers moving from capability design to implementation, a bounded AI pilot is a distinct next decision. It has separate questions about cohort, support and measures and does not replace the need to clarify institutional responsibility.

Schools do not need every teacher to become an AI expert. They do need every teacher to have a professional baseline, retain educational judgement and know where to turn when a question moves beyond that baseline.

They also need specialist depth somewhere in the institution, visible leadership accountability and cross-functional input for decisions that cannot be resolved safely inside one role.

This is not a case for a new AI department in every school. It is a case for making capability and responsibility deliberate.