When AI stops being an experiment: the choices facing schools in 2026–27

September always brings a sense of renewal. New pupils arrive, staff return with fresh energy and schools revisit their priorities for the year ahead. In 2026, however, artificial intelligence will return to school with us in a rather different position.

AI is no longer confined to a handful of enthusiasts testing new tools in their spare time. It is being used both formally and informally by teachers, support staff, school leaders, pupils and parents. It is increasingly embedded within the software schools already use, often without being especially visible. The question is no longer whether AI will affect education, but whether schools will shape its use deliberately or allow practice to develop by accident.

The period of experimentation has been valuable and necessary. It has allowed educators to explore what AI can do, discover where it is genuinely useful and identify some of its significant limitations. Many teachers have used it to generate initial ideas, adapt resources, draft communications or reduce routine administrative work. Schools have begun developing policies, delivering introductory training and discussing the implications for homework and assessment. However, experimentation cannot remain our long-term operating model.

If individual members of staff are using different tools, applying different standards and making their own judgements about data, accuracy and appropriate pupil use, schools accumulate invisible risks. They may also duplicate effort, create inconsistent expectations and unintentionally widen the gap between those pupils and staff who understand AI well and those who do not.

At AiEd, we believe the 2026–27 academic year should mark a shift from scattered experimentation to thoughtful, institution-wide practice. That does not mean adopting every new product or attempting to automate education. It means becoming clearer and more disciplined about why, where and how AI should be used.

There are five questions every school should consider this year.

1. What educational problem are we trying to solve?

The starting point should never be the technology itself. Schools do not need an AI strategy that sits apart from their educational priorities. They need to consider whether AI can help them address the challenges they have already identified: improving teaching, supporting pupils with additional needs, reducing unnecessary workload, strengthening communication or providing more responsive support.

AI use should therefore begin with a clearly defined problem. What are we trying to improve? Why might AI help? What would success look like? Is AI genuinely the best solution?

This matters because visible activity can easily be mistaken for progress. The number of staff using an AI tool tells us very little about whether teaching has improved, workload has reduced or pupils are learning more. Sometimes the most intelligent decision will be to use AI. At other times, the right decision will be not to use it at all.

2. Do we have clear and shared expectations?

Most schools now recognise the need for an AI policy. However, a policy stored on the staff intranet is not the same as a shared understanding of responsible practice. Staff need practical clarity about which tools they may use, what information must never be entered into them and who remains accountable for the finished work. They need to understand the risks associated with inaccurate information, bias, intellectual property, data protection and safeguarding.

Pupils also need consistent messages. They should not encounter entirely different expectations from one teacher, subject or year group to another. Schools need to be clear about when AI may support learning, when its use must be acknowledged and when it would undermine the purpose of the task or constitute malpractice.

This is not principally about catching pupils out. It is about establishing a new culture of honesty, responsibility and intellectual integrity. Young people need to understand that submitting an AI-generated answer is not the same as learning and that delegating thinking to a machine ultimately deprives them of the knowledge and capabilities education is intended to develop.

3. Are we investing sufficiently in people?

No AI policy can compensate for a lack of staff confidence and understanding. Training must now move beyond demonstrations of impressive tools or lists of suggested prompts. Staff need opportunities to consider how AI relates to their own roles, subjects and professional responsibilities. They need to know how to interrogate outputs, recognise errors, protect sensitive information and decide whether using AI adds genuine value.

Different groups will require different forms of support. Senior leaders and governors need to understand institutional risk and strategic opportunity. Teachers need practical and pedagogical guidance. Administrative and support staff need training relevant to the processes for which they are responsible. Pupils need a structured programme of AI literacy rather than occasional warnings about cheating.

Above all, everyone must understand that responsibility cannot be transferred to the technology. AI may produce the first draft, identify patterns or suggest possible approaches, but professional accountability remains human. The person using an AI-generated output must be able to understand it, challenge it and stand behind it.

4. Will AI narrow inequalities or deepen them?

AI has considerable potential to improve access to support. It may help teachers adapt explanations, improve the accessibility of materials, support language development or provide additional opportunities for practice and feedback. Emerging AI tutoring systems may eventually make more personalised support available to pupils who could never access private tuition. None of this is automatic.

Pupils have unequal access to devices, connectivity, paid tools and informed adult support. They will also differ significantly in their ability to judge whether an AI-generated response is accurate or appropriate. More confident pupils may use AI to extend their thinking, while others use it simply to replace their thinking.

Schools must therefore treat AI literacy as an issue of inclusion and entitlement. Every pupil, not only those with access and encouragement at home, should learn what AI is, how it influences their lives and how to use it critically, safely and ethically. The same principle applies to pupils with SEND. AI may create valuable opportunities, but tools must be evaluated for accessibility, bias and unintended consequences. Inclusion should be part of the design and evaluation of AI use, not an addition made afterwards.

5. How will we know whether it is making a difference?

This year, schools should begin moving from anecdotes about AI to meaningful evaluation.

That does not require a complicated research programme. Schools might select a small number of carefully chosen uses, establish what they are trying to achieve and collect proportionate evidence. Has a process taken less time? Has the quality of a resource improved? Are pupils receiving more useful feedback? Has accessibility increased? What have staff and pupils experienced? Have any new risks or unintended consequences emerged?

Schools must also be prepared to stop using a tool or approach that does not deliver sufficient benefit. AI adoption should not become a badge of modernity or an end in itself. Mature organisations are not those that use the most technology; they are those that make the best-informed decisions about it.

Keeping education human

There is a wider question running through all of this: what should remain distinctly human? Education depends upon relationships, expert teaching, encouragement, trust and an understanding of individual children that cannot be reduced to data. Teachers do far more than communicate information. They notice when a pupil is struggling, create a sense of belonging, inspire curiosity and help young people develop character, judgement and confidence.

AI should be used to strengthen this human work, not diminish it. If it releases teachers from low-value administrative tasks and gives them more time to plan, teach and support pupils, that is meaningful progress. If it generates more content, more checking and more distance between adults and children, it is not.

The year AI stops being an experiment should not be the year schools rush to automate more. It should be the year they become more purposeful, more consistent and more discerning.