AI for nurseries: a plain-English beginner's guide (England)

This guide is for nursery owners, managers and practitioners in England who keep hearing about AI and want a straight answer on what it can do in a setting, and what it must never do. It answers the questions early years staff actually ask when they start. Two rules run through every answer: no child or parent personal data goes into an AI tool, and every decision about a child stays with the adults who know that child.

Outside England, early years settings are regulated differently. In Scotland, see the Care Inspectorate; in Wales, Care Inspectorate Wales; in Northern Ireland, the Health and Social Care Trust Early Years Teams. This guide uses England's rules and sources.

Getting started

What actually is AI in a nursery, in plain English?

The kind of AI people usually mean is an AI chat tool, such as ChatGPT, Google Gemini or Microsoft Copilot. You type a request in ordinary words and it writes a draft back: a letter, a checklist, a job advert, a tidier version of your notes. It is a writing and sorting assistant.

It does not know your children, your families or your setting. It works only from what you type, and it can get facts wrong while sounding sure of itself. So in a nursery it belongs in the office, helping with paperwork, and never in decisions about a child.

Should we let staff use ChatGPT at all?

That is your setting's decision, and it is worth making on purpose rather than finding out later that people are already using it. A sensible starting point: staff may use an AI chat tool for admin writing with no child, parent or staff personal details in it, and anything it drafts is checked by a person before it is used.

Write that down in your policy, tell staff, and make sure the manager can see how it is being used. The Department for Education's guidance on generative AI, written for schools and colleges in England, says: "Teachers, leaders and staff must use their professional judgement when using these tools." (DfE: generative AI in education).

Where can staff get training on using AI safely?

Start with the sources that set the rules rather than with a product demo. The ICO's advice for small and medium organisations covers data protection basics in plain language. Your local authority's early years team and your safeguarding partners are worth asking about any local training. A short in-house session that walks through your own policy, with one or two safe examples, is a practical place to start.

Choosing a tool

Which AI tools are suitable for a nursery or childcare setting?

For office writing, the general AI chat tools do a similar job. Pick one and get used to it. Separately, some nursery management and learning journal software now includes AI features. Before switching any on, ask the supplier in writing what data the feature uses, where it is processed, and whether you can turn it off, and check the answer against your data protection policy.

This guide does not recommend or rank products.

Is ChatGPT (or Copilot) GDPR compliant in the UK?

There is no simple yes or no, because it depends on the version you use, how its settings are configured and what you put into it. That is a question for your data protection lead to answer with the tool's own terms and privacy information in front of them. The safe working rule does not depend on the answer: keep personal data out.

The DfE guidance for schools and colleges says: "It is recommended that personal data is not used in generative AI tools." The ICO has guidance on artificial intelligence and data protection.

Is the free version of ChatGPT OK for work, or do we need a paid business version?

You can start with a free version of ChatGPT, Google Gemini or Microsoft Copilot; each tool sets its own limits, so check its own page. Before deciding, read each version's own terms on how it handles data and what settings an administrator controls, or ask your data protection lead. Whichever version you choose, the rule stays the same: no child, parent or staff personal details go in.

Children's data, parents and your policy

Can I put a child's name, initials or details into ChatGPT?

No. That includes names, initials, nicknames, dates of birth, photos, addresses, health, allergy or medical information, behaviour notes, incident details, family circumstances and anything about a safeguarding concern. A first name plus a room and an age can be enough to identify a child in a small setting, so leave those out too.

The EYFS statutory framework for group and school-based providers says (the childminders framework has the same wording): "Confidential information and records about staff and children must be held securely and only accessible and available to those who have a right or professional need to see them" (DfE: early years foundation stage statutory framework, England). The ICO's guidance on children and the UK GDPR explains the extra care children's information needs.

Work with general wording instead ("a child in the toddler room", "a parent"), and add any real detail yourself in your own system, never in the AI tool.

Do we need to tell parents, or get their consent, before using AI?

If no personal data about children or families goes into AI tools, what you are deciding is mainly about openness and trust. Parents may want to know how their nursery uses new technology, so a short, plain note is a good idea (see the worked example below).

Whether consent or a change to your privacy notice is needed depends on what data you process, which is a question for your data protection lead and the ICO's guidance, not for this guide. If a software supplier's AI feature would process children's data, settle that question before it is switched on.

Do we need an AI policy, and is there a template?

A short written policy helps, even if your answer is "admin only". It can sit inside your existing data protection or acceptable use policy. Cover: which tools staff may use, which tasks, what never goes in (children's, parents' and staff personal data, safeguarding and medical details), who checks drafts, and who to ask. Review it when tools or guidance change.

Template wording from another setting is a starting point, not a policy: it must match how your setting actually works. An AI tool can help you draft or tidy your own policy from your own bullet points, and the manager approves the final version.

What Ofsted says

What does Ofsted think about nurseries using AI?

Ofsted has published a statement on how it looks at AI during inspection and regulation (England). It says: "We do not expect or require providers to use AI in a certain way, or to use it at all." It also says: "Inspectors may ask leaders how they make sure that any use of AI supports the best interests of children and learners." And on risk: "Safeguarding: AI can pose new and unique safeguarding risks." (Ofsted: how Ofsted looks at AI during inspection and regulation). Ofsted notes the statement does not represent a final position, so check the page for updates.

In practice, be ready to explain what staff use AI for, what never goes in, and who checks the results.

Observations and learning journals

Can AI help write observations, learning journals or EYFS reports?

Only with care, and only with the practitioner in charge. The observation itself, what it shows about the child and how it links to the EYFS are professional judgements that belong to the key person. An AI tool cannot see the child and must not be asked to assess them.

Where AI can help is with wording you have already written, without any identifying details: for example, turning your own rough notes about "the child" into clearer sentences, which you then check, correct and personalise in your learning journal system. Some settings decide observations are off-limits altogether; that is a reasonable choice.

Will AI make every observation sound the same and lose our voice?

It can, if staff paste its wording straight in. AI tends towards the same polished phrases. Parents may value hearing their child described by someone who knows them. If you allow any AI help with wording, ask staff to keep their own phrasing, specific details from the moment and their own voice, and to use the tool only as an editor.

Is it fair if one practitioner uses AI and the others don't?

It becomes a fairness and consistency problem when nobody has agreed the rules. Settle it at setting level: decide what AI may and may not be used for, apply it to everyone, and make clear that no one has to use it. Judge observations on accuracy and what they show about the child, not on polish.

Can AI track a child's development across all our observations?

No. Tracking development means making judgements about an individual child over time, from their records. That is the work of the key person, the SENCo and the manager, using your own systems. Pasting a child's records into an AI tool would put personal data into it, and asking it to judge progress would hand over a professional decision.

Admin and scams

Can AI take on admin: rotas, attendance, policies, paperwork?

This is the safest place to start, as long as personal details stay out. Good starting jobs: drafting a job advert, tidying a policy from your bullet points, turning a fee change into a clear parent letter, writing waiting-list replies, or making a checklist from your own notes. Use roles or labels instead of names for rota drafts, and add the real details in your own system.

AI makes no welfare, safeguarding, allergen or medical decisions. The EYFS framework for group and school-based providers says (the childminders framework has the same rule, for the childminder): "Before a child is admitted to the setting the provider must obtain information about any special dietary requirements, preferences, food allergies and intolerances that the child has, and any special health requirements." That information is gathered from parents and kept in your records, not handled by an AI tool. A meals or allergy letter can be drafted with AI from your own policy, then checked by the manager against your records.

If you have a safeguarding concern about a child, follow your safeguarding policy and speak to your designated safeguarding lead. GOV.UK explains how to report child abuse to a local council; it notes: "This service is for England only."

How do we spot AI-made scams such as fake invoices?

AI makes scam emails, invoices and messages more convincing, because the spelling and tone are better. The DfE guidance for schools and colleges notes that generative AI "can generate believable content, including credible scam emails". Treat any unexpected request to change bank details, pay an urgent invoice or share login details with suspicion, and check it by phone using a number you already have. The National Cyber Security Centre explains how to spot and report phishing scams.

A worked example: a note to parents about how staff use AI

Example (made up, for illustration only). The setting and the wording below are invented.

The manager's own bullet points (typed into the AI tool):

• Office staff sometimes use an AI writing tool to draft letters, job adverts and checklists.
• No child's or parent's name, photo or details ever go into it.
• Observations, learning journals and anything about a child's care, health or progress are written by staff, not AI.
• Every draft is checked by a member of staff before it is used.
• Questions to the manager.

The prompt: "Turn these points into a short, warm note to parents from a day nursery in England, under 150 words, in plain British English. Do not add any promises or facts that are not in my points."

What came back (first draft, shortened): a friendly note covering each point. It also added "our AI tool is fully GDPR compliant" and "this saves our team hours every week".

What the manager changed before sending:

• Removed "fully GDPR compliant". It was not in the bullet points, and it is not something an AI tool can promise for the setting.
• Removed "saves our team hours every week", an invented claim.
• Checked the note matched the setting's written policy, and asked the deputy to read it.
• Saved the final version with the policy.

The tool did the typing. The manager supplied every fact and caught two claims that should never have gone out.

Where to go next

Pick one office job and try it for a week, with the no-personal-data rule written down first. The nurseries and pre-schools playbook has ready-made jobs for fee letters, job adverts, waiting-list replies and policy checklists, each with the exact prompt and a list of what to check. You can see three of the jobs in the package on the page. If you have never used an AI tool at all, read how to use AI at work, for beginners first.