# How Should a CPD Lead Plan a Year of AI Training for Staff?

Canonical URL: https://blog.theaieducator.io/posts/ai-training-plan-for-school-staff
Publication: Dan Fitzpatrick Insights
Author: Dan Fitzpatrick
Topic: AI Capability and Leadership Development
Published: 2026-11-07T07:15:22.000Z
Modified: 2026-11-07T16:42:10.296Z

A year of AI training works when each term has one job: know it, do it, keep it. Here is a term-by-term plan for CPD leads, built on EEF evidence and the Department for Education's own AI materials.

## In brief

Plan a year of AI training for staff by giving each term one job: autumn to know it (shared baseline and rules), spring to do it (technique practiced on real work in small groups), summer to keep it (a plan for September). Test every session with one question: will each person leave holding their own work, changed?

## Key takeaways

- A year of AI training works when each term has one job: autumn builds a shared baseline and rules, spring builds technique on real work, and summer makes the practice survive September.
- The Education Endowment Foundation's review of 104 professional development evaluations found no link between longer programs and greater impact, so a year is worth planning only for the gaps between sessions it creates.
- The EEF groups its 14 mechanisms into four jobs: build knowledge, motivate staff, develop technique and embed practice. A training plan should cover all four.
- The Own-Work Test is Dan Fitzpatrick's suggested check for any session: by the end, will every person hold a piece of their own work that they have changed, and know what they will do with it on Monday?
- The Department for Education's AI support materials, updated May 19, 2026, give England's schools a ready baseline, and its module on safe use is meant for all staff regardless of experience.
- Sort staff by where they start, not by seniority: confident users take the safety module and lead the spring groups, which saves their time and gives each group a colleague.
- In June, read samples of changed work, not attendance sheets or confidence scores.

Plan a year of AI training around what each person does with their own work, and let the calendar follow from that. A school that runs one training day and promises a follow-up has run an event. A school that gives each term one job, and tests every session against a single question, has built something staff can still feel in June.

This is a plan for whoever owns professional development: the CPD lead (continuing professional development, called professional learning in the US), the deputy who holds the training calendar, the assistant principal who has been told to "sort out AI training." It leans on England's Department for Education materials because they are the most complete public set. The sequence works anywhere that has terms, semesters or quarters.

## Where do schools actually start from?

When the Department for Education last measured it, about two in three primary schools and three in four secondary schools in England were either offering AI training or planning to, which makes the live question not whether to train but how. In its [Technology in Schools survey, published November 27, 2025](https://www.gov.uk/government/publications/technology-in-schools-survey-report-2024-to-2025) (fieldwork January to May 2025, primary and secondary schools in England, 795 senior leaders and 1,211 teachers), 21 percent of primary schools and 25 percent of secondary schools were offering generative AI training. A further 46 percent of primaries and 53 percent of secondaries planned to. In the same survey, 44 percent of teachers said they used generative AI for school activities.

Those figures are well over a year old and the position has moved, so read them as a shape, not a count. The shape is a staff room with two groups in it: people who already use the tools on their own work, and people who do not yet know what they are allowed to do. One session cannot serve both. That is the first argument for planning a year instead of a day.

## What does the research say a training plan should be built from?

The best evidence on professional development says to design around mechanisms, not hours. The Education Endowment Foundation's guidance report, [Effective Professional Development](https://educationendowmentfoundation.org.uk/education-evidence/guidance-reports/effective-professional-development), rests on a review of 104 evaluations of professional development programs from 2002 to 2020. It identifies 14 mechanisms in four groups: build knowledge, motivate staff, develop teaching techniques, and embed practice. Its second recommendation is that a program should do all four.

The report also says something that intuition resists. It cites reviews that failed to find a link between longer duration and greater impact, and concludes that "longer programmes are not necessarily more effective." What separates better programs from worse ones is what the time is used for.

So why plan a year at all? Because some of those mechanisms need gaps between sessions, not because elapsed time is good in itself. The report describes rehearsing a technique at least once outside the classroom, and notes that repeating the same action in the classroom at least twice can support embedding practice. You cannot rehearse, try, repeat and review inside one afternoon. A school year gives you the gaps. That reading is mine, not the EEF's: the report is about professional development in general and does not prescribe a three-term shape.

## What is the Own-Work Test?

The Own-Work Test is the question I suggest putting to every AI training session before it goes in the calendar. Here is the definition I use:

> The Own-Work Test is the question I ask before any AI training session goes in the calendar: by the time the session ends, will every person in the room be holding a piece of their own real work that they have changed, and know what they will do with it on Monday? A session that passes teaches a skill. A session that fails delivers a talk, and a talk does not survive the first busy week.

It is my suggested way of checking a plan, not a validated instrument. Its job is narrow. Most AI training fails for want of a person's own work in the room, not for want of content. The DfE materials are good. The trouble starts when everyone watches a presentation about prompts and nobody leaves holding anything that is theirs.

Write the test on the front of each session plan. If the honest answer is no, change the session until it is yes, even if that means covering less.

## What should each term do?

Give each term one job: autumn is for knowing, spring for doing, and summer for keeping. If you are starting mid-year, begin at the job that matches where your staff are, not at the date on the calendar.

### Autumn term: know it

The first term builds a shared baseline and shared rules, because staff who are afraid of breaking a rule will not practice. The DfE's [support materials for using AI in education settings](https://www.gov.uk/government/collections/using-ai-in-education-settings-support-materials), developed by the Chiltern Learning Trust and the Chartered College of Teaching and updated on May 19, 2026 for the 2026 to 2027 school year, are the obvious starting set. [Module 1](https://www.gov.uk/government/publications/understanding-ai-in-education-module-1) covers what AI is and how it works. The collection's own advice on module 3, on safe use, is that it "should be used regardless of your level of experience." There is also a separate page for school and college leaders. These materials apply to England; a US district should substitute its state guidance for the safety module.

The practitioner evidence agrees with the order. Alan Stupple, a director of mathematics at Bedfordshire Schools Trust, describes delivering whole-school CPD to three schools using the DfE slides. In [his account on the DfE Teaching Blog](https://teaching.blog.gov.uk/2026/03/16/how-im-using-ai-training-across-a-multi-academy-trust) (March 16, 2026), some staff did not know "what AI was and what it wasn't," and "Staff just wanted to know they weren't breaking any rules." It is one trust's experience, not an evaluation, but it shows the order the EEF's mechanisms suggest: knowledge and safety first.

Pair the modules with a one-page sheet that says what staff may do on Monday. A [short school AI policy](https://blog.theaieducator.io/posts/short-school-ai-policy) can supply it. The Own-Work Test shows up here as the smallest possible task: each person leaves with one low-stakes piece of their own work, such as a parent newsletter or a set of lesson notes, that they have rewritten with the tool and read critically.

### Spring term: do it

The second term develops technique, practiced on real work, in small groups, with someone to give feedback. This is where the DfE's modules 2 and 4 (interacting with generative AI, and use cases) earn their place, and it is where most one-day plans never arrive.

Stupple's sessions give a usable pattern. Staff "took work from their shared drives and rewrote it together in the sessions," and he advises dedicating "an hour or two to play around with AI tools." The EEF mechanisms behind it are modeling the technique, rehearsing it and getting feedback. Keep the groups small and split them by subject, phase or role: a science department, a Year 3 team, the teaching assistants (see [what to tell teaching assistants about AI](https://blog.theaieducator.io/posts/what-to-tell-teaching-assistants-about-ai)), the early-career teachers (see [early-career teachers and AI](https://blog.theaieducator.io/posts/early-career-teachers-and-ai)).

Each group's output is a before-and-after of real work plus a note of what the tool got wrong. That note matters most. It is where staff discover that use of AI stops somewhere, and [where that line sits for your staff](https://blog.theaieducator.io/posts/ai-skills-for-staff-beyond-basic-training) is the thing the next session should be about.

### Summer term: keep it

The third term makes the new practice survive September's first busy week, using the EEF's fourth group: prompts and cues, action planning, self-monitoring and repetition in context. Ask each person to write a short plan for the first half-term of next year, naming the one task on which they will keep using the tool and the one thing they will stop doing by hand. Put a reminder in the diary for week two of autumn.

This is also the term to ask what the saved time becomes. [What happens after the initial AI training](https://blog.theaieducator.io/posts/what-happens-after-ai-training) sets out the decisions only a leader can make, and [how to evaluate AI training for teachers](https://blog.theaieducator.io/posts/how-to-evaluate-ai-training-for-teachers) covers what to collect. The short version: in June, read samples of changed work, not attendance sheets.

| Term | Job | Who | Each person leaves with | The check |
|---|---|---|---|---|
| Autumn | Know it | All staff | One page of rules, one piece of own work changed | Can staff say what they may use it for? |
| Spring | Do it | Small groups by role | A before-and-after of real work, plus what the tool got wrong | Does every group's example show a mistake caught? |
| Summer | Keep it | All staff | A plan for the first half-term of next year | Is there changed work to read in June? |

## What about staff who start in different places?

Group people by where they start, not by seniority. Stupple describes "three different camps," from petrified to over-enthusiastic, and any staff room has the same three. A two-question pulse at the start of autumn is enough to sort them: what have you used AI for on your own work in the last month, and what would you not use it for? The answers show who needs the baseline, who is ready to practice, and who should be leading a spring group.

Do not make confident users sit through the full autumn program. Give them the safety module and ask them to lead. It saves their time and gives the group a colleague, not an outside voice. The [three conditions for getting staff to actually use AI](https://blog.theaieducator.io/posts/how-leaders-build-ai-confidence-across-a-workforce) are a useful companion here.

## What I See in Practice

The training plan that fails is rarely a bad plan. It is a good plan whose second session never happens. I have trained more than 150,000 educators across more than 30 countries, and the pattern I see most often is that the first session is booked for an INSET (in-service) day while everything after it is scheduled for "when there is time." There is never time. The plans that last put the second session in the diary before the first one runs, and treat the first session as the start of something, not the whole of it.

The second pattern is quieter. A plan can have the right sessions and still fail the Own-Work Test in every one of them, because it was designed around what the presenter wanted to say. Expect to trade some content for a task when you run the test on your own plans.

## Where does this plan run out?

The plan runs out in four places, and you should know them before you commit the calendar.

First, the sequence is a suggestion. The EEF supplies the mechanisms and the DfE supplies the materials; the three-term shape and the Own-Work Test are my way of putting them together. Second, the evidence about AI-specific staff training is thin. The EEF's findings come from professional development in general, the DfE modules were last updated in May 2026, and the Stupple account describes a delivery, not an outcome. Third, the survey figures date from early 2025.

Fourth, a year is sometimes too much. If your staff already use these tools daily and need only the rules, autumn alone may be the right plan. Stretching it would run against the EEF's own warning that longer is not better.

## What should you do before the next staff meeting?

Do these six things, in this order, before you book anything.

1. Put the date of every session, in all three terms, in the calendar before the first one runs.
2. Run the Own-Work Test on each session plan, and rewrite any that fail.
3. Send the two-question pulse and sort staff into three groups.
4. Write the one-page sheet of what staff may do, and check it against your policy.
5. Choose who will lead the spring groups from among the staff who already use the tools.
6. Decide now what you will read in June: samples of changed work, not attendance.

If your leadership team is working through how to sequence AI training for staff, this is the kind of work I support through [AI training for schools](https://theaieducator.io/ai-training-for-schools?utm_source=blog.theaieducator.io&utm_medium=referral&utm_campaign=ai-training-plan-for-school-staff) and advisory work. For the wider case on building capability without creating dependence, start with [is your AI training building capability, or dependence?](https://blog.theaieducator.io/posts/build-ai-capability-not-dependence)

## Sources and further reading

- Department for Education, "Technology in schools survey report: 2024 to 2025," GOV.UK, published November 27, 2025 (fieldwork by IFF Research, January to May 2025, England). [gov.uk](https://www.gov.uk/government/publications/technology-in-schools-survey-report-2024-to-2025)
- Education Endowment Foundation, "Effective Professional Development," guidance report, first edition (the EEF page carries no publication date). [educationendowmentfoundation.org.uk](https://educationendowmentfoundation.org.uk/education-evidence/guidance-reports/effective-professional-development)
- Department for Education, "Using AI in education settings: support materials," GOV.UK, first published June 10, 2025, last updated May 19, 2026. [gov.uk](https://www.gov.uk/government/collections/using-ai-in-education-settings-support-materials)
- Department for Education, "Understanding AI in education: module 1," GOV.UK, published June 10, 2025, last updated May 19, 2026. [gov.uk](https://www.gov.uk/government/publications/understanding-ai-in-education-module-1)
- Alan Stupple, "How I'm using AI training across a multi-academy trust," DfE Teaching Blog, March 16, 2026. [teaching.blog.gov.uk](https://teaching.blog.gov.uk/2026/03/16/how-im-using-ai-training-across-a-multi-academy-trust)

*Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor and the author of bestselling books on AI in education. [Read more about Dan](https://www.theaieducator.io/about?utm_source=blog.theaieducator.io&utm_medium=referral&utm_campaign=ai-training-plan-for-school-staff).*


## Frequently asked questions

### What should an AI training plan for school staff include?

It should cover four jobs: a shared baseline and rules, practice on real work, feedback on that practice, and a plan for keeping it going. One way to sequence them is one job per term: know it in autumn, do it in spring, keep it in summer.

### How long should AI training for teachers be?

Length is not the lever. The Education Endowment Foundation's review found no link between longer professional development and greater impact, and says what the time is used for matters more. Plan enough sessions to practice, repeat and review, and no more.

### Should all staff get the same AI training?

No. Staff start in different places, from anxious to enthusiastic. Run a short pulse at the start of the year, give everyone the safety basics, and let confident users skip the introduction and lead small practice groups by subject, phase or role.

### Which DfE materials can schools use for AI training?

The Department for Education's Using AI in education settings support materials include four modules for all staff and a page for leaders. Updated May 19, 2026 for 2026 to 2027, they apply to England. The DfE says module 3, on safe use, should be used regardless of experience.

### How do you know whether staff AI training worked?

Look at the work, not the room. Attendance and confidence scores look the same after poor training as after good. In June, read samples of changed work from each group and ask whether staff can show something they now do differently.

### Who should deliver AI training in a school?

An internal lead using the DfE modules can deliver the baseline, as one trust director did across three schools. Practice groups work best when led by a colleague who already uses the tools on real work, with an outside voice kept for specific gaps.

---
Source: [How Should a CPD Lead Plan a Year of AI Training for Staff?](https://blog.theaieducator.io/posts/ai-training-plan-for-school-staff)
Publisher: [The AI Educator](https://theaieducator.io)
