How teachers are using AI in the classroom: 12 real examples

Talk to any group of tutors about AI and you get two reactions in the same breath: relief that a chore has got quicker, and worry about what it might be doing to standards. Both reactions are fair. The honest picture is that AI has become a genuinely useful classroom tool for FE staff, but only when it is used with a clear head about what it is good at and what it is not.

Below are twelve examples pulled from real practice across colleges and training providers, not marketing case studies.

1. Differentiating worksheets

A tutor writes one core worksheet, then asks an AI tool to produce a simplified version for lower-level readers and an extension version for advanced learners, keeping the same learning objective. This used to take an evening; it now takes twenty minutes of editing.

2. Generating starter and plenary activities

Rather than a blank page, teachers ask for five starter activity ideas for a specific topic and group size, then pick and adapt one. It replaces staring at a lesson plan template, not the planning itself.

3. Drafting feedback comments

Some assessors use AI to turn brief marking notes into fuller written feedback, then check and personalise every comment before it goes to the learner. The AI never sees the learner's actual work in these cases, only the assessor's shorthand notes, which avoids data protection issues.

4. Creating case studies and scenarios

Vocational tutors ask for realistic workplace scenarios, for example a health and safety incident in a warehouse, to use in assessment prep. This works well because it is generative and low stakes, not something a learner's grade depends on being accurate.

5. Summarising long reference documents

Awarding body handbooks and qualification specifications are long. Staff use AI to produce a first-pass summary of a specification, then verify it against the actual document because summaries can miss or misstate assessment criteria.

6. Text to speech and reading support

AI-powered read-aloud tools help learners with dyslexia or visual impairment access written materials without a separate support worker for every session.

7. Translating materials for ESOL learners

Handouts get translated into a learner's first language as a scaffold, alongside the English version, not as a replacement for it.

8. Adjusting reading age

Tutors paste in a policy document or news article and ask for it to be rewritten at a lower reading age, useful for functional skills and Entry Level learners.

9. Quiz and question bank generation

AI drafts multiple choice questions from a scheme of work, which the tutor then checks for accuracy before using them, because AI-generated questions sometimes contain factually wrong distractors.

10. Meeting and CPD note summarising

Staff record permission-based meeting notes and use AI to produce an action list, saving admin time that can go back into teaching.

11. Roleplay practice for assessors

New assessors use AI as a practice "candidate" to rehearse questioning technique before a real professional discussion, which builds confidence without using up a real assessment opportunity.

12. Drafting internal quality assurance reports

IQA staff use AI to structure observation feedback into a standard report format, again working from their own notes rather than uploading learner evidence.

What ties these together

None of these replace professional judgement. The AI drafts, the human checks, edits and takes responsibility for the final version. That single principle, human in the loop, separates safe use from risky use. It is worth building a simple structure into your own prompts too: a clear role, the context, the specific task and the format you want the output in, which we cover in detail in AI prompts for lesson planning that actually save time.

Where it goes wrong

The same staff who find AI useful are quick to list where it lets them down:

• Confidently wrong facts, especially names, dates, legislation references and statistics
• Generic tone that needs heavy editing to sound like an actual person
• Inconsistent formatting across long documents
• A tendency to invent sources or citations that do not exist
• Bias baked into training data, which can surface in scenario writing or example selection

A simple rule of thumb

If a mistake in the output would embarrass you, harm a learner, or breach a policy, check it properly before it goes anywhere. If it is a low-stakes first draft you would have written yourself anyway, a quick scan is usually enough.

Getting started without overreaching

Providers new to this do not need a big rollout plan on day one. Start with low-risk uses: starters, differentiation, admin summaries. Build staff confidence and a shared understanding of good practice before moving into anything that touches assessment decisions or learner data directly. Formal CPD helps here, and structured routes through our qualifications and CPD courses give staff a shared vocabulary and confidence rather than everyone experimenting alone with no guardrails.

Keeping data safe

Every one of the examples above avoids putting identifiable learner information into a public AI tool. That is not paranoia, it is a GDPR requirement. Names, addresses, health information, safeguarding notes and even full pieces of learner work can count as personal or sensitive data, and pasting them into a tool whose data handling you have not checked is a real risk, not a hypothetical one.

The bigger picture

AI is not going to write your scheme of work or replace the judgement calls that make good teaching good teaching. What it does well is take the first thirty percent off a task, the blank-page part, so staff spend their time on the parts that need a human: relationships, judgement and accountability.

Used this way, AI is simply another tool in the box, alongside the whiteboard, the VLE and the photocopier that never quite works. The providers getting the most value from it are the ones treating it exactly that ordinarily, with sensible rules rather than either banning it or leaving it unmanaged.