A college principal leading AI has one decision a school head does not: which standard the learner is being prepared for. The awarding body's rulebook restricts AI in assessed work, while the workplace the course leads to increasingly expects it. Policy, training and tool choices all follow from how you settle that gap.
Here is the gap in one hypothetical week. A Level 3 business administration learner (an invented example) spends Monday at her employer, where her manager asks her to draft a client letter with the company's AI assistant and check it before it goes out. On Tuesday she is back in class, finishing a unit that counts toward her qualification, and she has no idea whether the same move is a skill or an offense. Nobody lied to her. Two sets of adults wrote two sets of rules, and neither knew the other existed.
What makes AI in a college different from AI in a school?
A college serves learners a school does not, under rules a school does not face, and AI lands on all three differences at once. The same corridor holds 16-year-olds, adults retraining, and apprentices who are with an employer for much of the week. Much of the assessed work leads to vocational or technical qualifications owned by an awarding body outside your control. And the destination is a job, so the question "what will this learner be expected to do with AI on the first day of work?" has a real answer, which differs by occupation.
That last point is the one school-focused guidance rarely reaches. If you lead a school, the argument over AI is mostly about protecting learning. If you lead a college, it is also about protecting employability, and the two pull in opposite directions more often than anyone admits. The same decisions face US community colleges and technical colleges, with different rulebooks: accreditors, state boards and industry credential bodies in place of awarding bodies.
What do the assessment rules actually say?
The assessment rules are stricter and more specific than most college staff realize, and they bind the college as well as the learner. The Joint Council for Qualifications (JCQ), the body that sets shared rules for the main awarding organizations in the UK, published AI Use in Assessments: Your Role in Protecting the Integrity of Qualifications, effective April 30, 2025. It covers coursework and other non-exam assessment for both general qualifications and vocational and technical ones, with BTEC among its worked examples.
Four of its requirements matter to a principal. A college must have a malpractice policy that covers AI. Learners must name the tool and the date the content was generated, and keep their prompts and the output. Teachers must be satisfied that the work is the learner's own before they accept it, and may not use AI as the sole marker. And awarding organizations may sanction college staff who knowingly accept inauthentic work or fail to check it. That last sentence turns a learner-conduct matter into a staffing risk, and it belongs on the agenda of whoever owns quality.
One caution. JCQ speaks for its member awarding organizations, and not every qualification your college delivers comes from one of them. Check each awarding organization's own position, course by course, rather than assuming the JCQ text covers everything on your timetable.
Do students actually know what is allowed?
Many do not, and colleges are further behind universities in telling them. Jisc, the UK's digital body for tertiary education, reported in its Student perceptions of AI 2025 post, published May 21, 2025, that 86% of universities and 49% of colleges had student guidance in place, yet students still said they were unclear about what AI use was allowed. The post draws on discussion groups with 173 students and seven member surveys, with no split reported between college and university learners, so treat the figures as a signal of direction. Some students, it adds, avoid AI altogether because they fear it counts as cheating.
Read that last finding slowly. The harm of unclear rules does not fall only on those who cheat. It falls on the careful learner who stays away from a tool her future employer will hand her on day one.
The Two Rulebooks Check
The Two Rulebooks Check is three questions a college asks of every course: What does the awarding body's rulebook allow in assessed work? What does the workplace this course leads to now expect? And where, before assessment, will learners practice the difference, with whom watching? A course that cannot answer all three is preparing learners for a job it forbids them to rehearse.
I offer this as a way of thinking, not a validated instrument. Its use is that it fails visibly, course by course, and the failures are usually not the ones a policy committee expects. Here is what one completed row might look like, for the invented business administration course above:
| Question | Hypothetical answer for Level 3 business administration |
|---|---|
| What does the awarding body allow in assessed work? | AI use must be named, dated and evidenced with prompts and output. No AI for the analysis the unit exists to assess. |
| What does the workplace now expect? | Drafting and summarizing with an approved assistant, with a human checking before anything leaves the building. |
| Where do learners practice the difference? | A named, unassessed task each term where they use the tool, record the prompts, then explain what they changed and why. |
| Who owns the answer? | The curriculum lead for the course, reviewed with two local employers each year. |
Notice what the third row does. It gives the learner a legitimate place to build the judgment that employers want, without touching the integrity of the assessed work. That one move dissolves most of the tension. The learner is not asked to choose between honesty and employability. She is shown which room she is in.
What are colleges that are further ahead doing?
They are making the rules visible to learners and building the practice into the course, and they are doing it with people rather than prohibitions. Jisc's AI in FE member stories roundup, published September 14, 2026, gathers examples from about two years of community sessions. Coleg Sir Gâr and Coleg Ceredigion adapted the AI Assessment Scale so that learners can show how they used AI in each assignment. South Devon College co-designed an AI tutor with learners and staff and moved its focus from detecting AI use toward helping students use it appropriately. Kirklees College offered AI training to everyone from governors to catering staff. Milton Keynes College wrote AI literacy into schemes of work, with learners sorting AI from non-AI examples and comparing outputs.
These are self-reported practice from member colleges, not evaluated results, and the roundup itself gives no outcome data. What they share is more useful than any single tactic: the work happens inside the curriculum, and the learner is told what is expected before the assignment, not after the accusation. If you are building your own approach to suspected misuse, what a school should do when a student is accused of using AI and how a head of department should handle AI in coursework set out the sequence, and assessment in the age of AI explains why design beats detection.
What I See in Practice
The most common mistake I see in colleges is a policy written by the people nearest the technology, for a problem owned by the people nearest the learner. I was once a Director of Digital Strategy in further education, and the pattern is familiar from the inside. The policy is accurate, approved and filed. Then a tutor in one department allows AI with a disclosure, a tutor next door bans it, and a learner who moves between the two rooms concludes that the rules are weather.
Across the education leadership teams I work with, the ones that get past this have done one unglamorous thing: they put curriculum leads, quality staff and an employer voice in the same meeting, and asked each to answer the same three questions for their own course. Nobody enjoys that meeting. It is the only one that produces a rule a tutor can apply on a Wednesday.
Who should own AI in a college?
One named senior leader should own it, with curriculum, quality, IT and governors each holding a defined part. Jisc's Leading AI in colleges: a strategic framework (the page carries no publication date) organizes the work into three pillars: skills, knowledge and culture; technology; and governance, resting on data maturity. Its recommendations include clear senior accountability, starting with the problem rather than the technology, and updating existing academic, data and IT policies rather than writing a standalone AI policy.
I would add one thing from experience: the senior owner must be able to settle disputes between curriculum and quality, because that is where the Two Rulebooks Check will land. The wider question of why a single AI lead is never enough is covered in who should own AI strategy, and what your corporation board should see is in what to report to governors about AI.
What should a college leadership team do this term?
Run the Two Rulebooks Check on a small number of courses before you write another page of policy. Five steps, in this order:
- Pick three courses: one where the awarding body is strict, one that leads straight into a workplace that uses AI heavily, and one where you suspect staff disagree.
- Put the curriculum lead, a quality manager and one employer in a room for an hour per course, and fill in the four rows above.
- For every course that cannot answer question three, design one unassessed practice task and name the tutor who will run it.
- Publish one page per course telling learners what is allowed in assessed work and where they can practice the rest. Learners should hear it before the assignment, not after.
- Agree who tells the corporation board what you found, and when. A principal who reports gaps early is more credible than one who reports compliance late.
Two longer threads sit behind this. The workplace question is changing fastest at the bottom of the career ladder, which is where your learners start; what happens to entry-level work when AI does it and which skills matter when everyone has access to AI are the places to start. And the staff side, where a single training day rarely changes practice, is the subject of AI skills for staff beyond basic training.
A college that gets this right gives a learner one thing a prohibition never can: the chance to rehearse the job honestly. The ones that do not will keep producing two kinds of graduate, those who learned to hide their use and those who never learned to use it.
The next step
If your leadership team is working through these questions, this is the kind of work I support through AI strategy sessions for schools and colleges, turning a set of rulebooks into a plan each course can apply. If you are planning a leadership conference or an away day, I also speak on leading AI in education. I write about these decisions most weeks in my leadership newsletter.
Sources and further reading
- Joint Council for Qualifications, "AI Use in Assessments: Your role in protecting the integrity of qualifications", effective April 30, 2025: jcq.org.uk
- Sue Attewell, "Student perceptions of AI 2025", Jisc National Centre for AI, May 21, 2025: nationalcentreforai.jiscinvolve.org
- Helen Nicholson-Benn, "AI in FE member stories: Learning from real college experiences", Jisc National Centre for AI, September 14, 2026: nationalcentreforai.jiscinvolve.org
- Jisc, "Leading AI in colleges: a strategic framework", undated page read October 22, 2026: jisc.ac.uk
Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor and an international keynote speaker who has delivered AI training to more than 150,000 educators across more than 30 countries. More about Dan.
