Show Your Working: Why Students Can Tell Who Has Done the Work

You can tell within a minute whether the person at the front of the room has built the thing they teach. Not from the slides. From how they answer the awkward question.

I sit on both sides of this line. I am a computer science student, and I run a small AI engineering practice in Plymouth. In the same week I take a lecture and ship a system for a client. I have built things, and I have sat in rooms where the person teaching them had not.

The question I keep asking is not whether a course is well run, but whether it is honest. Education is a claim about capability, and such a claim is worth only what can be evidenced.

Education is not a transaction

Students do not enrol to buy information. They enrol to be changed. They want the knowledge, yes, but also the judgement, the habits and the nerve to build something new. A course that serves slides and collects a fee is a transaction. A course that reshapes how someone thinks is a transformation.

Full is not the same as good. The question is whether the programme is equipping people, or standardising them.

The test universities avoid

This matters most in the newest fields. Universities are launching AI degrees at speed, teaching students who intend to build real systems. So a fair question follows the instructor into the room. Have you built one? Not advised on one, not read about one: built one, with users, constraints and failures that taught you something the reading list could not.

The old test was the publication list. The new test is the work.

A claim is easy to print. The evidence behind it is not. In a practical discipline, it is the second that convinces.

What the numbers say

The demand is not in doubt. Nor is the gap between what is promised and what is taught.

15%rise in UK AI degree applications, 2025BCS, from UCAS data

695UK 18-year-olds placed on AI degrees, 2025/26up 39% on the year

35%of firms with AI roles they could not fillmainly a lack of work experience

Why the answer shows

Students are perceptive assessors. They watch how a question is handled, whether the failures described were survived or imagined, and they notice when enthusiasm meets something the slide deck never anticipated. Word travels, and employers learn, intake by intake, which graduates can build.

The honest counterpoint

I am not arguing that only practitioners should teach, or that a working engineer is automatically a good teacher. Teaching is its own craft, and some of the finest lecturers spent whole careers in research. Theory is the foundation students build on.

The claim is narrower. Where a course promises industry competence, the people teaching it should show the work. Theory from experience and theory from a textbook are not the same in the room.

The rule I would put in every prospectus

Before a course promises industry-readiness, name the person who will teach it, and show what they have built. Ask four questions and do not accept certificates as answers: what have you made, who used it, what broke, and what did that teach you?

Do that and you do more than protect students from bad teaching. You build what every institution claims to want: a reputation that is earned, not asserted. The students can tell. Give them something true to tell it about.

Readout: what is confirmed, what is argued, and what is not

• This is an argument from a practitioner's vantage, not a survey. It is labelled as opinion where it makes a judgement.
• My dual role, a computer science student and a working AI engineer, is factual and is the basis of the first-hand element. The specific lecture and delivery examples are mine.
• The three figures are drawn from the sources below: BCS on UCAS data (April and December 2025) and the DSIT AI Labour Market Survey 2025. Each number is confirmed.
• No institution or individual is named or accused. The piece argues a standard for evidence, not a case against anyone.
• What would change my mind: an AI course with no teaching staff who have shipped anything, whose graduates nonetheless build well and are hired on the work they can show. I have not seen one, and I would be glad to.

Sources & references

1. BCS, AI degree applications rise 15%, 29 April 2025, from UCAS data.
2. BCS, Demand for UK computing degrees in 2025, 23 December 2025.
3. Department for Science, Innovation and Technology, AI Labour Market Survey 2025.
4. Times Higher Education, Funding challenges mean UK risks falling behind on AI education, 10 June 2025.

If you run a course, a department, or an AI programme and want an honest industry view of who should be in front of the room, email brandon@kreostudio.co.uk.