Colleges don't need more AI literacy. They need human literacy.

For most of modern history, education rested on a simple bargain. Develop knowledge, literacy, numeracy and occupational skill, and young people become economically useful, socially included and personally capable adults.

AI is quietly dismantling that bargain. When a machine can draft the essay, produce the drawing, run the calculation and summarise the evidence in seconds, “preparation for cognitive work” stops being a good enough answer to the question what is education for?

I’ve spent nineteen years teaching and leading engineering in further and higher education, and this year I set out my answer in a position paper, Human Literacy in an Age of Non-Human Intelligence. This post is the short version.

”More critical thinking” isn’t the answer

The standard policy reflex goes: AI makes critical thinking more important, so teach more of it. That’s partly right. Learners do need to question AI output and spot its mistakes. But it falls short for three reasons.

• It’s vague. In most courses, “critical thinking” means “evaluate your sources” or “write a balanced argument”. That’s much narrower than the phrase suggests.
• It depends on knowledge. You can’t critique AI-generated engineering analysis without enough engineering to know where to probe. Skills without substance collapse into vibes.
• AI is getting good at it. Systems already compare evidence and generate counterarguments. A learner whose job is to be a slower fact-checker for a faster machine hasn’t been educated. They’ve been parked.

Critical thinking asks whether something is valid. Judgement asks what matters, who’s affected, what risk is acceptable and what should actually be done. AI can assist with the analysis. People stay accountable for acting on it.

From task completion to human formation

The paper’s central idea is what I call human formation: deliberately building knowledge, skill, identity, ethics and agency together, so learners can act responsibly in real situations.

It isn’t a new label for personal development or employability skills. It changes the organising question of a course, from “Can the learner complete the task?” to “Is the learner becoming capable of responsible judgement in a world where more and more tasks can be automated?”

In engineering terms, the task at risk is CAD execution, drawing production and following a specification. What the curriculum now has to build is design intent, safety reasoning, materials judgement, manufacturability and professional responsibility, the things that sit above the task.

One point matters a great deal: this isn’t a shift from knowledge to “soft skills”. Internal fluency is the foundation for judgement, not an alternative to it. A learner who has never done the calculation themselves won’t develop the gut feel that tells them when the AI’s answer is wrong.

The risk nobody wants to name

AI won’t arrive in education evenly. For advantaged learners, it means richer tutoring and faster feedback, on top of plenty of human contact. For learners in under-funded provision, the risk is what I call automated containment: AI platforms replacing time with a skilled lecturer, presented as “personalised learning”.

That temptation is real, and it isn’t a failure of character. Colleges have absorbed years of funding pressure, and specialist engineering, construction and digital posts are the hardest to fill. A platform that promises to cover the gap looks like survival.

But for a learner with interrupted schooling or fragile confidence, the relationship with a skilled educator is often the thing that holds them in education at all. So any AI decision in a college should face one question first: does this give learners more human contact, or less? Automating registers, tracking and compliance paperwork to free up teaching time is the legitimate use. Using AI to cut contact hours is not.

Assess readiness, not completion

AI has also exposed something the sector has quietly known for years. Unit-by-unit accumulation can produce the appearance of capability. Everyone has an incentive to move learners through the gate, and now a learner can submit fully compliant work with very little engagement.

The paper proposes a different question for qualifications. Not did the learner complete the programme? but is the learner ready for what comes next? That means university, employment or an apprenticeship, assessed through a substantial learner-led project and a human-led professional discussion, much like an apprenticeship end-point assessment.

The obvious objection is assessor time. AI helps here, without replacing judgement: it can map evidence, flag gaps and prepare targeted questions, so the assessor walks in properly briefed. The assessor still has to read the work themselves. In short: AI carries the coverage. Humans carry the judgement.

Three things you can do this term

None of this needs to wait for national reform.

1. Audit one scheme of work for agency. How much of it asks learners to choose, frame a problem or justify a decision, rather than follow instructions?
2. Replace your AI ban with an AI policy. Some tasks done without AI, to build real fluency. Some done with AI and assessed on the learner’s direction and critique. Some that ask learners to find where AI gets it wrong.
3. Make your reasoning visible. Narrate how you reach a judgement, not just the conclusion. Learners who watch expert judgement develop judgement. Learners who only see expert answers develop compliance.

The point

AI doesn’t make education less important. It makes its real purpose impossible to ignore: forming people who can live and work wisely with powerful tools, not people who can merely recall what the machine already knows.

The future of education isn’t AI literacy. It’s human literacy.

The full paper covers the evidence, the policy changes I think are needed, and practical frameworks for colleges. Read it here. If you’re working through what AI means for your provision, I’d be glad to talk.