Last week I presided over two graduation ceremonies at the University of Wollongong and attended a third. Thousands of students. Thousands of families. The same question, unspoken, in every row.

Is this still worth something?

It is a question shaped by a specific anxiety: that generative AI is about to render graduate skills redundant, that the credential is devaluing in real time, that the years of study and the debt were a bet placed on the wrong future.

I think that anxiety is asking the wrong question.

The standard debate frames graduate employability as an automation problem. AI can now write code, summarise legal documents, draft reports, build financial models. These are tasks associated with graduate entry-level work. If AI does them, who hires the graduate?

The problem with this framing is that most jobs are not collections of tasks. They are bundles: tasks, yes, but also coordination, judgment, relationships, institutional knowledge, and accountability. When one component of that bundle changes, the bundle reshuffles. It doesn’t just disappear.

A junior analyst whose routine modelling work is automated does not simply become unnecessary. Tasks that previously belonged to a more senior analyst begin flowing to them. New tasks emerge: validating model output, managing AI-generated analysis, catching the errors a confident model makes with quiet authority. The bundle reconstitutes around a different set of demands.

This is what actually happens in organisations when a component of work becomes cheaper or faster. The junior analyst’s role changes. It doesn’t just vanish.

But more importantly, to do the new tasks well you need to know more than the old ones required, not less. AI output requires a human who can evaluate it. That means knowing enough to recognise when the model is wrong: when the summarisation has missed the critical caveat, when the financial model has made an assumption that does not hold, when the code does exactly what it was asked and not what was needed.

That knowledge has to be built somewhere. It does not come from prompting. It comes from the process of learning a discipline: from getting things wrong under supervision, from understanding the underlying logic rather than the surface output. AI does not eliminate the knowledge floor. It raises the cost of not having it.

The graduates who struggle will not be the ones who know their field and also know how to use AI tools. They will be the ones who used AI tools to avoid developing the knowledge that would let them evaluate the output.

There is a second argument for university education that rarely appears in the employability debate, because it is harder to quantify. University produces a person, not just a graduate.

The gap between a student who arrives at university and the one who leaves is not primarily a gap in what they know. It is a gap in how they operate: judgment under ambiguity, professional identity, the capacity to take responsibility for outcomes. These are formed over years, through friction. Bad drafts, difficult feedback, group work with people who are nothing like you, deadlines that matter.

Generative AI can produce a first draft. It cannot form the person who knows what to do with it.

We should also be honest about the short term.

Some of what looks like AI-driven displacement right now is not. It is economic and geopolitical instability, with AI as a convenient cover story for decisions that were going to be made anyway. Executives who have been looking for a reason to downsize have found one. That is real, and it is not fair to the graduates caught in it.

The period is genuinely volatile, and pretending otherwise is dishonest.

But volatility has always sorted for the same qualities: the ability to adapt without losing your sense of what you are for, to find the problem worth solving in a changed environment, to contribute in ways that are not scripted. Universities, at their best, build that.

The graduates I watched walk across the stage last week are going to be fine. Not because the credential protects them, but because of what forming it required.