The debate over artificial intelligence and jobs has become oddly binary: either machines will wipe out work or leave employment largely untouched. The evidence so far points somewhere in between. Employment has held up in aggregate in recent years, with no clear, broad-based decline attributable to AI. Yet, beneath that stability, entry-level roles are evolving faster than the post-secondary institutions that prepare graduates to fill them.

The routine, low-discretion tasks that used to be the training ground for early career workers are the easiest to automate. While this might eliminate some roles, others will be created – but these will demand something post-secondary institutions have long assumed they already teach: human judgement.

This is not a soft skill. It is the operational layer of decision-making: the ability to frame a problem, weigh competing constraints, assess risk and take responsibility for outcomes when the answer is not obvious. In an AI-enabled workplace, this is the “last mile”: algorithms generate options but people decide what those options mean and what to do next.

Job postings show how quickly this shift is taking hold. The share of all mid- and entry-level job ads asking for human judgement has risen sharply over the past two years, with nearly half asking for skills in decision-making, problem-framing and accountability.

ߣߣ��Ƶ

This demand cuts across sectors but is most visible in sectors where day-to-day decisions carry financial, safety or regulatory consequences, such as banking, energy, utilities and healthcare. The trend also extends to more hands-on roles. Construction sites, supply chains and clinical settings increasingly rely on workers who can interpret signals, manage trade-offs and respond in real time. Automation does not remove human involvement in these contexts; it raises the cost of poor judgement.

The market rewards this accordingly. Jobs requiring human judgement skills consistently offer higher wages, with the premium in roles tied to risk and accountability. Notably, the advantage is clearest for diploma and bachelor’s graduates – the bulk of the entry-level workforce pipeline.

ߣߣ��Ƶ

This creates an uncomfortable question for post-secondary education. If human judgement is now a core labour market skill, why is it too often treated as an incidental outcome of higher education, rather than a deliberate one?

Most programmes remain organised around knowledge transmission and technical proficiency. Assessment rewards accuracy, not reasoning. Students learn to produce answers, often with increasingly powerful AI tools, but are rarely required to defend decisions or reflect on consequences. The result is graduates who are technically capable but judgement-light.

The adjustment required is structural, not cosmetic.

First, human judgement needs to be embedded explicitly into curricula as a defined, assessable outcome – not left as a by-product of disciplinary learning. That means designing coursework where students must operate under ambiguity: interpreting incomplete information, weighing competing constraints and making defensible decisions when there is no single correct answer. The assessment of reasoning quality is more important than technical accuracy.

ߣߣ��Ƶ

Second, AI literacy must be paired with oversight. Teaching students how to use AI tools is insufficient. They need to understand when to question outputs, how to identify bias or error, and where accountability ultimately sits. In a world of probabilistic systems, knowing when not to trust the machine is as important as knowing how to use it.

Finally, experiential learning needs teeth. Work-integrated learning often places students in observational or execution roles, shielded from decision-making. This is a missed opportunity. Placements should include scoped responsibility: assessing options, identifying risks, making recommendations and reflecting on outcomes. Judgement develops through exposure to consequences, not just tasks.

None of this is entirely new. Universities have long claimed to teach critical thinking, and colleges have long emphasised applied learning. What is new is the clarity of the demand for it.

The ladder is still there for junior employees but the first rung is higher. Graduates are expected to arrive not just with knowledge but with the capacity to make decisions. In that sense, the future of work is not less human but the bar is raised on what only humans can do. Post-secondary institutions will have to work ever harder to keep up.

ߣߣ��Ƶ

Boxi Yang is senior research associate, education and skills, at Signal49 Research, Canada.