AI Isn't Just Taking Jobs.
It's Making It Harder for People to Prove What They Can Do.

Dr. Murgatroyd: No. Job loss matters, but it is not the whole story. The bigger shift is that AI is changing how people learn about work, build experience and show employers what they can do. For a long time, entry-level jobs helped turn learning into visible ability. People learned by doing real tasks, they made mistakes, got feedback, improved and gradually earned trust. Employers could watch that happen. Now, that pathway is getting weaker.

Junior roles used to do more than fill routine needs. They were training grounds. Assistants, analysts, trainees and junior staff learned how the work actually worked while coming to understand pace, judgment, standards and how to recover from mistakes. Employers did not need perfect proof on day one because they expected people to grow into the role.

That model was already under strain, and AI is speeding it up. In many knowledge-heavy jobs, AI is starting to do the work junior
employees once did first. Not everywhere, of course. The picture is different in skilled trades, care work and many service roles. But across administrative, analytical, technical and professional work, the first rung of the ladder is getting thinner.

So the real problem is not just that some work is disappearing. It is that one of the main ways people used to build trust and prove ability early in their careers is starting to break down. A quick fact will help: 21% of Canadian companies have already stopped hiring for entry-level jobs, and an additional 47% say they will stop hiring for entry-level work by 2027.

Dr. Murgatroyd: Yes. AI did not create this problem. It exposed it and made it worse.

For years, employers in Canada have invested too little in training and development. They spend less than any country in the G7. Entry-level pathways have narrowed. Hiring has become tougher at the bottom, with employers expecting people to arrive already productive. At the same time, more people have gone through higher education without a matching expansion in stable career-track opportunities.

The result is familiar: Youth unemployment stays high (currently at 14%), underemployment is common and many graduates struggle to get a real foothold in the labour market. The link between education and work was already under pressure. AI accelerates that pressure, making an existing weakness harder to ignore.

The critical question is not whether people are learning enough. It is whether the system still turns that learning into something employers can see, trust and reward.

Dr. Murgatroyd: It is changing the system in three clear ways. First, it is removing many of the starter tasks junior workers once learned from. Early-career employees used to draft first versions, summarize documents, do basic research, review routine material and handle repetitive coding or administrative work. Those tasks were not minor. They were where people learned the craft, picking up judgment, workflow, standards and confidence. When those tasks disappear, the learning does not just speed up. In some cases, the learning path disappears with them.

Second, employers are taking on fewer development risks. As AI raises expectations around speed and output, firms have less reason to hire people who need time, coaching and patience. They want people who can contribute right away. That helps explain why so many so-called entry-level jobs now ask for years of experience. It also helps explain why employers are relying more on automated screening and ranking tools. The burden of becoming job-ready is being pushed away from employers and onto individuals, colleges and universities.

Third, AI makes finished work harder to read as proof of individual ability. A report may be polished. A presentation may be strong. A code sample may work well. But what, exactly, does that prove? Did the person do the thinking, shape the structure, solve the problem and make the key decisions? Or did AI do much of the heavy lifting

Dr. Murgatroyd: It means more uncertainty on both sides. Graduates want their learning to lead to decent work, good wages and some stability. Employers want to hire with confidence.

AI makes both sides less certain. When employers are no longer sure what a credential, portfolio, writing sample or work portfolio really proves, they fall back on safer signals. They ask for more experience. They lean more on referrals, institutional brand, internships and tightly controlled assessments. Some firms are also building their own pipelines, boot camps, internal academies, rotational programs and employer-designed qualifications because they trust what they can observe and test for themselves.

The wider risk is that access narrows. People who already have strong networks and strong signals keep moving ahead. People at the start of their careers, or those coming through less traditional routes, face a steeper climb.

Dr. Murgatroyd: Exactly. Employers are not abandoning standards. They are changing what they trust.

For a long time, the move from learning to work relied on fairly stable signals: credentials, work-integrated learning, probation periods, supervised practice and time on the job. People entered real workplaces, performed tasks, learned by doing, made mistakes, improved and built evidence that others could observe. Employers could see ability developing in context.

AI weakens that model. It simplifies or removes some of the tasks through which novices once learned. It raises expectations for immediate contribution. And it makes outputs harder to interpret. A polished piece of work no longer tells you as much about the person behind it. As those direct signals weaken, hiring becomes more cautious and more dependent on indirect signals, brand-name institutions, prior experience, social networks, elite internships, interviews and cultural fit. Those are not neutral substitutes. They are unevenly distributed and often socially biased.

So yes, AI is changing work. But it’s also changing how people are judged, how risk is managed in hiring by employers and how opportunity gets allocated.

Dr. Murgatroyd: I mean the set of steps that turns study, practice and experience into something employers believe.

With -level work, employers could see ability taking shape in real conditions, which is what created trust.

When the proving ground shrinks, the problem is not just fewer tasks. The problem is fewer chances for people to show what they can do in a way others believe.

Dr. Murgatroyd: No. A skills gap suggests a simple answer: Teach people the missing skills and the problem goes away. That is too neat a response for what’s really happening now.

The deeper problem is that the systems that develop, test and recognize ability are weakening at the same time work is changing faster. AI is changing what counts as useful work, raising expectations for immediate contribution and undermining some of the old ways people proved themselves.

So yes, people still need skills. But they need a credible way to demonstrate those skills and have them recognized. This is a problem of capability, recognition and trust.

Dr. Murgatroyd: We already produce graduates, certificate holders and people with more training than most countries in the world. Yet, many still struggle to turn that learning into stable work. That tells us the issue is not just how much learning is happening. It’s whether employers trust the signals attached to it.

If AI makes the final output easier to polish but harder for the employer to interpret, then more training on its own does not fix the missing link.

And if all we do is add more courses to a system with weaker recognition, we may end up with more learning on paper and no more opportunities at work.

Dr. Murgatroyd: It looks like this: Someone learns useful skills through education, projects, work experience or self-directed learning. They can produce strong outputs, and often they do so with AI support. The work may look impressive. But that’s where the uncertainty starts.

Take a graduate applying for a risk role at a bank. They submit a sharp brief and a polished presentation. The employer may still wonder: Can this person actually think through a messy problem on their own or did AI do most of the heavy lifting?

Or consider junior software development. A candidate may show working apps and clean code. But the real question is whether they understand the structure, can debug unfamiliar problems and can make sound decisions without leaning on the tool every step of the way.

The issue is not that the person learned nothing. It is that the employer no longer feels sure what the evidence actually proves.

Dr. Murgatroyd: Several, and they are serious. Pathways into the labour market weaken. Early-career workers face more instability. Employers rely more on exclusionary proxies. Capable people are overlooked or underused.

Over time, the system becomes less efficient because skills are not being matched well to real opportunities. And it becomes more unequal because access depends more on existing advantages than on demonstrated ability.

There is also a wider consequence. When the link between learning and work becomes less reliable, trust in the value of education itself begins to weaken. People may still pursue credentials and training, but with less confidence that they will lead to meaningful employment. Employers may complain about labour shortages while doing less to develop talent. Institutions may produce more learning without creating more recognized opportunity.

That is not just a labour market problem. It is also a social contract problem.

Dr. Murgatroyd: We need to rebuild the pathway from learning to recognized ability. That is the central task.

Recognition can no longer be treated as a side effect of education or hiring. It has to be treated as a basic part of how people get a fair shot at work in an AI-shaped economy.

That means, first, building better evidence of what people can actually do. Not just transcripts or course completions, but clear proof of performance: what a person can do, under what conditions, to what standard and with how much independence.

Second, we need stronger ways to validate that evidence. If outputs are increasingly shaped with AI, trust cannot rest on polished artifacts alone. We need assessments, demonstrations, observation and verification that can distinguish familiarity from real competence.

Third, we need portable records of competence. People should be able to carry forward trusted evidence of what they have demonstrated across institutions, sectors and stages of life. That record needs to be more specific than a degree and more credible than a self-assembled portfolio.

We also need to expand and redesign work-integrated learning. If AI is shrinking the tasks through which early ability was once developed, we cannot simply assume workplaces will keep providing those opportunities on their own. We need deliberate structures in which people can still learn, be coached, be observed and build trust: co-ops, apprenticeships, embedded practice, strong simulations tied to performance and better employer-educator partnerships.

In other words, we need a better bridge between learning and opportunity.

Dr. Murgatroyd: My main point is this: AI is not only changing jobs. It is weakening the bridge between learning and work. If we do not rebuild that bridge, training will matter less, hiring will become more exclusive and more people will struggle to get a fair start.

The real question is not whether Canada can produce educated people. We can and we do. It is whether we can make their abilities visible, credible and usable in an AI-shaped labour market.