AI is Killing Entry-Level Jobs
The 13% Drop Nobody is Talking About
Yesterday, Bank of Canada governor Tiff Mackem warned that early evidence shows AI is reducing the number of entry-level jobs available in some occupations.
Are we heading toward a future of mass unemployment, or is AI just the latest “calculator” to change how we work? Cara Stern and Mike Moffatt dive deep into the data behind AI’s impact on demographics and the workforce.
While Mike leans into historical optimism, Cara brings the receipts: a recent Stanford study showing a 13% drop in employment for young workers in AI-exposed fields since the release of ChatGPT. We explore which jobs are “AI-proof,” why Gen X seems to be winning (again), and what policy changes could help young people get a foot in the door.
If you enjoy the show and would like to support our work, please consider subscribing to our YouTube channel. The pod is also available on various audio-only platforms, including:
Below is an AI-generated transcript of the Missing Middle podcast, which has been lightly edited.
Cara Stern: So, Mike, are you worried AI is going to take your job?
Mike Moffatt: Not at all. Technology has killed about 99% of the jobs that existed even just a couple hundred years ago. Yet somehow you and I and everybody else, we’re not all unemployed farmers. One of the examples I’d use for this is: the terms “calculator” and “computer” used to refer to people. These were job titles that people had, and more often than not, women who undertook long and tedious calculations on behalf of organizations.
There were hundreds of thousands of them with these jobs all over the place, but nowadays, a computer is a small object that you can fit in your pocket, and a calculator is an app on that object. And the unemployment rate of women is actually much higher than it was 100 years ago, despite the fact that I don’t know a single woman who has the job title of computer or calculator.
So technology does destroy jobs, but it also creates them.
Cara Stern: You’re so much more optimistic than me. I look at how AI is progressing and it seems like can come for any job, and it seems like it’s so hard to predict. It makes me really nervous to think about. Yes, we aren’t all unemployed farmers, but at the same time, how do you know if the job you went into is going to be one that’s going to be needed in the future?
So it’s kind of scary, and I just keep thinking about people who are just about to start their careers. It’s got to be scary because, like, I have to look ahead until, you know, I’m 65ish, so another 30 years or so, but looking ahead, another 40 years is even harder than that. I don’t know how you know which jobs AI is going to take and which ones it’s not going to take.
Mike Moffatt: And I think that’s absolutely a fair comment that it’s not always clear who the winners and losers are going to be, what jobs are going to be lost, which ones are going to be created. So there’s a lot of uncertainty, and uncertainty is scary, but what we can do is put some evidence on this discussion. We could shed some light here.
In our group chat, you were telling me about a study by three Stanford economists on how AI has affected the U.S. job market over the last four years. What did that study find?
Cara Stern: Yeah, it was nice to be able to find some data to be like, “See my fear is not completely unfounded!”
So they analyzed tasks that language learning models like ChatGPT could do easily, and they tracked jobs that do those tasks. They looked at them over the course of four years from 2021 to 2025, and they measured the number of jobs over time that do those tasks.
They did find a drop from when ChatGPT was released to the public in late 2022. They found that workers aged 22 to 25 in the most AI exposed occupations – so those are things like software developers, customer service representatives – have seen a large number of employment declines since late 2022, when ChatGPT came out.
It’s around 13% when comparing with workers in less exposed industries, so that’s actually quite huge, and it’s nearly 20% lower compared to their late 2022 peak in some occupations, right before these AI became pretty easy to access by the public.
Mike Moffatt: I have to admit this surprised me. This was a larger impact than I would have guessed, given that these technologies are relatively new. So what kind of occupations were most exposed, according to this paper?
Cara Stern: It had a chart of ones that were most exposed vs. least exposed. Some of the most exposed ones were accountants – which surprised me a little bit – auditors, programmers, administrative assistants, customer service reps, operations managers. And then some of the least exposed were a lot of trades. There were housekeepers, there were health and psychiatric aides.
Older workers in very exposed occupations and workers in less exposed occupations across all ages, such as health aides, have seen stable or growing employment despite AI. That’s something in the range of 6 or 9%.
But overall, it makes it clear that the US economy, while overall employment is on the rise, employment growth for young workers is stagnant, and it’s likely that AI is playing a role in that.
Mike Moffatt: So it sounds like a lot of the discussions we’ve had here at the Missing Middle, where, you know, lately we’ve had some kind of doom for young workers where Boomers and Gen Xers like me are doing okay. Is that the theme of this paper?
Cara Stern: That is kind of right, but not entirely. It is the story in the industries that are most exposed to AI. AI is really good at general knowledge rather than the kind of knowledge you gain from working at a specific company or in an industry for a long period of time. Also, as you get older, relationships matter more than general knowledge.
The authors describe it as, AI replaces codified knowledge – the book learning that forms the core of a formal education. AI is less capable of replacing tacit knowledge – the idiosyncratic tips and tricks that accumulate with experience.
So what that led to is older workers in the same occupations where entry-level jobs are shrinking – they’re still seeing their employment growing or staying stable. They’re seeing between 6 and 9% growth for older workers in those AI-exposed industries, even where entry level jobs are disappearing.
Mike Moffatt: So even in those AI industries, Gen X wins again.
Cara Stern: You guys just can’t stop winning, right?
The key part, though, is that they distinguish between jobs where AI is used to automate and jobs where AI was used to augment the tasks. So if you’re using AI to organize data, maybe writing and checking technical code, that’s replacing jobs for young workers. But if you’re using it to collaborate better , get basic knowledge and check your logic, that’s augmenting.
So in areas where it’s augmenting tasks, there’s no effect on entry-level jobs, and in some areas, there are actually more jobs being created, which I was very pleased to see.
Mike Moffatt: So if I’m understanding this correctly, AI is having an impact on the number of jobs, particularly for entry-level workers. Is it having any impact on wages?
Cara Stern: Surprisingly, the study found that while employment dropped for young people, it doesn’t seem to be having much of an effect on wages yet. I say “yet” because, again, I’m not optimistic about the future with AI. But positions or demographics that are more likely to be impacted by AI have wage growth trends that are exactly the same as those that are less likely to be impacted by AI at this point.
Of course, that could change in the future – who knows? We’ll see. But that makes me wonder, how could someone or their kids avoid being affected by this barrier to entry that’s being created, as AI closes the door to young people getting into these companies through these entry-level jobs?
Mike Moffatt: Well, interestingly enough, I actually think this study provides that kind of blueprint because it makes it very clear, at least today, what types of occupations are being more affected and less affected. So you can kind of point to those areas and say, there are still lots of healthcare jobs, and so on.
My general philosophy, which really hasn’t changed with the advent of AI, is that you’re you’re best to develop a combination of full skills that go together. Most of us are never going to be superstars, the kind of 0.001%. But if you can combine a set of skills that has a rare combination – one I think is most valuable is having good verbal communication skills and also having some technical ability in some area. You can find people who can do one or the other, but it’s very rare to find someone who can do both. That combination is particularly valuable.
And finally, the people who can use these technologies are going to have a leg up on ones that refuse to learn them. So I don’t think we should, you know, go back and go, “Oh, I wish we had the same kind of labour market we had in 2004, 1987, or whatever.”
We have to understand these technologies aren’t going away. And if I was a young person, I would be really interested in learning how to use them.
Cara Stern: Looking at the study, trades seems like a good place to be because that seems hard to replace with AI, although it is very hard on your body, and jobs like health care, child care, education. Those are ones that I don’t see us replacing with AI – despite what Bill Gates said about how by 2035, we won’t need teachers any more, that we can just use AI. I don’t think our society’s going to be happy to send our kids into classrooms with robots, and have someone maybe supervising on the side.
But when I think about things like health care and child care, those are often quite low paid jobs. I’m thinking of those personal support workers that are going to be in demand and child care workers. Those are two jobs that are very much needed, very hard to replace with AI, and low pay. So it’s hard to know what you should actually go into if you’re starting a career now.
Mike Moffatt: Yeah, and I’m going to give you even more fear, unfortunately, because there often is conventional wisdom about all of this stuff that ends up being false. So I remember being a young academic hearing that I was never going to be able to maintain a teaching job because all of us were going to get replaced by these massive online courses where you could just listen to some superstar professor.
So Scott Galloway would end up teaching my class and every other class at business schools, and guys like me would be unemployed. I love Scott, but that hasn’t happened. Guys like me are still 20 years into this, despite that technology. We are still in classrooms because people value that experience rather than just learning on zoom.
On your point around the trades, I go the other way. I’ll give an example here about AI augmentation.
I recently fixed this condenser motor fan on a refrigerator. I used some online tools to figure out how to replace it – not only just replace it but to diagnose what was wrong with my fridge.
So in the past, maybe 25 years ago, I probably would have called someone up or bought a new fridge. But now, with these tools, I’m able to do these things myself. So you may see that in the future that small electrical jobs or what have you, you might not be calling that tradesperson.
Instead, you’re like, “Well, gee, I can do that myself. Or I can get ChatGPT to give me this kind of online course on how to do these things.” So it’s not even necessarily clear that there are these AI-proof industries, because even when you have AI technologies that are augmenting, they’re also augmenting for people who are not experts, like me, to be able to do these things so I don’t have to call the expert. This is where it gets really confusing.
To the last point on wages, that’s still important, right? I wouldn’t advise a young person, “Okay, well, this job is not going to get replaced by AI, so you should jump in with with two feet even if it’s low pay and it’s not something that you want to do.” It’s not really that much of a blessing that, okay, I’m not going to get replaced by Perplexity AI, but you still have to do something that you’re passionate about. And in a capitalist society, you probably want to do something that gives you a reasonable standard of living.
Cara Stern: Flexibility then is key, because if you don’t know which jobs are going to be taken over by AI, I guess you just need to be able to pivot as much as possible, which is easier said than done, of course, for a lot of people, because it’s hard to build a career when you don’t know where you’re going.
It’s much easier when there seems to be a path. But like you said, you got advice 20 years ago that didn’t actually pan out, so maybe there is too much fear.
Mike Moffatt: I think a lot of these predictions don’t come to pass, and I think you’re better off focusing on developing your skills and taking a skills-based approach rather than over-focusing on occupation, because a lot of occupation forecasts on the number of jobs that are going to exist 20 or 30 years from now are often based on things we can’t possibly know today.
Cara Stern: It also helps to look at the demographics of the country, because it can predict based on, “Are we going to have a whole bunch of seniors who need access to health care?” So maybe you go into different part of health care that is augmented by AI, but isn’t fully replaced. It’s a good way to figure out where you should be going, by looking at what the country’s going to need just based on the number of people who are going to have to access different services.
Mike Moffatt: And the number of people who are going to be retiring. So there is both a supply and demand element to this.
Cara Stern: That brings me to the next area that we have to cover, which is, given what we’re learning, that there are areas that AI is in fact replacing entry-level jobs, it makes me wonder, is there a role for policymakers?
Mike Moffatt: Yeah. So there’s this weird irony in public policymaking that, I think like 9/10 of the economic problems that Canadian governments and governments around the world try to solve are like one of two contradictory buckets.
The first one, which we hear a lot about in Canada, is that our productivity is too low. We’re using too many people to accomplish too little with too many person-hours and so on.
And then the second problem you often hear is like, automation is out of control, and we’re using too few people to accomplish so much more because technology is killing jobs, right? So it feels like we’re constantly looking for this Goldilocks point of productivity and technology.
In my view, we’ve had abysmal productivity growth over the last 40 years. So I think we should really be focusing on the first type of problem of using more technology, and getting people better skilled at that, getting companies to actually invest in that kind of thing.
My fear here with some of the AI concerns – and again, there are valid concerns – is that policymakers are likely going to do more harm than good, and are going to try and put the genie back in the bottle and harm one of the few areas where we’re actually seeing productivity growth in Canada.
Cara Stern: I was looking for some solutions for this, and I came across an article by Ilona Dougherty, who created the Youth and Innovation Project at the University of Waterloo. She’s a co-creator there. She made some suggestions that are worth considering, and so I want to run them by you. She suggests that we need more real-world experience in education.
And yes, there are programs already. We have a co-op student here at the Missing Middle – Hi Dylan!
But those programs need to be updated to make sure that those students are experiencing tasks that aren’t easy to automate. That’s what she suggests. What do you think?
Mike Moffatt: Well, surprise, surprise, a University of Waterloo person is suggesting co-op models. That’s just a joke. I’m a fan of co-op programs, so no argument here. And as a graduate of Ivy, I’m a fan of the case method, so, yeah, no arguments here, that practical education, real-world examples, real-world experience. That’s always quite valuable.
Cara Stern: I guess it makes sense because then you can start your career with that bit of experience that they’d be looking for. So you can gain those skills that you would get at an entry-level.
Her second suggestion is that companies continue to hire young workers and maybe trust them with tasks that make them a little bit uncomfortable. Things with consequences.
She argues that employers have to do this, otherwise they’re just not going to have people with experience coming up as they’re going to need to replace those senior employees. They need people who can replace them, people who have experience. If you don’t grow those employees and give them opportunities, then you’re not going to have them.
She talks about how a lot of young people will come into a workplace with very different ideas from what the status quo is there, and so they can be trusted with some problem-solving tasks. We should actually give them opportunities that they’re not getting right now with these entry-level very automatable tasks.
Mike Moffatt: Yeah, absolutely, and when I was in the private sector, I usually made a point to hire people right out of school. Part of that was, you know, I believe that they had fewer bad habits. Sometimes, working at companies, you get used to how things were done there.
But unfortunately, that’s not the mindset of a lot of Canadian businesses. Unfortunately, I can really only see most Canadian employers ever doing that kind of thing if economic circumstances required it. They’re just like, “I can’t find any more experienced workers, so I’m going to have to give this young graduate a chance.” So the only time they do that is in, overheated job market, and we have not had one of those in a very long time.
So I love the idea. I wish companies would act this way, but I’m a little bit skeptical that we’ll see this happen.
Cara Stern: The third thing she points to is that governments should play a role in funding these co-ops and work experience positions for high schoolers and for post-secondary students. So she talks about how Canada already has some of the most highly educated young people. We don’t need people getting more money to be able to get more credentials. That money should be spent on getting them more actual experience working.
Mike Moffatt: Yeah, and I think there’s a role for that as well. The federal government and some provincial governments have summer work programs and that kind of thing. A lot of times those can be kind of McJobs for lack of a better term. Trying to make those be ones that are more experiential and career-oriented would be valuable.
Absolutely, this is the kind of thing that governments have done in the past, and I’ll expect they will continue to do in the future.
Cara Stern: I saw another interesting one from Ryan Corona, who is contributing to the McDonald Laurie Institute, and here’s what he wrote:
“Canada should expand the Scientific Research and Experimental Development tax credit to include time spent integrating AI with human workflows and retraining employees for enhanced roles. When businesses can write off human-AI collaboration as readily as capital investments, they’ll choose augmentation over replacement.”
What do you think about that?
Mike Moffatt: Yeah, so this is interesting. And this is the kind of novel policy idea that I think we need to see more of. I don’t want to be too critical of folks, but a lot of the things I see coming out from policy shops is basically, “Do more good things and less bad things.” These kinds of vague things.
So I’m always a fan when policy shops get super specific like this one. I understand the logic of it and I think the logic makes sense, but I worry that that sounds like a paperwork nightmare. I can only imagine calling up the CRA and trying to figure out what is and isn’t an eligible expense for this. We’ve had past episodes on challenges at the CRA.
So I like the idea. I worry a little bit about the implementation.
Cara Stern: Thanks everyone for watching and listening. Our producer is Meredith Martin and our editor, Sean Foreman.
Mike Moffatt: And if you have any thoughts or questions about replacing a condenser fan on an LG refrigerator, please send us an email to [email protected].
Cara Stern: And we’ll see you next time.
Additional Reading/Listening that Helped Inform the Episode:
Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
Youth in Canada will need help gaining experience in the AI era
No to being young again; The struggles of Canadian youth employment - CIBC Capital Markets