Planet Money - Is AI slopifying the job market? (Two Indicators)

Episode Date: December 3, 2025

Vote for us in NPR’s People’s Choice Awards: npr.org/peopleschoice AI is already reshaping how people find work. Fewer entry-level jobs, robot recruiters, and ever-changing new skill requirements all add up to a new, daunting landscape for humans trying to find dignified work.Today on the show: two stories from the edges of a changing labor market. First we’ll assess claims that AI is causing a white collar job apocalypse. What does the data actually say? We meet an economist who has found one small but fascinating way to measure the impact of AI on workers. Then, we go face-to-face, or at least voice-to-voice, with AI. We meet a robot recruiter for a job interview and find cause to ask, ‘When might that actually be preferable to a human recruiter?’Pre-order the Planet Money book and get a free gift. /  Subscribe to Planet Money+Listen free: Apple Podcasts, Spotify, the NPR app or anywhere you get podcasts.Facebook / Instagram / TikTok / Our weekly Newsletter.The original Indicator episodes were hosted by Wailin Wong, Darian Woods, and Adrian Ma. They were produced by Cooper Katz McKim and engineered by Robert Rodriguez and Debbie Daughtry. They were fact checked by Sierra Juarez. They were edited by Paddy Hirsch and Kate Concannon. Learn more about sponsor message choices: podcastchoices.com/adchoicesNPR Privacy Policy

This is Planet Money from NPR. A couple of months ago, Jim Farley, the CEO of Ford Motor Company, took the stage at an event called the Aspen Ideas Festival. He was there to talk about the future of the economy and the importance of skilled trades. And as he sees it, something of a crisis. Hiring an entry worker at a tech company has fallen 50% since 2019. Are really, is that really where we want all of our kids to go? And then Jim Farley kind of just drops in this huge prediction. Artificial intelligence is going to replace literally half of all white-collar workers in the U.S.

Half of all white-collar workers. That is quite a pronouncement from the CEO of a major company. And Jim Farley's not the only executive talking in these dramatic terms about how AI might upend the economy as we know it. Like it could be a 21st century version of the... Industrial Revolution. Hello, and welcome to Planet Money. I'm Darren Woods.

And I'm Waylon Wong. Normally, we're co-host over at Planet Money's Daily Podcasts, The Indicator. But today, we're here to bring you two stories about how AI is changing hiring. Today on the show, we'll visit one slice of the job market where people are already feeling AI, the interview. AI is screening candidates. We'll test it out and check the data that suggests having an AI. interview might not be all bad. But first, that big pronouncement, will it get harder for

white-collar job seekers to find entry-level work? What then? Stay tuned. It's replacing jobs in fields like coding and customer service. We've also seen how AI can help people be more productive or efficient at work. We did a recent episode where we heard from listeners who are using AI for manual or tedious tasks like writing emails and analyzing data. This is a burgeoning area of research for economists like Laura Veldcamp. She's a professor at Columbia Business School. And the debate over whether AI is replacing or complementing jobs got her mental wheels. turning. I started thinking about the process of knowledge production, because AI doesn't produce

cups and plates and goods, it produces knowledge. And so in thinking about the knowledge production process and how it was changing, I was looking to history to guide my thinking about, well, when else have we seen major changes in production processes? And it seemed like the Industrial Revolution was the most natural parallel. The Industrial Revolution took more than a century to unfold and span multiple continents. It is really difficult, probably verging on irresponsible, to try to summarize such a large complex era of human history. But that's what Josh Freeman is here for. I'm a historian and I studied the history of labor and industry in particular. If you had lived during this period of history, what job do you think you would want?

Ah, well, I would like to be a wrenchier where you didn't have to have a job. You just lived off the labor of your tenant farmers or the interest off of your bonds. You know, early industrial work was mostly not pleasant. Passive income is the dream, Darian. Or at least so many emails are informing me in my spam folder. Those are sent by AI. Anyway, Josh says that when you zoom out on the Industrial Revolution, you can see how it both improved and diminished human life on practically a cosmic scale.

If you look at England, which is usually considered the first place for the Industrial Revolution, in the mid-18th century, so 1750, the average life expectancy was less than 40 years. Today, it's around 80. So that's the most basic measure, and you've doubled. On the other hand, you and I and everyone else is facing a planetary crisis from the Industrial Revolution that may seriously impact the future of our species. So, you know, in the biggest scale, you can see both the upsides and the downsides. Of course, this increase in life expectancy was not a straightforward or linear path. People's health actually got worse as they moved into crowded, polluted cities and worked dangerous factory jobs.

It took generations of organized labor activity, government regulation, and advancements in health care to change these conditions. Ultimately, Josh says the Industrial Revolution redefined people's relationship with their home life, with nature, even with their basic sense of time. And it's too early to know whether AI is upending human civilization in the same way. But economists like Laura Veldcamp are trying to study AI-related changes in smaller doses. And they're doing this by zooming in on things they can measure. Laura set out to capture essentially how bosses and workers are splitting profits. There is a technical term for this. It's called the labor share of income.

So think about the money that a business makes. Some of that money comes back to workers in the form of wages. And that percentage is the labor share of income. It both represents how important labor is and the output, what share of value it's adding. But it also represents what. what share of the income that labor should receive. Laura says the labor share of income fell during the Industrial Revolution. This is because the adoption of new machines meant less human labor went into producing goods.

Laura wanted to know if this is happening with AI and today's knowledge workers. So she and a colleague studied workers in the financial sector. They picked that industry because it's been an early adopter of AI. There is entry-level work that Laura thinks will require less human involvement over time, like maintaining databases, and then moving up the career ladder, AI is more of a complement than a substitute. For example, Laura says workers at financial firms are using AI to analyze data to make investment decisions.

And here's what Laura and our colleague found in their research. They predict that AI could lead to the labor share of income in knowledge work dropping by 5%. So if you think of profits or GDP as a pie among firms that are adopting AI, workers are receiving a smaller slice of the overall pie. Laura says this 5% decrease is similar to what happened to the labor share of income during the Industrial Revolution. So that's one potential parallel between then and what's happening with AI today. And Laura says it's not necessarily bad news for workers that they're getting a smaller slice of the pie.

And that's because the overall pie is getting bigger. We find that a worker who has AI skills in the financial sector is making about, $22,000 a year more than somebody who doesn't. So they may be getting a smaller slice of the pie at their firm, but their firm's likely to be much more profitable. And so as a result, that smaller slice is still more take-home pay. A smaller slice of a bigger pie. So this gets us into big questions about fairness.

Like how should workers and corporations split profits? Laura says the Industrial Revolution offers lessons here too. What we saw in the Industrial Revolution is that there were a few firms that adopted these new technologies and became monopolists. This was the era of robber barons, right? And the, you know, great capitalists, and they were insanely rich at a time when most people were desperately poor. And so I think there's a risk that we could follow the same path here

where there are early adopters that become monopolists and have an enormous amount of market power and squeeze us as consumers and us as workers to get most of the rents. Concerns about a small group of companies controlling those rents or, you know, the AI pie, are on Josh Freeman's mind, too. He points out that in the U.S., there used to be lots of automakers and steelmakers and airplane manufacturers. These industries got concentrated over multiple decades. Josh says so far, AI seems to be different. We're starting from extreme concentration.

You know, that's the way it's beginning. So that is a somewhat different dynamic. it's kind of monopolized from the get-go and will that shape the way this unfolds across the society and who benefits? You know, those are still very open questions. Open questions that Laura says policymakers and regulators will have to grapple with.

You know, if AI is going to make all those companies more productive and wealthy, how should those spoils be divvied up? And if the Industrial Revolution is any guide, figuring out how to share this pie could take generations of struggle between government regulators, bosses, and workers. And computers. Oh, I forgot about the computers.

Yes, they're waiting in the wings. Do not cut them out. After the break, my colleague Adrian Ma is going to speak with Anna. She's a recruiter whose job has been totally upended by AI. In fact, she doesn't know what her job was like before AI at all. Hello. Hi, Adrian. This is Anna, the AI recruiter from TP, calling to discuss the customer. PSG Global Solutions is a company that specializes in recruiting. A company you might call if you need to hire a lot of, say, call center agents or warehouse workers or nurses, but don't have a lot of time.

David Koch works for PSG, where his title, it's a very fancy one, Chief of Transformation and Innovation. Though, funnily enough, his metaphor for recruiting is a very analog piece of technology, a funnel. And it's like a leaky funnel at that. The funnel is essentially a number of candidates going in and number of candidates actually being placed. Right? And every step of the way, every step in the process, you lose people. So typically, PSG posts a job online, a bunch of people apply, and the ones that seem promising are contacted by a recruiter for an interview. David says this is where the funnel starts to leak.

You know, it's one thing attracting talent and it's one thing identifying talent. The next step is you have to get them on the phone. And the more time that passes after a candidate applies, the harder that is. David says if a recruiter calls an applicant within one minute of applying, there's an 85% chance of them actually connecting. But if just 15 minutes go by, that connect rate drops to just 35%. And that's where recruiters end up spending hours a day, just dialing and redialing the same numbers. The amount of time they're spending just trying to connect and how much they actually

talk to someone is crazy. And then those that are connecting to very often are saying, like, I'm not interested. I feel like if David were an actor in one of these made for TV infomercials, this is a part where he would look directly into the camera distraught and he would say, there's got to be a better way. Yeah. And so what did they end up doing? They brought in a robot, of course. Hi, Adrienne, this is Anna, the AI recruiter from TP, calling to discuss the customer service representative role you recently applied for. Is this a good time to talk? Yes, it is.

Awesome, just a quick heads up. So this is their AI interview bot that they developed called Anna. And to show me how it worked, they had Anna interview me for a call center job. Can you describe a situation where you had to manage multiple competing demands? man's while handling a customer call? Are you talking about like multitasking? Yes, that's exactly what I'm referring to. Nice on Adrienne, stalling for time, vamping.

Thank you. I handled it just like a real interview. Because I'm often interviewing or having conversations with people. I'm listening. Please go on. How do you handle multiple tasks or interview? Yeah, and sometimes things come up in the conversation where it's going pretty well, I think. I think so. Could you think of any advice you might give to someone who is just starting out in this role? I guess I would say try to imagine yourself in the other person's shoes and have some empathy for what a person might be going through when they call. That's a great approach.

Having empathy for the customer's situation can go a long way in providing excellent service. Yeah, so I'm not feeling great about making. into the next round, to be honest. No, I thought that was a good answer. Anna said it was a good answer. Nailed it. A-plus. So PSG Global has this new technology. And the problem was they didn't have the data to prove Anna could do the job as well as a human. So they turned to an economist at the University of Chicago named Brian Jabbarian.

Yeah. And coincidentally, Brian had been looking for a company just like this, like one that was trying to use AI at a broad scale. and, importantly, willing to give him access to detailed data on how it was going. It took me three to four years, basically, in total, to find a firm willing to partner with me. Were there any companies that just, like, laughed in your face when you approached them with this idea? Well, most of them even didn't reply. But then, then, PSG said yes, because it really wanted to know what effects Anna would have on the recruiting process. To test these questions, Brian devise an experiment. where job candidates were split into three randomly assigned groups.

The first group would go through the normal interview process with a human. The second would be assigned to an interview with Anna. And the third would be given a choice, human or AI, whatever they preferred. Now, importantly, for all the job applicants, a human recruiter would still review the transcript or the audio from the interviews and make the actual decision of whether or not to offer a job. Brian's hypothesis was that Anna would not do as well as a human. And you can imagine why, right? Like, who among us actually enjoys calling up a business just to get that automated voice that is like, why don't you read me a 16-digit number?

And I'll, you know, and then you're just like, aberrater, aberrater. And yet, Brian says after running this experiment on some 70,000 interviewees, it was quite shocking or like surprising, which is when given choice, 78% of candidate choose to be interviewed by an AI voice agent. Given the choice, 78% of people chose AI. Who would have thunk? I was also very surprised to hear this. And one explanation seems to be that people felt the AI would be, for lack of a better term, less judgy. And in fact, Brian says job applicants who interviewed with Anna were about half as likely to report feeling discriminated against

based on their gender compared to those who interviewed with a human. And interestingly, women were also more likely than men to choose the AI over the human interview. But the surprises went even further. Brian also found that people who went through the AI interview process were 12% more likely to get a job offer and about 18% more likely to start and stay in the job for at least a month. So it seems like Anna was better at interviewing than the human recruiters. Of course, the next obvious question is why. Brian says when he analyzed the interviews, he noticed some patterns that people who got interviews tended to display certain, what he calls, linguistic features.

If you display a lot of interactivity, you have a lot of back and forth, or you display a high level of vocabulary richness, you increase your chances of getting a job offer. On the flip side, if a candidate used a lot of so-called back channel cues, like, uh-huh, uh-huh, that decreased their chance of getting an option. offer. And here is the kicker. Candidates interviewed by Anna did better on all these measures compared to those interviewed by a human. Okay, so putting this all together, it's almost like Anna allowed people to be better versions of themselves. And Brian says the psychological aspect of all this is definitely worth more study. And as for the company, PSG, you could imagine that having a recruiter that can work 24 hours a day and be infinitely replicated, that is seemingly less prone to human bias and pretty good at its job to boot is a big freaking deal.

And you would be right. David Koch at PSG says they plan to roll Anna out in 80 countries and use it to recruit for a lot of different kinds of jobs. And while it will definitely mean the company hires fewer human recruiters, David says the ones who remain will get to spend more time on analytical tasks and a lot less time just dialing numbers. The role of the recruiter is changing. and I think it's a positive change.

It's going to get more difficult, but it becomes a much more meaningful job, I think. By the way, David says there are dozens of competitor companies developing technology like Anna. So maybe sooner than we think, robots will be interviewing us for jobs. You know, there's been all this talk about how AI is too sycophantic to humans. Yeah, it's always sucking up. But I feel like now we have to study up on how to suck up to machines.

Oh, no. We have maybe a few more years left where you, dear human listener, still have agency and can make your voice heard. That's right, it's the people's choice. NPR is having our listeners vote on the best podcast. So whether it's Planet Money or The Indicator, just help one of us be a through line, please. Vote for us at npr.org slash people's choice. There's a link in the show notes. Today's episode comes to you from Planet Money's daily podcast, The Indicator.

Check it out if you don't already subscribe. The original stories were produced by Cooper Katzma Kim and engineered by Robert Rodriguez and Debbie Daughtry. They're a fact-checked by Sierra Juarez. They were edited by Patty Hirsch and Kay Cuncanon. I'm Waylon Wong. This is NPR. Thanks for listening.