Is AI a job-killer or an up-skiller?
SYLVIE DOUGLIS, BYLINE: NPR.
(SOUNDBITE OF DROP ELECTRIC SONG, "WAKING UP TO THE FIRE")
ADRIAN MA, HOST:
Lately we have been thinking a lot about the Luddites. Maybe you've heard of them? Back in the early 1800s in England, the Luddites were this group of skilled textile makers, you know, masters of their craft who spent years training and made textiles the old-school way.
WAILIN WONG, HOST:
But at the turn of the 19th century, this group of artisans had a problem. Big textile manufacturers were increasingly using machines and hiring lower-skilled workers to run them. The Luddites protested. They set fire to factory machines because they feared this trend would put them out of work. And of course, they were right.
MA: And we've been thinking about this lately because there seems to be a similar anxiety swirling around artificial intelligence. Now, leaving aside predictions about how AI may one day, you know, take over and kill off the entire human race, AI today is already showing its potential to kill off human jobs. AI is writing news articles and computer code. It's generating artwork and medical diagnoses. And yet, for all the worry about how AI will reshape the work landscape, there's new research to suggest that there might also be a positive story to tell here. This is THE INDICATOR FROM PLANET MONEY. I'm Adrian Ma.
WONG: And I'm Wailin Wong. Today on the show, we talk with researchers behind one of the first empirical studies to look at AI in the workplace. What they found might just give us one reason to be hopeful.
(SOUNDBITE OF MUSIC)
MA: When it comes to AI's potential to reshape the workplace, we're talking specifically about something called generative AI, basically a type of machine-learning system where you feed it a bunch of inputs - you know, they could be articles or movies or entire encyclopedias, basically things created by humans - and then based on that, the AI will generate something close, maybe even indistinguishable from what an actual human could make.
WONG: Now, there's no shortage of hot takes on how generative AI could affect the economy, but MIT economist Danielle Li is not really a hot-takes person.
DANIELLE LI: I get stressed out when there's sort of, like, too much hype around tech. And so usually what I do in those situations, which maybe is not the best thing to do, is I'm kind of more of an ostrich. I like to work.
MA: Where an ostrich buries their head in the sand, you want to bury your head in the data?
LI: Yeah. It's just like, that's, like, a thing I can do.
WONG: Does that make Danielle, like, a cold-takes person?
MA: (Laughter) I mean, that's not a bad thing. Like, cold pizza is just as good as hot pizza.
WONG: Yeah, I agree. I agree. I had to think about that for a second, but I decided I agree (laughter). In this case, Danielle wanted to study the effects of generative AI in the context of someone's actual job. So she and her colleagues, Lindsey Raymond and Erik Brynjolfsson, decided to look at customer contact centers - you know, the places you call or message when you can't log into your account or you keep getting an error message.
MA: Or you're wondering like, where is that thing that I ordered, like, three months ago? Basically...
MA: ...The places that you turn to in times of desperation.
WONG: Yes, when you have grievances with the company, right? So Danielle and her colleagues found a company that had begun experimenting with using AI in its customer contact centers.
MA: And here's how they did it. The company created this virtual AI assistant, which they trained up using thousands of examples of successful customer interactions. And then the customer support employees, when they were chatting with customers, they had this AI assistant on a screen in front of them, and it would instantaneously serve up suggested responses or solutions to the customer's problems.
WONG: Danielle and her fellow researchers got six months of data from this company, thousands of customer interactions, and what they learned is, well, first, that working in customer support is not easy.
LI: When we were looking through some of these conversations, the way that people talk to customer support agents - it's a very anonymous interaction in which people sort of don't think that, like, you know, God is watching. And so if you look, you see, you know, like, lots of yelling and people writing in all caps, people using frowny face emojis, people being quite - you know, saying things that they would never say to someone's face.
WONG: I would not last a day in a job where someone was typing at me in all caps all day long.
MA: (Laughter) That is hostile.
WONG: I would simply not survive. I would not survive. You know, despite the unpleasant customers, it turns out with an AI assistant, it's a little bit better. One of their main findings was that workers were more productive. They measured this by looking at the number of problems workers were able to resolve in an hour and found that productivity increased 14%.
MA: And not just that - they also found that customer satisfaction increased and employee turnover decreased, which suggests that the workers were more satisfied with their jobs. So in this case, Danielle says that the AI seemed to make a positive difference, not just for the customer or the company's bottom line but also for the employee.
LI: And I don't want to be necessarily overoptimistic about this 'cause you can imagine also a world in which just AI forces you to be more and more productive. It's tiring to be more productive. But the other view on it is that it actually sucks to be bad at your job. A lot of what customer service is, is about managing people's feelings 'cause people come, they're frustrated, and they're tired or whatever. And so in some sense there's kind of this sort of human soft skills component that these technologies are able to capture in a way that prior technologies couldn't.
WONG: Yeah, like an Excel spreadsheet may make it easier to automatically add up a bunch of numbers, but it's not going to tell you the best way to respond to an irate customer. So Danielle says in a sense, AI helps automate social skills.
MA: The researchers did expect generative AI to have some benefits for the employees, but what was really interesting to them was who benefited the most. So it turns out, the improvements in productivity and customer satisfaction were mostly coming from relatively inexperienced workers - so people who had been in the job for less than a couple of months. And meanwhile, the most experienced workers, they saw almost no improvement from using this AI. Danielle's co-author, Lindsey Raymond, says this makes sense when you think about it.
LINDSEY RAYMOND: Because the AI is learning from these human-generated training examples about, like, what makes a good interaction, the really experienced and really good workers are already doing that. So, like, access to those recommendations is not really going to have an effect, while the newest workers have the most to gain from access to those AI recommendations.
WONG: And both Lindsey and Danielle say this is pretty intriguing because of what it suggests about how AI could reshape our economy. For decades, advances in computer technology have mostly benefited highly paid, college-educated professionals. So think lawyers, software engineers, economists - technology made them better at their jobs, increased their productivity and wages.
MA: And while that was happening, that same technology made certain middle-skill jobs like clerks and secretaries and receptionists kind of obsolete.
LI: So what's happened is that a lot of the other IT has kind of led to this hollowing out of middle-class jobs. And then you have kind of high-skilled jobs benefiting from this. And a lot of the story around technology is that it increases inequality because that's empirically what we've seen.
MA: And while it's still early days and we don't have empirical research yet on how AI is affecting people's wages or job numbers, Danielle and Lindsey say just maybe what we're seeing here is the potential for generative AI to shift the old narrative - decades where technological advances have mainly benefited the most elite workers. And with these new AI tools, maybe workers who didn't have a ton of experience or a lead education can get a leg up this time.
(SOUNDBITE OF MUSIC)
WONG: Can't stop thinking about AI in the workforce? On the latest episode of our sister show, Planet Money, a renowned labor economist explains how AI could help rebuild the middle class if we play our cards right.
MA: It's really good. You should check it out. Also, this part of the show is normally where podcasts asks you to rate and review, right? But apparently that does not actually help people discover our show. So if you really love THE INDICATOR, a great way to show it is to spread the word. Tell your friends.
WONG: Tell your enemies, too. Tell a perfect stranger.
MA: (Laughter) Hey, the more listeners, the better, right?
(SOUNDBITE OF MUSIC)
MA: This episode of THE INDICATOR was produced by Corey Bridges and engineered by James Willetts. It was fact-checked by Dylan Sloan. Viet Le is our senior producer. Kate Concannon edits the show. And THE INDICATOR is a production of NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.