You’d think one line of work that would be safe from the robot takeover is human resources. But some companies are working on it.
A recent New York Times investigation laid out how Amazon’s automated systems for managing warehouse workers led to unreasonable scheduling, delayed benefits and inadvertent firings.
I spoke with Peter Cappelli, a professor of management and director of the Center for Human Resources at the University of Pennsylvania’s Wharton School. I asked him what a traditional HR department used to do that’s now done by software. The following is an edited transcript of our conversation.
Peter Cappelli: You know, I think there’s a lot of paper-pushing things that have been automated and work pretty well. It’s just putting stuff into software. So, filling out forms, for example, we probably all do those a lot. Also, chat bots, which are pretty common. The same sort of thing that you’d get if you call customer service, although, we hope it works better than that. It answers your questions and moves you along the trails. The things which are kind of on the horizon are things that involve machine learning, and building algorithms that will actually help you predict things like, predict who will be a good employee or possibly even predict who will be a bad employee. Who among your current people might steal from you or something like that, right? And those are the more controversial ones.
Amy Scott: That seems pretty scary, knowing what we know about mistakes that algorithms make.
Cappelli: One of the big differences that algorithms make in terms of mistakes is they make mistakes at scale. For example, if you’re thinking about hiring, generally speaking, we do the hiring really poorly. We push the problem often on the line managers because it’s cheap to do, and each manager does it differently. So, say they actually discriminate against somebody, well, there’s one manager and one problem that maybe you could identify. Once you have an algorithm that makes hiring decisions, and it makes a mistake, it’s a class action mistake, because it’s making the same mistake for everybody. And you’re now in big trouble as an employer.
Scott: And the legal risks are, I assume, discrimination lawsuits, wrongful termination, that kind of thing?
Cappelli: All those things, and generally, you have to show that the practices you’re using actually work. If you didn’t build the tool, you have to still be able to demonstrate that it works on your employees.
Scott: Are you seeing any innovations in this area that you think are positive?
Cappelli: Sure. It’s used to do things which we’re just not doing now. For example, people are starting to be able to move systematically inside their companies. You can apply for all kinds of different jobs, but how do you decide which jobs make sense for you? Sometimes it’s not so obvious. So, the algorithms could do that for you, and there are companies and vendors who will do that. Yeah, maybe some sophisticated counselor could do it for you, but we got rid of those people a while ago, so it’s not being done.
Scott: You wrote an article in Harvard Business Review called “Stop Overengineering People Management.” From your research, what did you decide is the right amount of automation and engineering?
Cappelli: I think the thing I was concerned about was the assumption, in particular, by a lot of business leaders that this is just the smart thing to do — that getting rid of employees, if you can, is the smart thing to do. And getting rid of their decisions, in particular, is smart, because you can do them better if an engineer makes them. Of course, there’s a long history of this in the U.S. It goes back to Frederick Taylor, in the 1920s in scientific management — the idea that engineers know better, how to do things like pick your schedule or your assignments. The problem is, first, they don’t always know better, and sometimes they can make big mistakes. And then second, even if they could do it as well as you do, people like to make those decisions themselves. They feel more invested in the result if they’re making decisions, and they’re involved in it, than if a piece of software is doing it for them. And the other thing is they just don’t understand, if software has made the decision, why it happened. Therefore you generate lots of complaints and lots of grievances and people are irritate if they don’t know, for example, why they were scheduled to work on Saturday, two weekends in a row. And there’s no supervisor to complain to because they didn’t make the decision.
Scott: In your piece, you wrote that the popularity of the engineering approach versus you know, empowering employees to make their own decisions and solve problems, has increased during economic downturns when workers don’t have the power really to quit even though they don’t like it, and has fallen in upturns when workers do jump ship or protest. And I wonder, are we seeing that now as the economy recovers, people starting to say, “you know what, I don’t want to work like that, at least not for what you’re paying me?”
Cappelli: It’s a really good question. You look at the labor market now, you see a lot of companies complaining that they can’t hire. I think that’s largely because they’re all trying to hire at once now, which doesn’t happen very often, but I think employers are starting to pay attention to what it is that employees might like. You’re hearing this a lot about work from home, for example, right? And the ability to control your own work includes things that software might be doing for you like, what days could I work from home, right? Or what does my schedule actually look like? And one of the reasons this is troubling is there’s been a ton of research showing that employee involvement in decisions really works. They make better decisions about things like how we do our jobs, because they can see more. They’re right on top of it in a way that an engineer might not. They care more about the work when they make some of the decisions or are involved in it. One of the downsides for managers is that it takes effort and a different kind of effort to get employee involvement to work. You actually have to come down and talk to people and be on the shop floor. You can’t just stay up on high. And if you care about managing culture, and you want to get your organization to operate well, you need the employees to be engaged in it, and that requires a very different kind of effort.
Scott: I’m wondering if there are things that people still need to do in HR, like managing conflict between employees? Or is that being automated too?
Cappelli: Nobody’s automating that, and good luck trying to do that, because one of the problems is you’ve got to try to figure out what the problem is, right? You know, I haven’t seen anything that’s gotten close to even trying to do that. And then you have to resolve it, right? And as we know, a lot of conflict resolution comes from simply getting people face to face with each other, especially with an authority figure like a boss in between them. I don’t think a computer monitor is going to be able to do that very well, right? It’s bringing people together, putting some peer pressure on them, or some social pressure on them, and it’s just almost impossible to imagine how you’re going to improve on that with software.
Scott: Well, yeah. It strikes me, the thing about the Amazon investigation, not just the people who were mistakenly getting fired, I mean, that’s awful, as well, but do you find that more companies are tracking every movement their workers make? You talked about truck drivers in your piece. Workers at Amazon reported being afraid to take a bathroom break, because it would show up as time off task. Is this good for the economy?
Cappelli: We have run this experiment before, after scientific management in the 1930s, and we tried to do it with an engineering-driven approach for the next 30 or 40 years. And we basically were taught by the Japanese that engaging employees in these issues is just way better. If you can trust your employees, you can save a lot of money on monitoring, and this is an almost ideological divide within management between those executives who feel you can trust your employees, or at the very least, you should spend the time and effort to create a kind of organization where you can trust them, versus those who believe you can’t. We’re seeing this even with work from home now, which was, frankly, a big exercise in trusting your employees, because we had no choice. But then, toward the end of the process, we see lots of companies investing a lot of money in tattle software of various kinds to monitor what you’re doing at home on your computer, so, kind of the opposite. And it’s a big ideological divide.
Related Links: Insight from Amy Scott
I talked about Peter Cappelli’s article in the Harvard Business Review. It’s a great overview of two opposing management theories about how to get most of out workers. Theory X says workers perform better when they’re tightly controlled and directed. Theory Y says we contribute much more when we have freedom to express ideas and take initiative. Theory Y had been on the rise for the past 40 years, but Cappelli says, after the Great Recession, the pendulum started swinging the other way.
Amazon is a good example. If you missed that New York Times investigation, entitled, “The Amazon Customers Don’t See,” it’s definitely worth a read. It focuses on the company’s JFK8 warehouse in New York. One thing that struck me was how David Niekerk, a former Amazon executive, characterized the company’s attitude toward its hourly workers. He said Amazon founder Jeff Bezos didn’t want those workers staying at the company very long, because company data showed people became less motivated to work hard over time. So, Amazon intentionally limited upward mobility, stopped guaranteed wage increases after three years, and even paid low-skilled workers to leave. Niekerk said Bezos believed people are inherently lazy, which is a pretty dim view of the frontline workforce that helped make Bezos the richest man on the planet.
On the topic of HR tech, Reuters and others reported earlier this month about Walmart’s plans to give free phones to about half its U.S. employees by the end of the year. More than 700,000 retail workers are eligible for the phones which come preloaded with a new app for scheduling shifts, clocking in to work, and accessing a voice-activated personal assistant to locate merchandise and answer customer questions. Workers will only be able to access work-related apps while on the clock, but can use them as personal devices after hours. The phones retail for about $500 each. We don’t know what Walmart paid, but it’s apparently worth it to have workers stay in “constant communication.”
The future of this podcast starts with you.
Every day, the “Marketplace Tech” team demystifies the digital economy with stories that explore more than just Big Tech. We’re committed to covering topics that matter to you and the world around us, diving deep into how technology intersects with climate change, inequity, and disinformation.
As part of a nonprofit newsroom, we’re counting on listeners like you to keep this public service paywall-free and available to all.