Skip to main content

The robots are coming for your office

A conversation with NYT’s Kevin Roose on how robotic automation will impact our future

Share this story

Photo Illustration by Grayson Blackmon / The Verge

As the editor-in-chief of The Verge, I can theoretically assign whatever I want. However, there is one topic I have failed to get people at The Verge to write about for years: robotic process automation, or RPA. 

Admittedly, it’s not that exciting, but it’s an increasingly important kind of workplace automation. RPA isn’t robots in factories, which is often what we think of when it comes to automation. This is different: RPA is software. Software that uses other software, like Excel or an Oracle database. 

On this week’s Decoder, I finally found someone who wants to talk about it with me: New York Times tech columnist Kevin Roose. His new book, Futureproof: 9 Rules for Humans in the Age of Automation, has just come out, and it features a lengthy discussion of RPA, who’s using it, who it will affect, and how to think about it as you design your career.

What struck me during our conversation were the jobs that Kevin talks about as he describes the impact of automation: they’re not factory workers and truck drivers. They’re accountants, lawyers, and even journalists. If you have the kind of job that involves sitting in front of a computer using the same software the same way every day, automation is coming for you. It won’t be cool or innovative or even work all that well — it’ll just be cheaper, faster, and less likely to complain. That might sound like a downer, but Kevin’s book is all about seeing that as an opportunity. You’ll see what I mean.

Okay, Kevin Roose, tech columnist, author, and the only reporter who has ever agreed to talk to me about RPAs. Here we go. 

This transcript has been lightly edited for clarity.

Kevin Roose, you’re a tech columnist at The New York Times and you have a new book, Futureproof: 9 Rules for Humans in the Age of Automation, which is out now. Welcome to Decoder

Thank you for having me. 

You’re ostensibly here to promote your book, which is great. And I wanna talk about your book. But there’s one piece of the book that I am absolutely fascinated by, which is this thing called “robotic process automation.” And I’m gonna do my best with you on this show, today, to make that super interesting.

But before we get there, let’s talk about your book for a minute. What is your book about? Because I read it, and it has a big idea and then there’s literally nine rules for regular people to survive. So, tell me how the book came together. 

So, the book is basically divided into two parts. And the first part is basically the diagnosis. It’s sort of, what is AI and automation doing today, in the economy, in our lives, in our homes, in our communities? How is it showing up? Who is it displacing, who is at risk of losing career opportunities or, you know, other things to these machines? What do we think about the arguments that this is all gonna turn out fine, what’s the evidence for that? And the second half of the book is really the sort of practical advice piece, that’s the nine rules that you mentioned. 

And so it was my attempt to basically say, “What can we do about AI and automation?” Because I think you and I have been to dozens of tech conferences, and there’s always some talk about AI and automation and jobs. And some people are very optimistic, some people are very pessimistic, but at the end there’s always this chart that shows how many jobs could be displaced by automation in the next 10 years. And then the talk ends. 

[Laughs

Everyone just goes to lunch, you know? And it’s like, “Okay, but...” I’m sitting there like, “What do I do?” I am a journalist, I work in an industry that is employing automation to do parts of my job; what should I, what should anyone, do to prepare for this? So, I wanted to write that, because I didn’t see that it existed anywhere. 

You just said, “We’re journalists, it’s an industry that employs automation to do parts of our job.” I think that gets kinda right to the heart of the matter, which is the definition of automation, right? 

I think when most people think of automation, they think of robots building cars and replacing factory workers in Detroit. You are talking about something much broader than that. 

Yeah. I mean, that’s sort of the classic model of automation. And still, every time there’s a story about automation — and I hate this, and it’s like my personal vendetta against newspaper and magazine editors — every time you see a story about automation, there’s always a picture of a physical robot. And I get it. Most robots that we think of from sci-fi are physical robots. But most robots that exist in the world today, by a vast majority, are software.

And so, what you’re seeing today in corporate environments, in journalism, in lots of places, is that automation is showing up as software, that does parts of the job that, frankly, I used to do. My first job in journalism was writing corporate earnings stories. And that’s a job that has been largely automated by these software products now. 

So an earnings story is, just to put in sort of an abstract framework, a company releases its earnings, those earnings are usually in a format, because the SEC dictates that earnings are released in a format. 

You say, “Okay, here’s the earnings per share, here is the revenue. Here’s what the consensus analyst estimates were. They either beat the earnings or didn’t.” You can just write a script that makes that a story, you don’t really need a person in the mix because there’s almost no analysis to that. Right? 

Right. And that’s not even a very hard form of automation. I mean, that technology existed years ago, because it’s very much like filling in Mad Libs. You know, it’s like, “Put the share price here, put the estimate here, put the revenue here.” 

But now, what we’re seeing with GPT-3 and other language models that are based on machine learning, is that it’s not just Mad Libs anymore. These generated texts are getting much better, they’re much more convincing and compelling. They’re much more original, they’re not just sort of repeating things that they’ve picked up from other places. So I think we’ll see a lot more AI in journalism in the coming years. 

So, we cover earnings at The Verge, we do it with a very different lens than a business publication, but we pay attention to a lot of companies. We care about their earnings, we cover them. If I could hire the robot to write the first two paragraphs of an earnings story for a reporter, I think all of my reporters would be like, “Great. I don’t wanna do that part. I wanna get to the fun part where Tim Cook on the call said something shocking about the future of the Mac.” Right? And that’s the part of the story that’s interesting to us, anyway. 

It seems like a lot of the automation story is doing jobs that are really boring, that people don’t necessarily like to do. The tension there is, “Well, shouldn’t we automate the jobs that people don’t like to do?” 

Yeah, this is the argument for automation in the workplace, is that all the jobs that are automatable are repetitive and boring and people don’t wanna be doing them anyway. And so that’s what you’ll hear if you call up a CEO of a company that sells automating software, I mean, RPA (robotic process automation) software. And that’s what I heard over and over writing this book. But it’s a little simplistic, because automation can also take away the fun parts of people’s jobs that they enjoy. 

There’s a lot of examples of this through history, where a factory automates, and the owners of the factory are like, “This is great for workers, they hated lugging big pieces of steel and so now we’ll have machines do that and they’ll be able to do the fun and creative parts of the job.” And then they install the automation and the robots, and it turns out that the workers don’t like it because that was part of the job that they enjoyed. It wasn’t necessarily lugging the pieces of steel, but was the camaraderie that built around that. And the downtime between big tasks. 

we’re seeing automation that is designed not just to replace one task or two tasks, but is really designed to replace an entire human’s workload

Ideally, it would be the case that automation only took away the bad and boring and dull parts of people’s jobs, but in practice that’s not always how it works. And now, with things like RPA, we’re seeing automation that is designed not just to replace one task or two tasks, but is really designed to replace an entire human’s workload. The RPA companies now are selling what they call digital workers. 

So instead of automating earnings reports, you can automate entry-level corporate journalism. Or you can automate internal communications. There are various ways that this is appearing in the corporate world. But I think there’s a gap between what the sort of utopian vision of this is, and how it’s actually being put into practice. 

Let’s talk about RPA. I’m very excited. You’re the only person who’s ever wanted who’s ever volunteered an hour of their life to talk about RPA with me. So, RPA is robotic process automation, which is an incredible name. In my opinion, made to sound as dull as possible

It’s like ASMR, if you wanna fall asleep you could just read a story about RPA. 

[Laughs] The first time anyone told me about RPA, it was a consultant at a big consulting firm, and they were like, “Our fastest growing line of business is going into hospitals and insurance companies where they have an old computer system, and it is actually cheaper and easier for us to replace the workers who use the old computer system, than it is to upgrade the computer system.” 

“So, we install scripts that automate medical billing, and are basically KVM switches, so keyboard-video-mouse switches that use an old computer, like they click on the buttons. The mouse moves around and clicks on the old computer system, and that is faster and easier to replace the people, than it is to migrate the data out of the old system into a new system. Because everyone knows how complicated and expensive that is, and this is our fastest-growing line of business.” 

And I thought that was just the most dystopian thing I’d ever heard. But then it turns out to be this massive industry that has grown tentacles everywhere. 

Yeah, it’s amazing. I mean, my introduction to this world was sort of the same as yours. I was talking to a consultant. I was actually in Davos. That’s not my favorite way to start a story.

[Laughs] 

But we’ll go with it. And in Davos, you know, it’s this big conference. I call it “the Coachella of capitalism.” It’s like a week-long festival of rich people and heads of state. The main drag, the promenade, is all corporate-sponsored buildings and tents and, you know, corporations rent out restaurants and turn them into sort of branded hang-out zones for their people and guests during the week. And by far the biggest displays on the promenade the year that I went were from consulting companies. Consulting companies like Deloitte and Accenture and Cognizant and Infosys, and all these companies that are doing massive amounts of business in RPA, or what they sometimes refer to as digital transformation. That’s sort of a euphemism. 

They were spending millions of dollars and bringing in millions of dollars. And it was like, “What is going on here?” Like, “What are these people actually selling?” And it turns out that a lot of what they’re selling is stuff that’ll plug into your Oracle database, that’ll allow it to talk to this other software suite that you use. The kind of human replacement that you’re talking about. It’s very expensive to rebuild your entire tech stack if you’re an old-line Fortune 500 company. But it’s relatively cheap to plug in an RPA bot that’ll take out, you know, three to five humans in the billing department. 

One of the things in your book that you mention, you call this boring bots. And you go into the process by which, yeah, you don’t show up to work one day and there’s a robot sitting at your desk. As a company grows and scales, it just stops hiring some of these people. It lets their jobs get smaller and smaller, it doesn’t give them pathways up. 

I see that very clearly, right? Like if their entire job is pasting from one Excel database, one Excel spreadsheet to another Excel spreadsheet all day, they might themselves just write a macro to do it. Why wouldn’t you as a company be like, “We’re just gonna automate that”? But all that other stuff in an office is the stuff that you’re saying is important. The social camaraderie, the culture of a company. Is that even on the table for these digital transformation companies? 

It’s not really what they’re incentivized to think about. I mean, these consulting firms get brought in to cut costs. And cut costs pretty rapidly. And so that’s their mandate and that’s what they’re doing. Some of the way that they’re doing that is by taking out humans. They’re also streamlining processes so that maybe you can reorg some of the people who used to work in accounts payable into a different division, give them something to do. But a big piece of the sales pitch is like, “you can do as much or more work with many fewer people.” And I talked to one consultant in Davos, and I’m sorry, this is the last time I will ever mention Davos on this podcast. 

I’m putting your over/under on Davos mentions at five. 

[Laughs] It’s like the worst name drop in the world. But I talked to one consultant and he said that executives were coming up to him and saying, “How can I basically get rid of 99 percent of the people that I employ?” Like the target was not, “How do we automate a few jobs around the edges? How do we save some money here and here?” It was like, “Can we wipe out basically the entire payroll?” 

And “Is that plausible? And how do we get there as quickly as possible?” 

How big is the total RPA market right now? 

It’s in the billions of dollars. I don’t know the exact figure, but the biggest companies in this are called UiPath and Automation Anywhere and there are other companies in this space, like Blue Prism. But  just UiPath alone is valued at something like $35 billion and is expected to IPO later this year. So, these are large companies that are doing many billions of dollars in revenue a year, and they’re working with most of the Fortune 500 at this point. 

And the actual product they sell, is it basically software that uses other software? 

A lot of it is that. A lot of it is, this bot will convert between these two file formats or it’ll do sort of basic-level optical character recognition so that you can scan expense reports and import that data into Excel, or something like that. So, a lot of it is pretty simple. You know, a lot of AI researchers don’t even consider RPA AI, because so much of it is just like static, rule-based algorithms. But a lot of them are starting to layer on more AI and predictive capability and things like that. 

So you get some that are, you know, this plugs into your Salesforce and allows it to talk to this other program that maybe is a little bit older. Some of it is converting between one currency and another. But then there are these kind of digital workers, like you can hire — I’m making air quotes — you can “hire” a tax auditor, who you just install, it’s a robot, and theoretically that can do the work that a person whose job title was tax auditor, did before. 

So let’s say I run like a mid-size manufacturing company, I’m already thinking about “Okay, on the line, there are lots of jobs that are dangerous or difficult or super repetitive, and I can run my line 24 hours a day, if I just put a robot on there.” Then I’m looking at my back office and I’m saying, “Oh, I’ve got a lot of accountants and tax lawyers, and, I don’t know, invoice preparers and all these people just doing stuff. I wanna hire Automation Anywhere, to come in and replace them.” What does that pitch look like from the RPA company?

Well, I went to a conference for Automation Anywhere. This was pre-pandemic when conferences were still a thing. 

And, you know, there were executives on stage talking to an audience of corporate executives and telling them that they could save between 20 and 40 percent of their operating costs by automating jobs in their back and middle offices. And so that pitch, you know, some companies might save less than that, some companies might save more than that, but that’s the sales pitch: You can be more productive, you can free up workers to focus on higher-value tasks. Oh, and also you can shave 20 to 40 percent off your operating budget. 

And so they would come in and they would assess, okay, you use Salesforce, you use an old database, you use some other program, right? I mean, at the end of the day back office work is people sitting down in front of a Windows PC and using it. So they’re like, which of these tasks are repetitive?

Yeah. Which are repetitive? What are the steps involved? There are some stories that I’ve heard of people being sort of asked to train their robot replacements.

To kind of like, walk the RPA vendor or the consultant through the steps of their jobs so that, that can then be programmed into a script. So there’s a lot of that, but there’s also sort of reimagining processes and like, “Do you really need people in three separate offices touching this piece of paper or could it be one person and a bot”? I think part of what they market as “digital transformation,” is just going in and asking people, “What outdated stuff are you using and how could we modernize that a little bit?”

One of the themes here is that maybe the entire national political and cultural conversation about automation is pointed at blue-collar work. Right? It’s a deindustrialized society, we don’t make a lot of things here. Blue-collar workers are hurting all over America. You are talking very much about white-collar workers in corporate America getting replaced by, I mean, let’s be honest, very fancy Windows scripting programs.

Yeah, that’s where the sort of excess is in the economy. I mean, if you go into a factory today, they’re very lean. Most of the jobs in factories that could be automated were automated many years ago. And especially if you go to places like China, I mean, there’re factories that have very few humans at all, it’s mostly robots. So there isn’t a lot of excess there to trim. 

On the other hand, a lot of white-collar workplaces are still brimming with people in the back office who are doing these kinds of repetitive tasks. And so that’s sort of the strike zone right now. If you are doing repetitive tasks in a corporate environment, in a back office somewhere, your job is not long for this world. But now there’s also some more advanced AI that can do kind of more repetitive cognitive work. 

One example I talk about in the book is there’s a guy I met, who’s making essentially production planning software. So this would be not replacing the people in the factories who are working on the assembly line, it’d be replacing their bosses who tell them, “Okay, this part needs to be made in this quantity, on this day, on this machine.” And then, you know, “Two days later we’re gonna switch to making this part and we need this many units, and they need to go to this part of the warehouse.” 

All that used to be done by supervisors. And now that work can be mostly automated too. So it’s not purely the kind of entry-level data clerks that are getting automated, it’s also their bosses in some cases.

That feels like I could map it to a pretty familiar consumer story. You’ve got a factory, it’s got some output. It’s almost like a video game, right? You’ve got a factory, it’s got some output, you need to make X, Y, and Z parts in various quantities and you need to deliver on a certain time. And to some extent, your job is to play tower defense and just fill all the bins at the right time. Or you could just play against the computer and the computer will beat you every time. That’s what that seems like. It seems very obvious that you should just let the computer do it. 

Totally. And that’s the logic that a lot of executives have. And I don’t even know that that’s the wrong logic. Like I don’t think we should be preserving jobs that can be automated just to preserve jobs. The concern, I think I, and some other folks who watch this industry have, is that this type of automation is purely substitutive. 

So in the past we’ve had automation that carried positive consequences and negative consequences. So the factory machines put some people out of their jobs, but they created many more jobs and they lowered the cost of the factories’ goods and they made it more accessible to people and so people bought more of them. And it had this kind of offsetting effect where you had some workers losing their jobs, but more jobs being created elsewhere in the economy that those people could then go do. 

And the concern that the economists that I’ve talked to had, was that this kind of RPA, like replacing people in the back office, like it’s not actually that good. 

It’s not the good kind of automation that actually does move the economy forward. It’s kind of this crappy, patchwork automation that purely takes out people and doesn’t give them anything else to do. And so I think on a macroeconomic level, the problem with this kind of automation is not actually how advanced it is, it’s how simple it is. And if we are worried about the sort of future of the economy and jobs, we should actually want more sophisticated AI, more sophisticated automation that could actually create sort of dynamic, new jobs for these people who are displaced, to go into. 

One of the things I think about a lot is, yeah, a lot of white-collar jobs are pretty boring, they’re pretty repetitive. One of my favorite TikTok paths to go down is Microsoft Excel TikTok. And there’s just a lot of people who are bored at work who have come up with a lot of wild ways to use Excel and they make TikToks about it. And it’s great. And I highly recommend it to anyone. 

But their jobs are boring. Like the reason they have fodder for their TikTok careers is because Excel is boring and they’ve made it entertaining. Those jobs, apart from the social element, are sort of unfulfilling, but at the same time, those are the people who might catch mistakes, might come up with a new way of doing something, might flag a new idea. Is that cost baked into the automation puzzle? 

No. And in fact, I’ve heard some stories from companies that did a big RPA implementation, you know, took out a bunch of workers, and then had to start hiring people back because the machines were making mistakes and they weren’t catching errors and the quality suffered as a result. So I think there’s a danger of overselling the benefits of this kind of automation to these companies. I think some of the firms that are doing this, it’s a little more snake oil than real innovation. 

So yeah, I think there is a danger of kind of over-automating. But I think the problem is that executives in a lot of companies, and I would say this applies largely outside of tech, this is largely in your beverage companies, hotel chains, Fortune 500 companies that maybe are running on a little bit of outdated technology. 

I think the executives at those companies have come to view labor as purely a cost center. It’s like, you’re optimizing your workforce the same way that you would optimize your factory production. You’re trying to do things as efficiently as possible and I don’t think there’s a lot of appreciation for the benefit that even someone like an Excel number cruncher could have in the organization. Or maybe if you retrain that person to do something different, they could be more productive and more valuable to the organization.

But right now it’s just a numbers game. They’re trying to hit next quarter’s targets and if automating 500 jobs in the back office is the way to do that, then that’s what they’re gonna do.

You just brought up retraining. In the book you’re not so hot on retraining. You don’t think it has a lot of benefits. How does that play out? 

Well, the data just isn’t there on retraining. I mean, this is the sort of go-to stock response when you ask politicians or corporate executives, what do we do about automation and AI displacing jobs? And there’s re-skilling, there’s up-skilling. 

There’s telling journalists to learn to code. 

Right, there’s telling journalists to learn to code. [laughing]

And like, you know, you hear these heartwarming stories about coal miners who got laid off and then went to coding bootcamp and became Python engineers, and started doing front-end software development. But those are the exception rather than the rule. There’s a lot of evidence that re-skilling programs actually don’t have a long-term positive impact on the people who go through them, in economic terms. And some of that is probably, you know, about the kind of humans who are participating in them. 

If you are a coal miner, your skill set is maybe not well-matched to be a software engineer. It’s not that they’re not smart enough to do it, it’s that they frankly sometimes don’t want to do it. It’s not rewarding in the same way that the old job was. So the long-term benefit of these re-skilling programs is still something that we don’t have a lot of evidence for. And there’s been some estimates that say private sector re-skilling, companies retraining their own workers, there’ve been some estimates that something like only one out of every four private sector workers can be profitably retrained. 

So we’re really talking about something that needs to happen at the federal level if it’s gonna happen at all. And right now there’s no momentum on that from either side of the aisle in Washington, to do any kind of federal retraining program. 

The politician who comes to mind, first and most clearly in this conversation is obviously, Andrew Yang, who ran in the Democratic primary. He only talked about automation, basically. He’s advocated for universal basic income because he says automation is coming for all of our jobs. Is his approach more focused on the “boring bot” white-collar automation? Or is it at the manufacturing level?

No. And I think this is a place where he and I disagree. I mean, I like Andrew. I think he was right on a lot, but I think, you know, when he’s talking on the trail about automation, he’s largely talking about blue-collar automation. He talks a lot about truck drivers and manufacturing workers and even retail workers. And I’m sort of sold on this idea that those industries are actually not the issue right now; the more pressing and urgent issue is white-collar automation. 

And I think something like self-driving trucks is a great example of something that I am not as worried about as he is, because absolutely there will be self-driving trucks, and absolutely some truck drivers will lose their jobs. And the same goes for self-driving cars and, you know, taxi drivers and delivery drivers. I mean, there’s going to be disruption there, but those are actually like gigantic technological achievements.

They will unlock huge new industries. I mean, you can just imagine, when there are self-driving cars, there will be self-driving hotels and restaurants and gyms, and there’ll be all kinds of jobs popping up for people who are making and selling these cars, who are repairing them, who are programming them, who are developing the hospitality around them. It’s like, there’s gonna be a lot of dynamism in that industry. So while, yes, it will crush some jobs, it will also save lives because it’ll be safer than the human drivers and it’ll open up new opportunities for people. So that’s an area where I’m actually not as pessimistic as Andrew Yang is.

What do you think about universal basic income? 

I think it’s a pretty good idea. I mean, what we’re learning now with the stimulus checks is that giving people direct cash transfers is a really good idea in times when things are perilous and you need to give people a way to stay afloat. And there are other ideas that I think are wise too. I mean right now the tax rate for labor is a lot higher than for capital and for equipment. So companies are actually financially incentivized to automate more jobs because they get taxed less on money that they spend on robots versus on employing humans. So I think equalizing those tax rates could be a way to deal with this on a policy level.

But ultimately I think we have a long way to go on any of this stuff. There aren’t really a lot of politicians agitating for this except for Andrew Yang. So I think my goal is not to give people perfect policy recommendations. I’m assuming some sort of stasis on the government level, and I’m trying to convince people that it’s in their interest to take this into their own hands and come up with their own plans. Because I don’t think the cavalry are coming. 

One of the things that I have talked about, on maybe every episode of the show is how trends have accelerated in the pandemic. And obviously we’re moving to remote work, we’re out of offices. Even maybe three years ago, I was at a Microsoft event and I saw Satya Nadella, CEO of Microsoft. And he was talking about all the things they were doing, and at the end he’s like, “And I just heard about this robotic process automation. It sounds amazing.” 

And now it’s like, oh, everyone’s doing it. Microsoft is in that business. He went from, “I thought it was interesting” to, “If you’re writing robots to use Excel, we’re gonna write the robots for you.” That is a huge business. That’s a great business for Microsoft to be in. Google’s doing it. You mentioned the other two companies that are already big. How much has the pandemic accelerated this curve? 

A huge amount. I mean, I talked to a bunch of consultants who get these calls to come in and automate, you know, the call center or the finance department at big companies. And they said, there are basically two reasons why things have accelerated. One is that, I think, the pandemic has created a lot more demand for certain types of services and goods and created some supply chain issues. And so companies actually need to automate parts of their operations just to keep up with the demand. 

But they also mention that there’s been this kind of political cover that the pandemic gave the executives, because a lot of this technology, the RPA technology, is not new. Like this has been around. It’s not sophisticated, it’s not mind-blowing in its complexity. But it’s fairly obviously displacing workers, and so a lot of executives have resisted it because, you know, it doesn’t save them that much money, it’s not that much more productive or accurate than the humans doing those jobs, and if they implement RPA in normal times, workers get freaked out. There’s a backlash, maybe the mayor of their city calls and asks them why they’re automating jobs. It’s a political headache in the instances when it happens publicly. 

But during COVID there’s been no real backlash to that. In fact, customers want automation because it let’s them get goods and services without coming into contact with humans who might potentially be sick. So it kind of freed up executives to do the kind of RPA automation that they had been wanting to do and have been capable of doing for years. And so the consultants I talked to said, “Yeah, we’re fielding calls from a lot of people who are saying, ‘Yeah, let’s do that automation project we talked about a couple years ago. Now is the right time.’”

You’re gonna come into our back office, while everyone’s out of the office, and figure out which accountants we don’t need anymore.

Exactly, and you know, there’s some precedent for this. I mean, economic disruption is often when big changes happen in the workplace. You’ve already seen millions of jobs disappearing during the pandemic, and some of those jobs might not come back. It might just be that these companies are able to operate with many fewer people.

So you’ve called them boring bots. You say the technology is not so sophisticated. The industry calls it RPA. Like, there’s a lot of pressure on making this seem not the most technologically sophisticated or exciting thing. It comes with a lot of change, but I’m wondering, are there any stories of RPA going horribly wrong? 

I’m just imagining like, I think the most consumer-facing automation is, you call the customer support line and you go through the phone tree. It makes all the sense in the world on paper: if all I need is the balance of my credit card, I should just press 5 and a robot will read it to me, but like I just want to talk to a person every time. Because that phone tree never has the options I want or it’s always confused or something is wrong. There has to be a similar story in the back office where the accounting software went completely sideways and no one caught it, right?

Yeah, I mean, there’s several stories like that in the book. There’s a trading firm called Knight Capital that had an algorithm go haywire and it lost millions of dollars in milliseconds. There was actually just a story in the financial markets — I forget who it was, it was one of the big banks — accidentally wired hundreds of millions of dollars to someone else and couldn’t get it back. And so it was just like, they just lost that. I’m sure that automation had some role in that, but that might have been a human error. 

But there are also lower-level instances of this going haywire. One of the examples I talk about in the book is this guy Mike Fowler, who is an Australian entrepreneur who came up with a way to automate T-shirt design. So, I don’t know if you remember like five or six years ago, but there were all these auto-generated T-shirts on Facebook that were advertised. So, you know, it’d be like, “Kiss me, I’m a tech blogger who loves punk rock.” You know, and those would just be like Mad Libs, you know?

Hang on, I gotta buy a T-shirt.

[Laughing] Or like, “My other car is a flying bike,” or whatever. You know, it was just the weirdest, most nonsensical combinations of demographic targeting IDs, like plugged into T-shirt designs and uploaded to the internet. And Mike Fowler was one of the people who was making that, and he pioneered this algorithm that would take, you know, sort of catchphrases, and plug words into them and then automatically generate the designs and list the SKUs on Amazon and make the ads for Facebook. 

And so he made a lot of money doing this, and then one day it went totally wrong because he hadn’t cleaned up the word bank that this algorithm drew from. So there were people noticing shirts for sale on Amazon that were saying things like “Keep calm and hit her,” or, “Keep calm and rape a lot.” Like just words that he had forgotten to clean out of the database, and so as a result, his store got taken down. He lost all his business. He had to change jobs, like it was a traumatic event for him. And that’s a colorful example but there are, I’m sure, lots of more mundane examples of this happening at places that have implemented RPA.

Is that cost baked in? I’m imagining, you know, the mid-sized bottling firm in the Midwest and the slick top five consulting companies selling RPA, “Everything’s gonna be great.” Then they leave. The software is going sideways. No one really knows how to use it. Like, is that all baked into the cost? Is that just, the consulting company gets to come back in and charge you more money to fix it?

I think that’s how it’s going a lot of the time. The consulting companies end up sort of playing a kind of oversight role with the bots when they malfunction. Because there just isn’t a whole lot of tech expertise in a lot of these companies, and certainly not for things like this. So, yeah, the consulting companies are making money hand over fist on this. There’s no question about it. And this has been a transformative line of business for them because it’s actually like, it’s not that hard, frankly.

And a lot of the stuff is off the shelf. You can go into a company, you know, maybe they haven’t updated their servers in 30 years. And so you’re arriving with this thing that they think is very fancy, but is actually just like a couple lines of code that plug into the Oracle database. So, it makes them look like wizards and it doesn’t require a whole lot of new technology and innovation.

One of the other things you cover for the Times is misinformation, the dark side of the internet. You’re talking a lot about white-collar workers, accountants, back office people, they’re often men. It seems like there’s a real apocalypse coming where a lot of sort of mid-level white dudes in seemingly safe corporate jobs get pushed out of the workplace. Literally your podcast is called Rabbit Hole. Fall down the rabbit hole of YouTube disinformation. Like, I can just add all that up in my head, but it’s not too rigorous. Do you see that connection? 

I do. There is no one stereotype of a person who gets radicalized on the internet. But a lot of people that I’ve run into in reporting on extremist communities have a fairly similar origin story. Which is like, “I graduated from college or community college. I had a lot of debt. There wasn’t a lot of opportunity for me. And you know, I needed a social life and so this was sort of the way that I found status and meaning and friends and a purpose, was by joining an extremist community.” 

I don’t know that the link is sort of causal but I think it’s probably correlated. There’s a reason so many people are out of the labor force, the participation rate is quite low, historically. And so there are just a lot of people who are sitting at home looking for things to do, things to entertain them, things to keep their attention, a sort of mission to plug into. And so maybe for some people that’s an extremist community.

Yeah, I just, for better or worse, I’m thinking about Fight Club, right? Which, you know, it’s a movie that has been framed and reframed many times over the years, but at the heart of it there’s a guy with a really, really boring white-collar job that he hates, and he finds a community that is outside of that. And then they blow up some credit agencies. 

I’m not saying that’s happening here, but the population of disaffected people being pushed out of the workforce has second order effects, some of which can be positive, but many of which are negative. And that doesn’t seem to be factored into the RPA equation, either at the consultancy level, certainly not, and definitely not at the political level.

This is the big error that I think has resulted from giving this whole conversation about automation and AI over to economists and technologists. Because those communities in particular look at things in the long run and in the aggregate. So they’ll say, “Yeah, the industrial revolution wasn’t great in all ways and there was some child labor and, you know, some factories with gross safety violations. But in the long run, people’s lives improved. And you know, we had more time to spend with our families and we weren’t working back-breaking jobs on the farms.” 

“What really sticks out is how much this sucks for people.”

And I think that when I went back and started researching kind of contemporaneous accounts of these past technological shifts, what really sticks out is how much this sucks for people. Like, it’s not a happy experience for a lot of them. I mean, the industrial revolution was horrible for workers. There were these squalid boarding houses where the factory workers would be put and they would be paid barely subsistence wages and they would basically be tortured at work. They would all get sick and it was Dickensian and horrible. 

And so, I think if you had gone to those people and said, “Well, you know, on the plus side, 30 years from now GDP will have risen 20 percent.” They’re gonna be like, “Screw you. Like, I don’t like this. This is not making a material difference in my life for the better, in fact, it’s made it much worse in my immediate circumstances.” And so I think, yes, it’s important to look at what happens in the long run, in the aggregate with new technology. But it’s also important to just listen to the people who are telling us what it’s doing in their lives right now. 

I want to talk about the second half of your book. So the first chunk of your book, Kevin, is very much, “Here’s the conditions of automation, pay attention to this. It’s happening way faster than you think.” The second half of your book is like an instruction manual to you, as an individual, how to dance around the wave of change that’s coming. Walk me through that.

Yeah, this is the happier portion of the book.

Yeah, I save the smallest chunk of time in the podcast for the happy part.

I think it’s important to give people the good and the bad news. The bad news is that you know automation’s coming and it’s gonna displace people. But the good news is that there’s something you can do about it, and it doesn’t require becoming a coder, it doesn’t require going back to school for a STEM degree. It doesn’t involve any sort of productivity hacking. 

What I found in talking to people who work on AI is that it’s actually just about being more human. The things that we can do to protect ourselves, I have nine of them in the book, but more of them revolve around this idea that we are going to need to move toward jobs and activities that can only be done by humans. And that just makes sense, right? When the robots come into your workplace, the stuff that’s left is the stuff that the robots can’t do. So, I was trying to figure out, what can’t the robots do? What is only done by humans right now, and what is likely to only be done by humans into the future? 

And so those are by definition the very human things that I think we’ve been steering people away from, unfortunately, for years. Telling them, “Don’t major in humanities.” You know, I think Vinod Khosla and Marc Andreeson made some form of tirade against how the liberal arts are worthless and everyone should major in engineering and anything else is a waste of your time. 

But if you look at just what the AI researchers are doing, they’re not sending their kids to coding bootcamps. They’re sending their kids to the Waldorf School, where they can learn to dance and play and be creative and express themselves, and they’re not idiots. Like, they know that the skills that are gonna be valuable in the future are those softer human skills.

Give me an example of some of those softer human skills that apply broadly across the white-collar workforce?

I think the big one that people talk about is empathy and I think that is a key part of it. I mean, a lot of the jobs of the future will involve relating to other people. They will be interpersonal jobs, nursing, therapy, social work, that kind of job. But I think that the discussion often stops there. I think there’s a lot of pieces of empathy. One of them is sort of active listening, being able to focus. I mean that’s a really key piece of the puzzle here. You have to be able to control and direct your own attention. Which is why there’s an entire chapter in the book that’s about how to have a better relationship with your phone and the other screen-based devices in your life. I think one prerequisite for being a human is being able to sort of control what you think about. 

Other skills in the book, one of them I talk about is the ability to kind of read a room. This is something that I got from Jed Kolko, who’s an economist. And he is gay and he talks about the experience of growing up as an LGBTQ person and having to kind of fit in, to read people’s states to figure out, “How safe am I here? What kind of code do I have to switch into?” And obviously that’s not great that they have to do that, I wish they didn’t, but he said, basically, that skill of being able to quickly take the emotional temperature of a room is a really important skill for the future. 

And that doesn’t show up in any kind of skills inventory, but that’s gonna be very valuable for the people who are good at doing that. There are lots of others I could go into, but they all kind of boil down to the basic human skills that we nurture in little kids. Sharing, playing well with others, you know? Being a good partner, being a good collaborator, but that we often let sort of atrophy as people get older.

You have a little vignette in your book of the guy who does your taxes, and how he effectively competes with —  I’m sure I’ve even read these ads on these podcasts, like Quicken or QuickBooks, you just like dump the data on them directly from your horrible employee management software at work and then some taxes are generated and they cost $50. But you actually use a person and your vignette is like why his job still exists and how he saw competing with Quicken.

My accountant is this guy named Russ Garafalo, and he is a former stand-up comedian. And one of the things I was interested in when I was looking at this book is finding the survival stories, like who are the people who should have been automated out of their jobs but weren’t, and why? And Russ is a classic example of that. Tax preparation is largely an automated business now. Most people use TurboTax or some form of software to do their taxes. And yet, Russ is there. His firm’s growing. He’s doing well. So, I wanted to figure out why that is.

And it’s because he’s a former stand-up comedian. He’s really funny. It’s really interesting to talk to him, and he’s really good at relating to people in a thoughtful and interesting way, and he hires other creative people and pays for them all to take improv classes because he thinks that those skills will make them better accountants. And he’s right, like it is genuinely an enjoyable experience.

I have to call him soon because taxes are due in less than a month, and I’m looking forward to that. That’s not gonna be a chore for me because I actually enjoy talking to him. So, the sort of human side of any profession is just getting more and more valuable, as automation takes over more and more of the actual functional work of doing taxes. 

How you’re able to differentiate yourself from TurboTax as an accountant, if you’re Russ, is by giving people an experience that they want, and not necessarily being the most eagle-eyed tax preparer. It’s about being the best human.

One of the tropes of all coverage of Gen Z or millennials, or whatever, is, we now pay for experiences over products, right? We spend more money on vacations. I think every generation does this, but these are the tropes of covering particularly people in their 20s because their dollars shift the economy very fast. 

But the idea that we pay for experiences over products, that we pay for interactions over, you know, a fancier car, is that what you’re getting at? Is it, at the end of the day, your accountant is still using Excel, and you could have TurboTax do that for free, essentially, but you want to talk to a person who’s funny, so you’re willing to pay a premium for that?

Yeah. I think that’s the lesson of the past little while here, is that experiences are really valuable for people. And so it’s not just gonna be that people are paying for experiences in travel and retail and hospitality. They’re gonna be paying for experiences when they hire a lawyer or go to a doctor or engage a marketing firm. They’re not going to be paying for, necessarily, efficiency and expertise. They’re gonna be paying to feel something. And that’s one of the sort of rubrics that I’ve used to figure out which jobs are gonna be more stable as we get more and more automated as a society. 

The jobs that involve making things for people are going to become less and less valuable, and a smaller and smaller piece of the economy. And the bigger piece, the growing piece, is gonna be jobs that involve making people feel things. So that’s not an original idea, I’ve gotten that from a number of AI researchers because they point out, this is already happening. You can already see this happening, this kind of artisanal boom in goods and services that sort of have more of a human touch to them than something that’s mass produced in a factory by robots.

Isn’t the counterexample of this already that customer service at big companies is horrible? Like I use Google every day, I use all of their products and services every day. If something goes wrong with Google, my only real recourse is to Google it, which has always seemed Kafka-esque to me. 

That I’m turning to this company that has a broken product to figure out how to fix this broken product. And there’s no one to call. If I have to call AT&T — it’s funny to me that you feel more excited about calling your tax professional than I feel about calling AT&T, right? They should be on the same spectrum. But I know that’s gonna be a negative experience. 

If that is an easy way for AT&T to boost its customer loyalty, to make people feel better about it, why wouldn’t they spend that cost if it’s so obvious?

Well, it’s not obvious right now because I think a lot of companies haven’t gotten very good at that. I mean they’re so involved in the mindset that, you know, customer service is a cost center that should be made as small as possible. But you see this happening on the edges right now. I mean, let’s take Google as an example. The only new successful email product — successful being defined as like, a lot of people I know are very excited about it — of the past 10 years is this app Superhuman, which actually is built on Gmail. It’s like a high-end luxury subscription email product that’s sort of a skin for Gmail, but that includes all this extra functionality.

And one of the key pieces of value that you get when you subscribe to Superhuman is a person, like a rep from the company, does a Zoom with you to walk you through how to use the email. It’s like a very bespoke, concierge model of something that, you know, is free when you just get it on the open market of Gmail. 

But people are willing to pay for that extra touch, that extra part that involves relating to humans and also allows them to get what they see as a better product. So I think that model is transferring to a lot of industries, where you’ll have the kind of mass experience that is purely machine-driven and there are very few humans involved in it, and then there’ll be kind of this luxury skin on top of it that involves much more human contact and connection.

I love the idea that it’s all software at the bottom and it’s just, you get to pay for various levels of people to help you use it in empathetic ways. 

Yeah. I mean, we might all have a team of IT tech coaches. One of the fascinating case studies I came across in the course of writing this book was Best Buy. Best Buy was supposed to die. Amazon was supposed to kill Best Buy many years ago because they sold all the same stuff. The big box was going away, Best Buy was largely dependent on new DVD and video game releases for profits, which went away. So I was interested in how they didn’t die, what they did. And it turns out that they moved to a very high-touch customer service model. 

They started this in-home adviser program, where, for a fee, they would come to your house, a Best Buy rep would take a look at your stereo system and your speakers and tell you which upgrade you needed, or they would sort of be there with you as kind of a personal tech consultant, and then they would sell you stuff on the back end. But the human connection was actually what drove the renaissance of Best Buy. 

It was not that they competed with Amazon on price or logistics, they did do those things, but the thing that set them apart was really that, unlike Amazon, where everything is done by robots and low-paid human pickers in warehouses, they would actually send someone to your house who would talk to you, who would talk you through it and answer your questions.

Several years ago, I talked to the CEO of a company called Asurion, which is a tech support company in Nashville. They sell phone insurance and all sorts of stuff, but their fastest growing line of business is they just sell a subscription to tech support. And they just know every problem that you might have with Bluetooth on your iPhone. And you can call them and they’ll just be friendly and help you. 

And people need it. And there’s just 10,000 people in Nashville who are helping people set up their Rokus every day. And that to me feels like a huge miss, right? Silicon Valley, particularly consumer products in Silicon Valley, pride themselves on being easy to use. But there is an entire company that’s built a business, Best Buy has built a business around how hard it is to actually use. And you can see that just bleeding into the enterprise space.

Totally. And I think companies are starting to realize this. I mean, one example I’ve been looking at recently is a company like Airbnb, which for many years had a very limited customer service department and ability. And then they started getting a lot of people who were angry at them.

The pandemic hit and the hosts were having their stuff canceled, and people were showing up to residences and they’d look nothing like what they looked in the photos. There was a lot of bad juju around that product and the customer service. And so they essentially de-automated that process. They hired a lot of humans and trained them in empathetic communication and so now they have many, many customer service people that you can actually call and talk to. So, I think when businesses get into trouble with the automated model is usually when they start de-automating and bringing in humans. Because there’s a lot that machines can’t do.

We’ve just got a couple minutes left. Your book, the headline is “Nine Rules.” What are the nine rules?

Well, I have to save something for the premium tier “Verge Plus” subscribers.

[Laughs] Yeah.

I’ll list them and we can leave some of the explanation to the people who actually buy the book. 

[Laughs]

Come on, man! I’ve gotta, there’s gotta be a curiosity gap.

Give them eight rules, but the ninth will surprise you.

Oh yeah, the ninth one is crazy, so I’m just gonna read eight. No, okay. Let’s go. Be surprising, social, and scarce. Resist machine drift. Demote your devices. Leave handprints. Don’t be an end point. Treat AI like a chimp army. Build big nets and small webs. Learn machine-age humanities. And number nine I’m not gonna reveal.

Amazing. All of those are curiosity gaps in and of themselves. I don’t know that you gave much away, but it’s a great book. I thoroughly enjoyed reading it. I am just so excited that I got to talk to somebody for almost an hour about RPA.

Finally. 

After years. 

I have been waiting for this my entire life. 

It’s like this dark cloud of consulting on the horizon. It’s just like sweeping over America, and I’m like, “I can see it.” And no one wants to see it except for you. So that was great. Thanks a lot.

Well, anytime you want to call up and just, you know, chat on a Sunday about RPA, you have my number.

And if I can’t get to sleep at night, I’ll give you a call.

Exactly.

Tremendous. Thanks a lot, Kevin.

Decoder with Nilay Patel /

A new podcast from The Verge about big ideas and other problems.

Subscribe now!