Daniel Rock - What Automation Means for How We Organize Jobs

What happens to a job when its tasks are exposed to AI? And why does the order in which tasks get automated matter just as much as which tasks get automated at all? My guest today is Daniel Rock, an assistant professor at the Wharton School and a co-founder of Work Helix. His research focuses on how technology changed the

nature of work within firms and the economy at large. Here's our talk.

Welcome to the Economics of Work. I'm excited to welcome Daniel Rock, assistant professor at Penn, co-founder of Work Helix. We'll talk more about Work Helix, but most importantly, the person that I call if I want to work through a half-baked idea. Daniel, welcome to the show.

Thanks for having me, Ben. Great to talk with you, man.

Yeah, for sure. I'll start with exposure, AI exposure. What exactly is exposure? How should we think about it?

AI exposure is a limited thing. It's a very limited thing. It just means broadly, like first question, exposure. what's exposed. So it could be a task, it could be a job, it could be a system.

Oh, jobs can be exposed? Yeah. What would that even mean? I understand the exposure of tasks. That makes sense. Is the exposure of a job really even a meaningful concept?

It can be. So the exposure for a task means that the task has potential to be changed with AI. Or you can broadly pick any technology and say, does that task have exposure? Could it be changed by that technology? You just need a decent way of generating the measure. The exposure of a job similarly is could the job be changed with the use of AI? And while I agree with you, that's like a little bit less of a meaningful concept. If you can't point to like, what does that mean for that job to be changed? But for the tasks underneath it, you might have like a summary measure at the job level to say like, there's a lot that this person should be doing using AI. So to the extent that there's like. Someone with agency making choices about how they do their work at the job level, they will pick certain things to do with AI and pick ways to allocate their time across tasks. You know, exposure, I think, is often misconstrued with automation. And that's one where, you know, I often have to correct people to say, like, no, these exposure scores, the ones I've worked on, the ones that others have worked on, they don't correspond to automation on its own. And that sort of measure is much more difficult to create, if at all possible, right?

Well, okay. So let me take the pro automation position just to play with it. So I totally get that like interpreting the exposure at some aggregate level, like a job. would definitely not translate to automation of that job. But like, let's say you had tasks at a very, very, very granular level. Could you then interpret it as automation? Like even if you think of a task as being supported by AI, isn't that really just like kind of automating one little component of it and like complementing the remaining component of it? Like, yeah, how should we think through how we could interpret exposure? Like, do you think we could interpret exposure as automation if it were done at a granular enough level.

Okay. So this is where the what you are rating and