Abhast Gupta maps which tasks in medicine and retail can be automated. Routine work like EMR entry and cash handling are vulnerable, but rare complex cases and building trust aren't. He projects 80% of U.S. physicians and 12 million retail workers could become redundant, with 50-80% of all U.S. jobs at risk by decade's end.
Chapters
Let's talk about how this actually impacts jobs. So let's talk about a physician. So I took like a, you know, I thought it was maybe the top 5 things that a physician does and also the top 5 things that a retail sales or cashier does, right? And I just plotted what their points are. So, you know, you're putting things into the EMR, that's like relatively routine, there might be some nuances here and there, but relatively routine. You're writing a prescription for somebody that has diabetes, I mean, almost all diabetic patients fall into like 2 or 3 categories of medications, so you sort of just, you know, you're either starting them on a dose and you're ramping up. It's kind of pretty simplistic stuff on the complexity scale. A routine checkup is something that's, again, relatively routine, but takes a little higher order thinking to do. Let's like something— have you guys heard the term zebras?
In healthcare?
Yeah, so there's always this like, think horses, not zebras. Every medical student thinks that they're looking at the rarest form of some tumor when instead it's just like a zit. So that's like the idea that think horses, not zebras. But zebra cases, these very rare cases, really require someone with incredible expertise, and because they're rare, they're highly unique. So that's something that's like a very human task and will probably continue to be. People have tried to use— that's where Watson is trying to play right now, is in that space. They haven't had a lot of good success, to be honest. So I don't doubt that at some point they will. But diagnosing a disease, that's something that takes a lot of training to be able to do. But 90% of patients have the same problems. It's like cold, cough, diarrhea. Whatever. So that sort of falls in this category. So you kind of see this plot. And the same thing for people on the retail side. You start to see like a— I don't even know what to call it— like a heat map almost of sort of where their jobs are and what their tasks are. And so if you take this now and you kind of superimpose— sorry.
Mm-hmm.
I should do this better. You superimpose this block onto those 2 roles, you start to see now that for the doctor, literally these are the only 2 things that are not automatable, right?
Mm-hmm.
And for the retail and cashier, almost all of this stuff— folding clothes, greeting customers, getting new sizes, checking the inventory, checking the cash register, pulling something else from the inventory— These are all things that are also in the automatable bucket. Really just selling the customers, building that trust, and getting them to buy something, that's sort of the unique thing that machines are still not very effective at. Now, there's 2 points. One is if your entire job fits in the— let's just talk about this for a second. If your entire job fits in this gray area, if all the tasks that you do in a day fit in this gray area, then that is a job that is 100% threatened by automation. So I would think like ATM machines. So the bank tellers that were kind of handing cash out, those jobs are pretty much gone. There's no bank tellers anymore. There's people that work at banks that will help you open an account or kind of consult you on what's the right thing, but they're pretty much gone. So ATMs have sort of fully replaced that task. So if your job fits in this gray area, it's definitely at risk for automation. But then we have a few examples here of tasks that both doctors and— sorry— both doctors and retail folks do that are not in the automation buckets. So Things that are in this top right corner. And so the question is, does that mean that if— so I'd love to get your thoughts. So if most of what you do in a day is automatable, then does that mean that you keep your job or that you lose your job? So what do you guys think? Do you get to keep it even though some parts of it are not automatable? Or do you lose it and just lose those parts? I'd love to hear if anyone has any guesses. I think you keep your job.
Because of the risk, maybe security risk. Automated tellers, they can maybe be— something can hack it or something can be loophole. It's close to all this stuff if you're not hiring a physical human.
Sure. That makes sense.
Yep.
Go ahead.
Doesn't it depend on how important those parts in the right-hand corner To their job? Yeah, absolutely.
Yeah, so let's take the bank teller example. So like I said, bank tellers have kind of evolved now to not do cash dispensing. They're involved in now the account opening and those kinds of things. So what you're noticing in all— what you'll start seeing over the next decade in all jobs is that only the pieces of the job that are out here will actually retain— will remain. The rest will become automated. And that's called— that's basically productivity gains, right? Does that make sense?
Yes.
So here's the challenge, right, is that if I— let's say I had in a retail— let's say a retail— you probably have— let's take a random store. There's probably 2 people that are working the cash register. There's probably 4 people that are working on inventory. There's probably one person that's doing folding, another 2 people doing sizing and greeting, and then maybe another 3 people that are doing selling. So there's probably 20 persons in that retail store, right? But now you're able to automate all of these folks. So how many people are left in the retail store?
3 people.
Like 3 people, probably, right? There's robotics and sort of all the automation that's doing the rest of this stuff. So that's pretty scary, right? So retail makes up 7% of US jobs. Medical doctors, you would think— you might intuitively think, oh, a doctor is not going to be— I mean, that job is safe. Once you go to medical school, you're set for life. Except that the things that doctors are really doing that's not automatable, it's still— it's like 20% of their job. So whereas today there are 1 million doctors employed in the country, once this technology starts to really deploy, we're going to need maybe a fifth of them. So that's 800,000 doctors that'll be out of a job. Right? That's— on the retail side, that's 7%. So let's just do that math real quick. That's probably 12 million retail workers that won't have a job. In the US. And then there's a bunch of folks have done looking at specific jobs and their composition in the economy, and their prediction is that 50%, almost 50% of US jobs are in this automated— automation, are under threat of automation. I think this work was done before deep learning has really started to take force, so I think by the end of this decade I expect that to be closer to 80%. Thank you.
