Is AI Coming for Healthcare Jobs?

Block just laid off 4,000 people and CEO Jack Dorsey said AI made it possible. The AI jobs story is dominating the headlines in fintech and tech, but what does it mean for healthcare, where hiring has been growing and we hear so much about clinician shortages.

David Williams, President, Health Business Group, and John Driscoll, Chairman, UConn Health, discuss where AI is eliminating healthcare jobs, where it's creating new ones, and what leaders should be doing right now to navigate the uncertainty.

🎙️⚕️ABOUT CARETALK
CareTalk is a weekly podcast that provides an incisive, no B.S. view of the US healthcare industry. Join co-hosts John Driscoll (President U.S. Healthcare and EVP, Walgreens Boots Alliance) and David Williams (President, Health Business Group) as they debate the latest in US healthcare news, business and policy. 

GET IN TOUCH
Follow CareTalk on LinkedIn
Become a CareTalk sponsor
Guest appearance requests
Visit us on the web
Subscribe to the CareTalk Newsletter

Transcript

David: 0:00

Block just laid off. 4,000 of its 10,000 employees. And CEO, Jack Dorsey said AI made that possible. So this AI job story is dominating the headlines, at least in FinTech and tech. But what does it all mean for healthcare? Where employment has been growing and where we, where we hear so much about clinician shortages. Welcome to Care Talk, America's home for incisive debate about healthcare, business and policy. I'm David Williams, president of Health Business Group.

John: 0:35

And I'm John Driscoll, the chairman of the Board of Yukon Health. David, it's so good to see you. Thank you for, uh, fitting us into your very busy schedule.

David: 0:45

Well, John, it's very busy. The AI is driving me crazy telling me what to do, so I don't have much, as much time as I might otherwise.

John: 0:51

I I, I would think that with all of this AI stuff that you'd be kind of on the beach, I mean, you know, given all of your different engagements can probably been done better by AI agents. But let's be, before we get into, you know, how it's dis displacing you and empowering the tech overlords, what do you, what, where are we in this kind of ai? It's sort of a stock apocalypse. Um. This, this sort of AI fear about jobs, employment, and engagement?

David: 1:27

Well, John, what's pretty clear is that AI can do a lot of what white collar employees do, and so the threat of job loss is very real, but it's also been a little bit hypothetical.'cause there's some people who say, well, if it can do 40% of what a person can do, 40% of the jobs are gone. There's others that say, well, if it could do 40% of what a person can do. Well then that person will have time for other things. So I think this block example is very concrete, if you will, that says, yeah, some companies are really going and doing that. But in tech where we know companies overhired during the pandemic, is it just a matter of the reset? Whereas in healthcare, we always hear, hey, there's gonna be a shortage of X number of nurses and Y number, you know, of doctors. And so maybe it's just in healthcare is a little bit different. That's what I think is really interesting, John. I,

John: 2:18

I, I think you're absolutely right, as much as it pains me to say it. And by the way, block and concrete, cute, cute, cute, uh, cute, cute use of, uh, adjectives there. You know, Vinod Kla, I believe the, one of the famed technologists. Technologists and tech techno futurists, who's a brilliant investor and technologist in his own right. I believe it was he many years ago who said that, you know. Based on the dawn of, of, of data and the convergence of early stage AI that we would have virtually no need for radiologists going into the future. And what really has happened is that radiology employment is up. Not down as those models and the technology and the data has become, uh, more accessible. Um, part of it is that we still need humans to play an interpretive role. And part of it is there's more roles to play when you've got more insight, more data, and more complexity to what the radiologists who are reading these scans. And so I think one of the interesting things, uh, about the AI moment we are in. It's this, it's, it's, it's fear, hope, and confusion com at a time when the tools are getting more and more powerful, almost magically powerful. But I'm not convinced that in healthcare we're not gonna actually allow people, humans to do what humans do better. And we may actually have more jobs, um, for, or, or allow clinicians and non-clinicians to do the work they want to do as opposed to the busy work of administrivia.

David: 4:00

How about medical transcriptionist, John? So I think with radiology you've got this example of where, okay, well radiology is about pattern recognition and a computer can pick it up better and more accurately and make comparisons. And yet, as you said, the radiologists are, you know, now are, are busier. They have more things to do. Medical transcription, it used to be I'd listened to what the physician dictated and type it in, maybe use a little technology, but now it's 99% automated so you can see something very dramatic. As well.

John: 4:27

And I think, you know, the, the Becker's research, which we've talked about is interesting. It looked at about a hundred health hospitals and health systems and found that there were some workforce reductions, but they were, the majority were hitting administrative and rev cycle roles. And I, and I think that there's an enormous amount of excess labor associated with, um, rev cycle. You know, I'm chairman of a. Software business that's in the rev cycle business. And it's not that the demands are getting any smaller, but to, for hospitals to get paid more effectively and faster and more efficiently. It's just that we can leverage software to do more of those jobs. But in many cases, the white collar jobs you're worried about. The higher end ones are ones that are, are. If anything, growing in importance. Where you do see an opportunity for cost reduction is in some of the entry level, truly administrative, moving, you know, moving paper from one place to another. Things like medical transcription, which is really trend taking the information you're getting from a patient or a doctor and accurately documenting it so that A, the doctors and the teams can figure out what you need to. When, when you need it as a patient, and secondly, that those doctors and hospitals get paid, those, those kinds of things, which were, are part of the 20 to 30% of healthcare that's administrative costs. That's two to three times what it is in other countries, I think, I do think that you're gonna see cost reduction and job elimination there, but I, I actually think on the core things that people think about in terms of healthcare. That we may be in a, a golden age. I mean, do you think I'm being too optimistic there?

David: 6:15

Well, you didn't, if you said platinum age, maybe I, I would, but, uh, or diamond Age, but no, I think there, there's some. There's some positive elements here other than just the pure math about adding and subtracting employees. So there's some examples where, um, you know, AI scribes, I think Tampa General Hospital did some work on this where they've got scribes, and it doesn't mean that you can just grind out more visits from the doctor. There's less burnout, less turnover, um, and lower recruiting costs to keep those physicians employed. So, you know, uh, healthcare has been pretty, a pretty difficult environment for people to work in. You see these polls about how many people are planning to leave their jobs, uh, you know, and so on. So in some ways, the technology can make the existing jobs easier, even without necessarily making them more productive. I think that's where you can start to get to the, at least the silver era, if not the golden one.

John: 7:07

Well, I think, you know, I was it Houston Methodist, that's projecting a 30% reduction in scheduling, registration, consents, prior auth and rev cycle through their application of agentic ai. I mean, David, that's gotta be good news. I mean, who wants to spend more time entering the same consents every time you move floors or move offices? I mean, it, it, it's sort of nuts the way that administrivia gets between the patient and the provider.

David: 7:36

You know, it's interesting, I was trying to think this through actually, based on our last visit out to San Francisco, during the JP Morgan Healthcare Conference, and I'm thinking about relative to Waymo. So a lot of times you'll get an agent or something on the, you know, on the computer or on the phone, and then he said, lemme talk to a person. Actually wanna talk to a person, an agent.

John: 7:54

Being a technology assisted, voice assisted, basically you're dealing, you're dealing with a bot, you're not dealing with the person.

David: 8:02

That's what I should have said, John, like a chat bot, uh, or something like that. And say, well, okay, let me talk to a person and they'll solve it. But I think increasingly people are having the experience. If the AI is good, they might actually rather talk to the AI than to a person. The reason I draw this, I tie this with Waymo, is a couple people who are. Tech Forward told me, yeah, try Waymo. And then also they, they like it better than an Uber 'cause they don't have to talk to the driver, they don't cancel on them. You know, it's more reliable and they think it's safer. And so I wonder whether we're getting to some of the point in healthcare, we'll say, actually I'd rather talk to the AI and certainly for scheduling, maybe I'd rather do that.

John: 8:39

And just to be clear, the Waymo is not a, a new version of a children's game. It's the audit. Those are the, the fully autonomous. Cabs that are available in places like San Francisco, Austin, some other cities. Um, but they're actually being banned in other cities. But they are Google's investment in driverless cars. And I, I agree. If you get out, you're out in San Francisco and you're, you're. You, you want to be taken care of on demand. The cabs are somewhat unreliable. The Ubers definitely can cancel and you don't quite know who you're gonna get at any moment. The Waymo's have a very consistent robotic simplicity to them, and they're very dependable. So I think, I think that's, it's we, we are at the point at which I think the combination of the large language models. The data that's available, our ability to actually compute, which is to say to make, get the right answers to your questions or drive safely. We are at a tipping point in some of these areas, particularly around things like transcript, transcription, problem resolution, and driving. But do you think that's a good thing or a bad thing for healthcare, David?

David: 9:54

I think in healthcare it, it is good to have more automation if it's reliable because the human processes are not very reliable. And also they can be very, um, intensive, uh, time intensive for people who are, who are sick. I, I'll give you an example from scheduling, which is one that we had mentioned in an example of Houston Methodist. So I was trying to get an appointment, uh, with the surgeon and. They said, well, they're full, but you can, you can call back, you know? And I said, well, do you have a waiting list? No, but you can call back as often as you want and try to reschedule. So they're, they're kind of telling you like, yeah, you can call every day. You know, that's crazy. On the other hand, um, I. I had a, a friend who I was trying to get into a surgeon, and because I know them, then they sort of did like a, you know, a workaround and they said, yeah, uh, call my assistant and I'll have you seen before or after my, you know, first or last patient there. So, I don't know. On balance, I do think the automation is actually gonna be helpful and it will lower cost. Which is helpful. So I, I would like to see more automation if it's really reliable.

John: 10:56

Well, and, and I think you raised some a point earlier, like, let, let's, let's the block example freaked people out and there's been a, a, a vast, a, a pretty material move. Downtick in software stocks is people think that Ag agentic, AI, and LMS are gonna take over everything. Whether, I don't think either of those is, I think both of them are probably a little bit overstated to your point. I mean, Jack Dorsey, who runs block, also ran Twitter and which had a 70% reduction in labor after Elon came in and apparently runs just fine. So he may not be the most reliable source of, of labor guidance, but when you are looking at the demographics of we're gonna be. You know, we're a couple of hundred thousand clinicians short in the not too distance future as nurses and doctors retire. I think, I think the AI agentic support for getting rid of administrivia and allowing certainly providers to practice the top of their license is, is gonna be a, gonna be a gift. The other thing we should talk about as we think about AI in healthcare, it's a pretty big priority for, uh, the Center for Medicare and Medicaid. Uh, and they are, they. You know, a lot of the emphasis of the priorities of administrator Oz and, and at times, uh, secretary Kennedy are around leveraging technology to simplify the interface of providers and patients, and also, frankly, to extend patient access to where providers aren't there. I mean, there's this CMS wiser model where they're gonna actually have a Medicare programs launching AI assisted prior auths. Pilots in six states. Uh, the, the premise being that if we could, if the prior auth is probably the, the thing that patients like the least and providers like even less because it means you have to do the mother may I thing with your insurance company. And I think both sides realize that it is a patience in. And providers at least, that it slows down and complicates things. Even with the best health insurance companies, it, it's a target for this administration to get rid of it because they believe that, look, if it's logic driven and rules driven, it should be automated and simple. I think that could be a, a big patient upgrade and again, allow providers to spend less time on the phone with their insurance companies and more time with patients.

David: 13:21

That would be a good idea, John. That's, that's a good use case for it. And as you say, like why should it actually need like a separate step that takes it outta the workflow, prior authorization as opposed to just, Hey, if it, if it fits these criteria, then it's, you know, then it's approved. I do think that's a good, that's a good spot for it. I. I want to talk about agentic AI again for a minute.'cause I think on the one hand, the, uh, the idea of if ai, um, having all these agents that can act on your behalf is, is wonderful. But as you know, I do a lot of work also in cybersecurity and there's a lot of questions now about, well, gee, can I use AI agents? Who is your agent working

John: 13:55

for?

David: 13:57

Well, so the, so what is, I, I talk to people who are more technical than I am on this and they say, well, the, the problem is that there's not a big overlap between things that the agent can do that are useful and that are secure, so you can make it secure. So as long as you don't let it do anything useful, otherwise there's some code that's kind of locked away and doesn't have access to anything that you're doing and, and kinda the emerging approaches that to you have to treat. Agentic AI is not just something that you like let loose on your machine or your account, but actually treat it as like a different, almost as an individual, like an assistant that should have its own email address and have its own access and, and it works for you, but isn't completely intermingled with what you're doing. There's a lot to come, uh, in that area. John and I, I think it's really too soon to say where it's gonna play out, and we could definitely have some disasters in the meantime.

John: 14:44

So are you in favor of, I mean, there's a lot of back and forth with the feds around whether we should do any regulatory oversight at all in AI that the, uh, czar of AI at the White House, David Sachs would suggest No. And yet you have. Some real concerns at the state level, and not just in blue states, but red states too. Uh uh, Flo very famously, governor Flo, uh, Florida, Ron DeSantis said There has to be some AI regulatory oversight, or we're gonna end up with the same disaster we've had. Social media not being regulated and, and driving an entire generation of kids to be, you know, to, to, to to of an epidemic of self harm and, uh, lack of attention. You know, there is a, there's a, there's a, there are movements in a number of states to try to regulate and that, but, but, but there's this federal preemption being what, what, what, what do you think should be the guidance, given that you've just made a really interesting point about if it's useful, it's. Unsecure and if it's secure, it's probably not useful.

David: 15:48

So social media has been, you know, not an unmitigated disaster, but there's some really some terrible things that have happened societally. And the potential for AI is an order of magnitude worse of what could happen. And you see that because whereas with social media, some people said, yeah, we have some concerns inside. Like about what's the impact on, uh, uh, self-image of young girls, you know, people at Facebook that said that. But you know, people that are working for the AI companies who really think it's gonna be the end of the. Humanity or the end of the earth is a, is a whole nother level. And, and really, yeah, I mean, we are in a, in a dangerous, um, spot. And, uh, it, you know, as things are kind of getting outta control and we have a chance to potentially do something about it, the states can't really do anything. Uh, they have no choice but to try. Um, but you need to see some change, uh, at the federal level. Um, and I don't see it coming this month. I'll say that.

John: 16:39

No. And, and no. And, and I I think it's interesting because, you know, president Biden had tried to collaborate with the large AI companies to come up with a compact around what safe AI looked like. And it, it's, it's, um, that was kind of trashed by. Folks like Mark Andreessen and others as, as, uh, as, as government controls. I'm not sure that was at all fair or accurate at the time. Um, and David Sax has gone, or the current Trump, AIS r has gone the other direction. I do think this is an area where we need to tread cautiously because there's a balance that the federal government has to place between competition with other countries, because I do think the AI race. As President Xi and, and, and, and, uh, prime Minister Putin have both talked about. It's a, it's an existential race if you can win it and, and sustainably stay ahead of your competitors from a global perspective. At the same time, these are super dangerous tools. I mean, we're already seeing AI. Local ais, the open clause, the the the Agentic program, which was hacked by a, a guy from I believe Australia, which allows you to deploy a local agentic AI to control some of the things on your computer. And one of the challenges is once you give the agentic AI some powers of au some semi-autonomous powers, it does some of the things you want and some of the things you can't predict, which is exactly where we do not wanna be. Healthcare. So I'm, I'm kind of with you. I'd love to see a regulatory framework that, uh, tried to balance those two things. Again, I think this is going to be, in general is gonna grow employment. It's going to lower costs and improve outcomes. I believe in terms of the healthcare we provide today, it doesn't mean healthcare is gonna cost any less.'cause we always seem to find new things to spend money on in healthcare. But it, you know, at a time when we need labor. AI can help. At a time when we need to get rid of administrative stuff, AI can help. I just worry that we're, we're sort of doing it without any regulatory oversight whatsoever when we know that there are some problems coming.

David: 18:56

So, John, let me bring it down to healthcare and specifically the role of executives, board leaders such as yourself can play. And one is to say, you know, how should healthcare executives be communicating with their workforces right now? I mean, obviously I don't think that we embrace the, the block approach of saying 40% of you are gone, get rid of people's. We'll see what happens. But what, like what? What is the right role? I mean, you see unions organizing around this. Point about fear of AI in healthcare, and I think that, you know, it's, it's, it's, it's right for people to be worried. What is the leader, what should leadership be doing now with all of this uncertainty?

John: 19:32

Well, as a chairman and as executive in some firms, what I'm trying to do is to make sure that everyone, even at the board level, has some. Working knowledge and use of AI and agents. If you're not playing with these tools, you don't really, it's really hard to have an opinion about it. So the first is to get comfortable with, uh, the tools themselves and really learn about it because it's as, it's as important as. Electricity was, it's gonna change and affect a lot of most things we deal with on a day-to-day basis, visibly and invisibly in ways that we can measure, in ways that we can't predict. So you wanna have some working knowledge of it. And I think, again, I think that for any executive you want kind of guardrails. On governance. Um, in my case, you know, one of the things we are communicating in the companies I'm working with is AI is not gonna be used to just eliminate jobs. It's going to be used to empower people who have the jobs they have today because frankly, there's more demand than there is supply of good care. And good solutions in healthcare. But that's a, that's a choice, David. I mean, I think some executives are gonna take the, um, the chainsaw Jack Dorsey approach of just hacking jobs and seeing what happens. I don't think that's a sustainable strategy in the healthcare ecosystem. And I think others are gonna try to take the ostrich strategy, which is, we'll, ignore it and see what happens. No, both of those are gonna fail. Um, this is, this is a massively important. Change in the way all work happens and the way people leverage information. And I think that, uh, and em and it and, and it, and it's gonna empower different constituencies in different ways. You gotta be playing with these tools and you gotta put up guardrails and have goals.

David: 21:21

Well, thanks for that, John. I look forward to continuing to explore these topics, uh, as, as the AI revolution unfolds. Well, that's it for another episode of Care Talk. We've been discussing today whether the AI Jobs Reckoning has arrived in healthcare and what it means for workers, executives, and policy makers. I'm David Williams, president of Health Business Group,

John: 21:42

and I'm John Driscoll, the chairman of the Board of Yukon Health. If you liked what you heard or you didn't. We'd love you to subscribe on your favorite service.