Listen and followApple PodcastsSpotifyYouTubeMore apps · Browse all episodes
This episode is brought to you by ModMed.
Fear of AI is not the problem. Refusing to test it is. Eric Jennings, a board-certified, fellowship-trained ophthalmologist and the senior medical director of ophthalmology at ModMed, argues that hesitancy about AI is a protective clinical instinct, and that the right response is the one physicians are already trained to use: appraise it the way you would appraise a new drug or a new procedure. You will hear the line he draws between cautious skepticism, which asks to see the validation data, and paralysis, which demands a guarantee no technology can give. Eric explains how to read a vendor’s training set the way you would read the N on a clinical trial, using the pneumonia algorithm that scored about 93 percent in the lab and then failed in clinic because it had learned to recognize portable X-ray machines rather than pneumonia. He lays out the guardrails to put in place on week one of a pilot, the human and technical signs that a practice moved too fast, and why the administrative side of the office, not the chart, is the safest place to start. This episode is brought to you by ModMed. Find out more at modmed.com. Press play to hear why he believes AI will arrive as an integrated ecosystem rather than one more bolted-on tool, and why the doctors who use it will replace the doctors who refuse it.
VISIT SPONSOR → https://www.modmed.com/
SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast
Transcript
Kevin Pho: Hi, and welcome to the show where we share the stories of the many who intersect with our health care system but are rarely heard from. This episode of The Podcast by KevinMD is brought to you by ModMed. Subscribe at KevinMD.com/podcast.
Today, we welcome Eric Jennings, a board-certified, fellowship-trained ophthalmologist and the senior medical director of ophthalmology at ModMed. He maintains an active surgical practice in Atlanta, Georgia, specializing in cataract, corneal, and refractive surgery. And at ModMed, he helps develop and train the company’s native AI tools, bringing a practicing physician’s perspective directly into the technology ophthalmology practices use.
Today, Dr. Jennings joins us to discuss why fear of AI in health care is not just pre-implementation anxiety, and why the right response is to treat a new AI tool the way you would treat newly published evidence. Pin down the risks, the benefits, and whether it actually helps you care for patients. Find out more at ModMed.com. That’s M-O-D-M-E-D.com, and a link will be in the show notes. Eric, welcome to the show.
Eric Jennings: Kevin, thanks so much for having me. I’m actually a longtime fan of yours, of KevinMD and your website, and all the amazing physician content. So thanks for having me.
Kevin Pho: Oh, thank you so much for listening. I couldn’t do it without physicians like yourself listening to my show and reading my site. So tell us about yourself and your path in medicine, and what led you to focus on AI and digital health.
Eric Jennings: Sure. So I appreciate the intro. I’m a cataract, cornea, and refractive surgeon. I practice out of Atlanta, Georgia. I’ve been in practice out of fellowship for about 10 years.
I was just like every other ophthalmologist. I had a little interest in technology. It contributed somewhat to my choice of specialty. We love our toys and gadgets and lasers and whatnot. And while I love the clinical side and patient care, over the years I saw a lot of the friction and inefficiencies firsthand, and that made me realize how much technology has been helping the way we deliver care, but could also continue to help.
And I decided I wanted to help solve those problems on a larger scale, potentially more upstream, which sort of drew me into digital health. So instead of waiting for it to get into our laps, I wanted to help out first. So now I blend both worlds. When I’m not seeing patients, I’m the medical director of ophthalmology at ModMed, so I use my experience in the clinics and the OR to help build and train AI tools that hopefully are making life easier for doctors and therefore translate to patient care.
Kevin Pho: So when it comes to AI, it goes without saying, and my audience knows this well, that the progress and innovation has been growing at an exponential rate. And when I talk to physicians, they’ve been burned before when it comes to that intersection between technology and medicine, right? I just have to say the words electronic health record, and a lot of doctors are fearful of another technology wave. So when it comes specifically to AI, what are some of the main reasons why you think clinicians are fearful of AI today?
Eric Jennings: It’s a great question, and I’ll join the club, right? Fear is warranted. I will give everybody an out here. It’s our protective clinical instinct, right? We’re trained to first do no harm.
These are algorithms and computer programs that are threatening our sense of control and our sense of liability. These are human lives at stake. In my world, it’s vision, it’s sight. No matter what you treat, you have things built just the way you like them, and anything new comes with some hesitancy.
We’ve gone through hours of training to get things just right, and you’re asking us to take on this new paradigm shift. I’m scared, too, but it’s what you do next, I believe, that separates those who are going to get stuck from those who may succeed. So I think the hesitancy is warranted, but fear may be taking it a little too far.
Kevin Pho: So tell us about some of your colleagues. Just looking at you, you look like you’re relatively in the first half of your career. But how about those colleagues, the physicians that you work with, fellow ophthalmologists? What are you seeing in terms of how they’re accepting AI or not?
Eric Jennings: It’s pretty split, right? First of all, thank you for saying I look like I’m in my first half. I suppose that’s so. My kids and my gray hair may beg to differ. But it is split. There are people who are so paralyzed with the fear that they say, “No, I’m not doing it. It’s never happening.” Then you have the other side of the spectrum that are gung ho, “Let me help. Let me implement as much as I can.”
Just like anything else, when you have such polarization, there are risks to both. You don’t want to move too quickly, but you don’t want to be paralyzed and not do anything. I think the people who are dipping their toes in and trying to figure out how it works best are probably setting themselves up for success.
There is a little bit of an age gap, too. There are people who say, “This is how I like it. It’s never going to happen,” and there are some who are a little bit more tolerant to change and more willing to shift their paradigm some. So I’m seeing a wide breadth across my specialty and medicine in general in terms of adoption of things like AI.
Kevin Pho: So it’s a fine line, right? The difference between productively cautious skepticism and paralysis and not using that technology at all. And I think that line is shifting every day as we’re discovering new uses in that intersection between AI and medicine. So today we’re speaking towards the end of September. How do you distinguish that line between cautious skepticism and paralysis?
Eric Jennings: I think caution, if you’re being careful, is you say, “OK, but prove it. Show me the data. Show me the validation. I will consider it.” Paralysis, a lot of what I’ve heard, is, “Guarantee me.” Right? “Guarantee that it’s never going to hiccup, it’s never going to make a mistake.”
I don’t think that’s a reasonable expectation, and if we’re translating this back to medicine, treat it like a new drug, a new procedure. This drug comes out for a condition that you treat, there are going to be side effects, but you’re going to review the literature, you’re going to start with a little bit of a trial, and then decide whether it fits your practice and your patient.
So instead of refusing it altogether, which is the paralysis, “Nope, nope, not happening,” go in with a little bit of cautious skepticism and say, “Well, OK, I’ll accept your premise, but show me the validation. Show me the data, and then I’ll take it from there.” And that’s what we’re trained to do.
Kevin Pho: So a lot of the AI tools today that physicians use require a certain amount of supervision. It could be AI scribes, it could be AI clinical decision support. Physicians always have to be that proverbial human in the loop to make sure that output is clinically accurate. Now, a lot of them feel like it’s another burden. In addition to everything they already have to do, it’s just one more thing that they have to do, right? How do you respond to a colleague who says, “I didn’t go to medical school to babysit an algorithm”? How do you respond to that line of reasoning?
Eric Jennings: I think, once again, it’s somewhat warranted. You’re right. My first reply is, “Well, we didn’t go to medical school to do a lot of the things that we’re being asked to do these days,” right? Data entry clerks, memorizing the LCDs for payers. I didn’t go to medical school to do all that.
But you’re going to have a choice. Either you let a well-trained AI agent do it for you with some oversight, or you’re going to have to do all this yourself. If you already have somebody, let’s say it’s an MA, a technician, a scribe, as we have in ophthalmology, they’re already potentially doing things for you. It’s the same thing. You, as the clinician, are responsible for the output and the outcome. It all comes back to us, which is fine. We’re up for it, I think.
So it’s really not babysitting, it’s responsible oversight. But because it’s different from what you’re used to, it feels a little scary. So you didn’t go to medical school to babysit, but you’re ultimately responsible. If you have an assistant, you are supervising them. The same thing holds for an AI tool.
Kevin Pho: Well, the question always comes up: Is AI ever going to replace physicians? And I think I had a guest on previously, and they said that AI isn’t necessarily going to replace physicians, but the physicians who use AI appropriately are likely going to displace those that don’t. What do you think about that?
Eric Jennings: I completely agree. That quote’s been circulating, and I think it’s well said. There are some cases for AI replacing physicians, but I don’t think in the way that we tend to think about it.
The best example I have is, shout out to our radiology colleagues, right? Because AI is trained on patterns. So what do radiologists do? They read films. They look at X-rays and CTs and MRIs and films, and so you can train an AI to find abnormalities. Well, the American College of Radiology said, “Well, you know what? Let’s embrace it. Let’s build this ourselves and put some guardrails.” So now the AI can say, “Look, I’m going to separate out definitely normal and definitely abnormal, and we’re going to send you the abnormal ones. And you can still practice and do what you want to do, but we’re going to offload some of the redundancies and low-level stuff that are clogging up your days.”
And I think that’s really how things are going to pan out if we as physicians really help build this and have our voices heard in terms of the uses, and maybe even fight back about, “No, it shouldn’t do everything that I do.” There’s some clinical intuition that may not be able to be matched by AI.
Kevin Pho: So whenever you go to a conference, right? AI is the buzzword everywhere. Every vendor at every conference describes their product as AI-powered. Now, how can practices differentiate between AI tools that actually help their workflows versus act as just fancy window dressing? How do you tell the difference?
Eric Jennings: So if anybody goes to conferences, trade shows, and you go to the expo floor, you’re right, Kevin. You’re going to see AI-powered this and AI-powered that.
The first step is to look for things that actually solve problems. There’s a big push to say, “I’m going to develop this cool AI tool that looks great, and we have a great logo,” but it actually creates more work for you to do. So it’s adding work. I don’t care if it’s AI or not. It’s not solving a problem.
Also ask, what does AI even mean? Anything can be AI-powered. Was it built natively? Is it internal? How is it trained? You don’t have to know everything about AI. For fun, for the listeners, you can go on ChatGPT and say, “Explain to me basically how AI works and what questions can I ask?”
But when it comes to these companies, you can slap your name on anything and call it AI-powered. Let’s say you’re into revenue cycle, right? I want to help doctors get paid. OK, well, the user has to manually enter in all their data, and you plug it into OpenAI’s ChatGPT, and it returns something cool. Well, congratulations, that’s AI-powered, but it’s adding work to what you need to do, and it’s not going to be helpful. So ask those questions about does it solve problems, how is it built, and what does it mean by AI-powered? Because that’s a nice marketing tool, but the clarification matters.
Kevin Pho: So when evaluating an AI tool, how can providers verify that the training data actually reflects their practice’s specific demographic and acuity mix?
Eric Jennings: I would say the same as we do for any clinical trial, right? If you’re looking at a new study that comes out, if you look at the N, and the N is 12, that’s not a really good training set. So number one is ask about the demographic and acuity breakdown of the training set.
The way AI works, if you’re not familiar, is you really do get a huge data set, similar to a clinical trial. You feed it through this algorithm, and you look at the output. If you’re a physician who treats an elderly rural population and this algorithm by a company was validated using healthy 20-somethings from California, the algorithm is not going to work.
Don’t just accept the response of, “Well, it was lots of visits, lots of data.” Ask the questions. How many? How many visits? Was it specific to your specialty? Was it not? Either could be OK. And who validated it? I’m proud that I’m a practicing clinician who gets to vet a lot of these visits to make sure they are for my specialty.
I’ll explain this interesting example in terms of validating. There was an algorithm that could detect pneumonia with high accuracy in the lab. I think they did this in the AI lab. It had 93 percent accuracy or something to catch pneumonia. But it failed clinically. Why did it do that? Because the AI wasn’t really finding the pneumonia. It was tracking that the portable X-rays had a higher correlation with pneumonia because those patients were sick. They couldn’t get out of the bed, so the portable X-ray machine was used for sick patients. The ones who were captured as an outpatient were healthy. So when they deployed that to the outpatient setting, it messed up left and right.
So once again, you don’t have to know everything about AI, but know: Was this a broad training set that will really work for your clinic and what you do in practice?
Kevin Pho: So let’s talk about adopting an AI tool. In a lot of practices, they run pilots first. I know in my organization, before things are rolled out, it’s run as a pilot in a small setting with a certain number of clinicians. Now, talk to us about the operational guardrails a practice should establish on week one of an AI pilot.
Eric Jennings: I would say the first thing you need to do is establish a human in the loop. I think you alluded to that at the beginning of our conversation, but even though these things may be marketed as, “Oh, we’ll take the place of this and replace that,” you still have to have human oversight. So 100 percent, the human has to be there.
Start with tasks that are low stakes. Over-communicate with your staff and with the patients about the pilot. This is no judgment. This is something brand new that we’re implementing in our clinic or OR that’s going to change things, so start slow.
There has to be a feedback mechanism, right? So how is it working? How is it going? Nothing that gets pushed out is perfect, and even if the company says it is, once you put it in practice, it may be different for you. So human in the loop, oversight, and feedback.
Don’t start out, when you bring something into your workflow, by saying, “Well, this tool is here to take over X, Y, and Z.” That’s never going to go over well and you’re going to be met with a lot of friction. So vet your tool, analyze it, see how it’s going, rinse and repeat, especially week one, and I would say even through month one.
Kevin Pho: So are there signs that a practice has implemented AI too quickly, without sufficient operational guardrails? What are some examples of what they may look like? And if you have a story or a case study to share, I’d be interested in hearing that as well.
Eric Jennings: I think some signs I could probably break down into the human and the technical side. From a human standpoint, if you went too fast, you just get low buy-in from your staff. They think it’s replacing them, so they’re going to be silent or they’re going to complain nonstop. It’s just like somebody else coming to take your job. You’re going to potentially try to get them fired or you’re going to complain about them nonstop.
I don’t have a specific use case, but as a medical director, I’ve seen this when the doctors will come back to me and say, “Hey, my staff hates this. They don’t like it.” It’s like, “Well, what happened?” Well, they tried too fast. They didn’t communicate well that, “No, this is not here to take your job. This is what we’re going to use together to take care of our patients.”
From the tech side, it’s going to take time to correct workflow flaws, and no AI is perfect, nor is it fully customizable. So if you rush the implementation, if you get this new tool and you put it into play, it’s actually going to slow you down. And what’s that going to cause? More friction, more complaints. The staff are going to complain. You’re going to get rid of it and not really see it to its full potential.
Kevin Pho: So I want to talk about support staff or clinical staff not embracing AI even though it does help the physician. So there’s a little bit of a disconnect there, right? If the AI does its job well, chances are it may threaten the jobs of some of the clinical staff, right? So how do you resolve that workflow disconnect when an AI is working, I guess, quote unquote, “too well” from a physician standpoint, but the nursing or administrative staff are hesitant to use it?
Eric Jennings: Yeah. You don’t force it from the top down. You don’t say, “Here’s what we’re doing.” Especially if you’re a practice owner or if you’re running a team, you don’t force it.
If you’ve done your homework and you have an implementation plan, ideally this doesn’t happen. But usually if there’s some resistance, then it’s either clunky and not really helping, in which case maybe it’s not best for you, or there’s this quiet fear of job replacement.
So the way I see it, this is about ownership and involvement. Identify super users. Especially with my technicians as an ophthalmologist, I tell them, “You’re not just here to click buttons. You have to own these patients too. They’re your patients. So don’t just go through the process. I want you to think. You don’t have to make the diagnoses, but what could be going on here?”
And the same thing holds for any AI tool. So get your staff involved. I like to identify a super user, potentially for each portion. There’s the front office super user, there’s the clinical lead, the back end super user. Let’s champion the tools. They’ll be the go-to. So not only does it take some work off of your plate, but it makes them involved and they own the solution, and then you meet about it and hear what’s working and what’s not working. When you involve them as a team, you’re less likely to get a lot of kickback.
Kevin Pho: So it sounds like communication and transparency definitely would be keys in terms of potentially implementing AI tools, just like everything else.
Eric Jennings: Definitely, 100 percent.
Kevin Pho: So most of the conversation is generally focused on a chart, but sometimes on the administrative side, there is more to gain and less to lose. How should physicians evaluate whether or not AI is needed in the front or back office?
Eric Jennings: The clinical note is what we all surround ourselves with, right? In medical school, it’s the SOAP note. How do we take care of our patients? Why are they here? What do we see, and what are we going to do about it? But the reality is, in order to take care of patients, there’s a business around it, whether it’s the hospital or your own private practice.
Given a lot of the administrative burden we’re being put through, I would say that we need AI everywhere, in the front and the back offices. These people are our infrastructure. They’re being asked to do things that are potentially twice their job description and that constantly change.
As an example, take payer LCDs. For those of you who aren’t as familiar, the LCD is a policy created by these private companies that process Medicare claims for a specific region. So they’re rules for getting paid, and they can change on a constant rolling basis. So how are you supposed to find and chase down the rules for what you need to do to get paid? Well, wouldn’t it be nice to have an AI tool or a bot, if you will, that can automatically look at these and check your charts against these rules to make sure you get paid? Because if you don’t, then you can’t take care of patients.
So even though you hear a lot in the news about the AI scribes, and they’re phenomenal, I’m privileged to have built one, the administrative tasks are the low-hanging fruit. And frankly, if you’re dipping your toe into AI, it’s the safest sandbox to build some trust because you still have a lot of oversight into what eventually goes out the door and comes in. So it’s a good place to start.
Kevin Pho: Are you seeing specific administrative tasks have the highest return on investment? You mentioned things like prior authorization, automated coding, AI phone agents. Any specific area of administration that is garnering the fastest operational ROI?
Eric Jennings: I think a lot of those things do live within the EHR, and there are some attempts at connecting to the administrative tasks. But if we’re thinking about completely outside of the clinical experience, prior authorizations, right? Huge ROI, because how much of that is a burden? If you’re listening to this, raise your hand, right? Prior authorizations take time. They take effort, so AI has a huge ROI there.
Claims denial assessments, same concept, where you send out this claim, “Hey, I did my work. I take care of my patient,” for a payer to say, “No, thanks. We’re not going to do that.” So then you have to go track information down and send it back out. It’s all about time and effort.
And the last one I think I’ll mention is patient communication, right? These patients want to talk to you. They have lots of questions. They want to move their appointments. “Why was my surgery canceled?” And I think at the core of it, a lot of the patients still want to talk to a human, but if given the option, if the AI solution can readily reschedule you faster, it’s a really easy way to offload some of that work from your staff.
Kevin Pho: So we talked earlier about the impact AI has on clinical staff, and certainly through transparency we want to get their buy-in. But even though AI may not replace that staff, it is definitely going to impact and perhaps change their job description, right? So how should practice leaders redesign administrative job descriptions in the era of AI?
Eric Jennings: I think there’s going to be a shift, right? We’ve talked about this paradigm shift for the clinician, but to your point, everyone else may have a different experience. So there’s going to be a shift from data entry clerks to something like patient experience coordinators. And we’ve talked about human in the loop; there will be AI auditors.
So going back to the radiology example, that’s a super professional example, but if you have somebody who is doing billing and claims, it’s not that we don’t need you anymore, but the AI is going to categorize and send you the hard ones. So your expertise will be there. So there will be an opportunity to upskill and manage these people and to handle the technology exceptions, but keep that empathetic human interaction that algorithms shouldn’t and frankly can’t touch. They’re also already overloaded with what they’re doing and doing what’s potentially twice their job description.
So I think redesigning it in that way, a good use example here is ophthalmic scribes. We in ophthalmology see anywhere from 40 to 80 to 100 patients a day, so we needed human scribes. Well, now they’re fearful of an AI scribe. I would advocate that it’s not going to replace them. I think maybe if you have three scribes, maybe you only need one, but having a person in the room to help you is still a great idea. But now you have an ambient listening tool that can listen to everything, and your scribe can handle more things for you to make your flow easier. So I think shifting things towards that is really going to make the biggest difference.
Kevin Pho: So we touched upon return on investment. So how exactly should return on investment for these AI tools be measured? Because there is of course the direct revenue management, which is more objective, but there are also more subjective impacts of AI in terms of impacting things like burnout reduction, for instance. So how do you measure return on investment when it comes to these AI tools?
Eric Jennings: The first, the obvious one, is the direct one you alluded to. So, the numbers. When you look at your P&L and you’re weighing a revenue cycle AI tool, well, are my claims being processed? Is my revenue up? But that’s easy and sort of a given.
The metric that you alluded to is potentially quality-based. I’d say the first one, and we’ve had this challenge, myself and others, is to look at the clock. It’s sort of silly, but if an ambient listening tool gets you home for dinner at 6:00 instead of 8:00, the massive reduction in burnout and turnover for your staff, because they don’t like it either, can significantly outweigh a potential marginal boost in your RVUs.
If you’re a practice owner, if tasks are offloaded from some of your staff members, they’re happier, right? They can probably do what they were hired to do, and that translates directly to a positive experience for the patients. Happy staff, happy doctor, happy patients, for the most part. So there’s a hidden ROI, but it’s real, and it can be tracked. If the clinician and the staff and administrators choose to, it can be tracked and should be heavily considered.
Kevin Pho: So I’m going to ask you next to look into your crystal ball. Is there a specific clinical workflow you think will be completely transformed by AI in the foreseeable future, and how should practices prepare for that today?
Eric Jennings: This is one of the more exciting things, and I love that you said crystal ball because that’s really what it is, right? We don’t know what’s going to happen over the next three to five years, but we can guess based on where the AI algorithms and things are going.
And I have a pretty high confidence to say that it’s not going to be a single workflow. It’s not going to be a new tool that you’re going to use. It’s going to be an integrated ecosystem. Right now, what’s becoming a problem for some of you is that a lot of the AI tools are bolted on. So you have this solution and this solution and this solution, and they all plug into all aspects of what you’re doing, and that can’t scale. Some of those companies might go out of business, or it’s not going to work.
So imagine a network of agents. We call these agents, right? Different AI systems that are designed to handle different parts of the workflow. You have the clinical note, you have the billing and claims, your patient outreach and education, but they’re designed to work seamlessly together and simultaneously. So they hand off things to one another in the background. It all makes sense.
You can already see this in other AI systems, software engineering and HR and payroll solutions. But obviously it’s going to take time in medicine because of the clinical nuances and some of the regulatory hurdles. Going back to our first point of why you should be afraid of multiple agents, you know, what’s going on back there? What are you guys doing? You’re supposed to be hesitant, so it’s going to take time. But I have high confidence that that’s where we’re going to go in terms of the crystal ball and how to do it.
Kevin Pho: So just like we talked earlier about physicians who best utilize AI may replace physicians who don’t, the same goes for practices, right? Practices that thrive with AI over the next few years are going to do better than practices that don’t. So what are some of the metrics? What would separate, in your eyes, practices that thrive with AI? What does that look like?
Eric Jennings: Going back to our last point, I think the thriving practices will invest in these sort of integrated platforms. Patchworking things together sounds good, but these point solutions don’t talk to one another. So looking towards the future, it’s not going to be a nice-to-have. We call it interoperability. Things that play well with one another are going to be required. So I think those that thrive are going to look towards that and vet things heavily in terms of, well, how well does this play with everything that I have and everything that I want to do?
And also, the ones that thrive are going to educate themselves now. Again, alluding to our earlier conversation about paralysis versus skepticism, it’s OK to be skeptical. You should be. But if you’re going to thrive, be skeptical and move forward. Whether or not you initiate an AI tool, at least be interested, at least dip your toe in and educate yourself. If you’re going to be completely hesitant and resistant and not adopt anything, you will likely be left behind.
Kevin Pho: You’re listening to a special sponsored episode of The Podcast by KevinMD. We’re talking to Eric Jennings, a board-certified, fellowship-trained ophthalmologist and the senior medical director of ophthalmology at ModMed. Find out more at ModMed.com. That’s M-O-D-M-E-D dot com. The link will be in the show notes. Eric, thank you so much for joining us. My last question is always to ask you your take-home messages to the KevinMD audience.
Eric Jennings: First of all, once again, thanks for having me. Don’t be afraid of AI. Don’t be afraid of it. Critically vet it and validate it. AI is an algorithm, but just like anything new you learn, whether it’s a new drug, a procedure, or a preferred practice pattern you hear about at your academy meetings, appraise it critically. Figure out what it’s all about. Critique the evidence and do a pilot. Safely figure out how it works for you.
The analogy I use when going into AI is that you don’t have to start off at the full surgery. For those of us who do procedures, or everybody did throughout training, you don’t just jump into the surgery. Learn to throw the stitch. Throw the stitch, tie the knot, and then keep building from there.
I would also say be patient with AI, right? It’s still a computer program. There is something called hallucination, which is fascinating if you read into it, so be patient with these solutions and don’t expect perfection, just like you still go to CME and try to perfect your craft over time.
And I’ll end with what you brought up, Kevin. I don’t think AI is going to completely replace physicians in the way people think. The doctors who use AI will likely replace those who totally refuse to adopt.
Kevin Pho: Eric, thank you so much for sharing your perspective and insight. Thanks again for coming on the show.
Eric Jennings: Thank you so much.
Listen and followApple PodcastsSpotifyYouTubeMore apps · Browse all episodes
