2. How AI Improves Clinical Insight
The panel discusses how AI can unify fragmented patient data, reduce cognitive burden, and support better clinical decisions. They emphasize chart summarization and surfacing information clinicians might otherwise miss.
Oh, great meeting you all.I'm not a tech guy, so I'm going to try to moderate this as good as possible.But one of the first things that maybe opening up the conversation, you know, how do you how do you think AI can help us with getting better insights, better coordination and quality of care?
Oh wow.So that's a loaded question, but I think part of it number one is giving the clinician and the physicians access to what's going on with the patient because a lot of times the basics of what you need toknow from what the patient is kept telling you and the imagery and then the EMR and the EHR, all of it sits in different data warehouses or data sets.And so AI has the capability of being able to really source all that information bringing together and giving you kind of a one snapshot to the to the client.And having that ability to do that, I think gives the physician the time that's needed to spend with the patient, but it also gives you the ability to actually really discern what's really happening.So having that one view of everything in the same place and not having to look in four or five different systems to determine that, I think that's one of the things that AI can help and assist with.
I think from a data and analytics perspective, prior to AI, we make decisions off of data, right?Instead, with AI, we now have the ability to make decisions at point of time while in workflow, while we're seeing eye-to-eye with patient, if not hand-to-hand with patient from a clinician perspective.And so as we think about the cognitive burden, if not the overload of for physicians or clinical care teams, is it allows us to have greater augmentation, allow that cognitive load to be put elsewhere, and it just givesus more time for in-person care, which us as humans is we desire that on the day-to-day.And as we think about other elements around clinical documentation, if not the ability to summarize an overall visit, if not a patient encounter, as that data gets collected and these systems become more integrated, it also allows us tomake next best recommendation for clinical care.
It allows us to give more power back to the to our patients on knowing what that next step is, but also on the flip side, it allows us to understand what is that next best action that we cantake within side of actual workflow.
Yeah.I think on one hand, AI has changed a lot of the way that we develop software, the way we work together with with health systems.Epic is the patient's story, essentially their health record.And something that hasn't changed is our overall goals, which are to help clinicians love their jobs, to make doctors feel like they're a good doctor and they don't have to spend all this time on administrative work or clicksor scrolling, and also to help healthcare organizations be successful, to streamline things on the revenue cycle like appeals and and processing denials and coding and things like that.And then finally to help patients be healthy.So it's provided a new set of tools in service of the goals that really have been sustained in healthcare.
I think if so I'm a psychologist by training and so when I just think about the complexity of the human condition and the human body, right?There's only so much that a human can actually process all at once, right?They can hold seven working thing things in their working memory at one time.Now the human body and what happens to the human there's so much so many more things than seven things that you can process, right?And so that's where I think AI really plays a vital role in terms of the augmentation of being able to hold hundreds or thousands of things all in constant at at the same time to really help to understandand the physician to understand what needs to happen, right?So so I think that's where AI really plays a vital role in sort of augmenting care is expanding their possibility to really understand what's happening with the patient cuz it can understand all of those different inputs.
You know I I know there's a lot of clinicians in the room and they know this very very well.When we went to the EMR the notes the clinical notes went from one page could read and understand what's going on with the patient to now by the time you're discharged from the hospital I mean it's 25 pagesand a lot of data in there that frankly is it like I'm not sure that we were able to process.It just pulled into the chart and because it's available.
It's true.There's actually a paper written recently called Call Me Dr.Ishmael and it says that the average size of a patient's chart who comes into the emergency department is as long as Moby Dick.So it's literally impossible to read the full chart when a physician is seeing a patient for the first time.So one of the things in Epic is a summary of a patient's chart that shows the clinician, here's all the pertinent information for your specialty that you need to know.And what we find when we survey the clinicians is that they usually it usually surfaces something they would not have found in the time that they have.It's like 80% of the time they're finding new information that changed how they were caring for the patient.
3. Barriers to Healthcare AI Adoption
The conversation shifts to the main obstacles to implementation: data silos, workflow fit, trust, and change management. The speakers stress that AI must solve real problems and fit existing clinical processes.
So in your experiences, I mean the the positives, the why, the potential indications and many, what has been the barriers?
Oh well, so my team covers payer, provider, medtech, and pharma, right?And so uh one of the things that always comes up is that, you know, why is the ecosystem of healthcare for the patient so distributed to where it's not intersecting together?And I find for me at least four key things that always come up.Number one is data.It's like the data is sitting in multiple places and you have both structured and unstructured data that's in there.So uh the best way to think of that is part of it's easy to read, part of it's not easy to read.So how do I actually get the value out of what's in that unstructured data?Uh I think the second part of it is is workflow paradigm.There's a workflow inside of every health system that already exists.I mean I'm in multiple systems throughout the year and there's still yellow sticky pads and sticky notes up in almost every cubicle and every desk you walk through.
And if it's there, that means that I can't just give you an AI tool and don't take into account, wait a minute, that Rev Cycle team has 22 yellow sticky notes, the Claims team has another 10, uh theScheduling Prior Auth team has another 30 because regardless if it's written down, it's part of the workflow.And if you don't take that into account, then you're going to miss it.So you put a tool in place that never gets adopted.And so because the lack of adoption, I think is is the workflow.The third part I think is trust.Uh the patient trusting that this is good for me, getting the new team to be able to trust what the tool is going to do because there's such a fear out there that, you know, AI is going to,you know, come alive because we saw the the movie AI Robot and we're thinking that, you know, it's going to come in and take over things.
But the reality is it's supposed to it should be augmenting, it should be assessing and helping provide insight to the actual caregiver so that they can do things better.The last thing that I see most enterprises fail with is change management.The intent around AI is not replacement, it's to make things better.And I use this example and it seems kind of extreme, but if I'm going to take a trip from New York to California, I can still take the Wells Fargo stage coach.I still have two stage coach drivers.Or I can take a a Boeing airplane and I still have two pilots.Two people still leading and driving and getting me to the destination, their role has changed.AI is going to change the role and the way that people were functioning, enabling the support to the clinician to still be able to provide care to the patient, which I think is the ultimate goal.
We're trying to make the administrative tasks better.And I think if we can get through those four barriers, then the implementation and the rollout across the enterprise becomes a lot easier.
I think one of the things to realize with where we are in history, if we think back whenever the computer hit modern America, it took us 18 18 years to adopt the computer.It took us nine years to adopt the internet.Took us six years to adopt the iPhone.Well, since 2023 is whenever ChatGPT hit the modern street, right?And we're still very new.2% of the average or 2% of all households in the United States for pay for an AI oriented subscription.So we're still very early in this journey.And I think like the ability to scale AI comes with an element of discipline, but also an element of what is the value that we're doing it?But what has to come along with that is AI fluency, which is oftentimes in academic medicine back in the I mean like 2010s, we always used to talk about data fluency.
How do we read, write, argue and decision make and make decisions off of data or metrics or analytics.But now we have to think about even when our schools and our communities as do we drive AI adoption and fluency for consumers or even clinicians or back-office staff?And that's going to take some time on how do we read, write, argue, and decision-make with data and now with AI, which is now putting even more data in front of the clinician than ever before.And And that just takes time to scale.I think so the ability of being disciplined, being able to create that ability for change capacity for our clinical care teams and our physicians, and allowing that to scale with the right discipline, but also the right governance andmonitoring put in place because fundamentally these tools are not humans.They are It is a digital tool.
It does not feel a cognitive thing uh unless if we tell it to.And so it's being able to have that right monitoring in place is so so important.
Yeah, in my experience things continue to get harder basically as time goes on.You know, the amount of treatments that are available, the amount of personalized treatments, and from a clinician's perspective they're also trying to manage helping patients navigate their insurance and getting transportation and finding an appointment and and it's likethe burden just centers on the single people.And handoffs in healthcare, the number of people involved in these processes like authorizations and scheduling and follow-up and getting the medications from the pharmacy.Now there are studies that say reason patients don't take medications, the most common reason is they can't afford it.And so I I think one of our jobs is to make solutions that clearly solve a problem.And what I've found is if you if you give someone a technology that solves a big problem they have they will adopt it.You don't really need change management.
It We've seen that with AI charting, which if any of you have gone to the doctor and your doctor said can we Can I use my phone to record this visit and it will do the charting for me?That has has been taken up quicker than any change in the last 20 years as far as healthcare technology is concerned.So I think change management certainly is a challenge and the clinicians who already have very full days, no extra time, how can you do something that may be more painful short-term?You know, maybe you can see one less patient because you have to learn something new, but long-term it's going to help you see more people, have better efficacy of visits because you don't have to spend so much timeon all the typing and administrative part of it.
That's exactly what we see in our customer base as well as like if you're not painting the vision or the reason for that change, they're not going to take it on, right?And so you really have to create that those conditions where it's not forced upon them, but it's like this actually will help me do what I actually want to do better.And it might take some time.It might take a tradeoff in terms of like re-altering these processes, but ultimately it frees up that capacity for me to do the thing that I actually want to do as opposed to the other things that I don'twant to do, but I need to do, right?So there are always these administrative tasks that you need to do, but nobody wants to do.And so then that question is like, hey, is that a great place for AI to come in to play because it's something that needs to happen, nobody wants to do it.
It's pretty standardized hopefully in a way.And now you can have the AI go about doing it.And so you just have to pose it in in that sort of way to get people to want to buy in.
4. Building Trust Across Stakeholders
The panel explores how to earn trust from patients, clinicians, and staff with different levels of comfort and fluency. They discuss governance, outcome-focused messaging, and examples of AI creating value without replacing people.
Follow up on on something you said, Tony.You said trust and and trust is interesting.And my impression that there are so many stakeholders and the speed of trust of these stakeholders is very variable.
And so how do you think about approaching such a dynamic range if you want of tech enablement, tech understanding you mentioned earlier, the lingo, the comfort with the technology.You have patients that are 18 20 years old and age is not an issue, but just that a generational thing where AI first, right?Like before I go to the doctor, I would go in there and see my symptoms and then I'm going to show up at the doctor with and then you have the other extreme where I want the human touchand then that's on the patient side before you get to the nurses, the doctors, the other clinicians.How how do you design something when you have so many different stakeholders with very very different
Um I think there's two parts.Number one is that you need to have a minimum governance and an assurance of this.Your team is using AI whether you know it or not.So if you don't have governance in place, that means that they're using something that's not being regulated by the enterprise.Almost every enterprise implementation that I've done, I've never found that there was I mean you have people that's copying and pasting out of you know a secure hip environment and they're putting it over into an online chat toolthat they found that was free and so now all of a sudden you have a patient information that's going back and forth in an unsecure environment.So number one, you want to have the right kind of governance in place in the enterprise that that does it.Number two is exposing the teams to the tool and the benefit and the outcome.
I really do believe that you cannot introduce AI as a technology.You have to introduce it as an outcome for the business.This is what's going to be better for this particular group.This is what's going to be better for this particular organization or this is going to be uh the financial impact for the CFO or and I give you a natural example.So I met recently with a West Coast CEO of a large hospital and his chief of staff had applied for a role.And he was like, she's going to help me prepare for a board meeting very similar to this large group.And I said, well, how long does it normally take her to prepare for your board meeting?He's like, well, six weeks.She's got to pull all this paperwork.She's got to put presentations together.She's got to align speakers.She has to do all these different things.
And we use one of our AI tools to help him.And all of a sudden now, he's using Copilot as his AI chief of staff.Now, it didn't replace anything.The person got promoted to a higher level role.They're doing something now to advance their career.And now the role that he was going to post for his chief of staff, he was able to post for a new nurse because that's what he really needed in the hospital.And AI is now taking over an administrative task that is no longer needed.And now they redirect redirected funds to a place to where it really needed, which is care for the patient.That now told the team, hey, we can trust this tool better because they saw it being utilized in a way that provided value to the overall system as a whole.So that's just kind of one example.
I think from a trust perspective as from Bon Secours Mercy Health aspect is we started with our workforce in mind.And as you think about each one of your own consumer journey, we all have different levels of trust, but a different level of fluency and comfort, right?Most recently, uh, Meta launched what was known as Muse.And the idea of Muse is you can have your own consumer personalized agents to go do what you desire.Um, now use it cautiously because it may order a dog bowl that shows up on your door because you're complaining about your current dog's water bowl and getting water on the floor.True story for me at least.Um, but as you think about an, uh, an enterprise workforces, we've taken more of a persona-driven based pathway of to where we know that we're going to have a patient AI journey.
We know that we're going to have a clinical AI journey.But as you start getting into the clinical staff, the way that we provide care is so different from a neuro perspective in comparison to an ortho element.And so as you standardize and you imagine the art of the possible with inside the clinical workflow is you have to begin to start understanding the roles of where AI can be applied and where it doesn't need tobe applied.And being very methodical around on how we communicate, how we think about and on how we utilize AI is AI doesn't need to be applied to everything.Um but it's something that you have to build trust with over a course of time and uh you can't just full on throw it at I mean throw at and hoping people use it.
Muse is an interesting example.I was all in on Muse until last night I was reading about all the privacy breaches.So talk about trust.I took away access to all of my personal information and I have a saying that I try to remember all the time which is don't piss off your users.And I think that's a big mistake that we make a lot of times in AI.I was in a meeting a couple weeks ago and I'm I'm probably the most AI forward experimenter at our company.And I was in a meeting with someone and they said don't worry, we're going to upskill you.And I thought isn't that what everyone always wanted is to be upskilled?So I I think it's also about how you talk to people, how you value the um just the humanity that goes into especially healthcare.It's like we we're irrational as humans.
We have hope, we have belief, we support each other and all these things are not that are not really encoded into the computer or into technology.So you know one of the things that I I really hope for and that that we try to do is to remember that saying don't piss off your customers because that's really critical.
Yeah ultimately trust is earned and it's earned over time.And I think it's earned with the interaction that you have with uh with the people either delivering the product or using the product, right?And so uh you can't take trust as a given.You have to actively every single day earn it.Um and if you're delivering a product in the same way you have to actively every single day earn that trust of that end user uh with how well you deliver the outcomes for
5. Enterprise Strategy and Vendor Consolidation
The discussion turns to market fragmentation, enterprise vision, and how health systems should choose AI investments. The speakers explain why many organizations prefer a fast-follower approach and want AI embedded in familiar platforms.
Yeah, I mean, it's fascinating the amount of um solutions that are coming to the market to maybe put a little bit of a more of a system hat for a second.Um there's so many coming in.
And the market has not consolidated yet.Do you foresee that?Is that going to happen at some point?And then what's your advice for healthcare systems like Roper and like us?Do you Do you invest in the whatever is available today and then consolidate later?Do you wait to consolidate?
Yeah.Um so number one, I I really do believe that you have to have a real true enterprise vision around this, right?Be it if you are looking six months out to uh a three horizon two three year out.I mean, both the panels along with, you know, obviously you all being a client of ours, uh Epic and Workday are very strong partners with Microsoft.We work with them day in and day out.Uh we have dedicated teams that actually work specifically in that to try to help integrate some of that, right?Um so that's number one.Number two is it's a cost issue, right?It's like, you know, who bears the cost?We were talking about this actually right before we came up is that, you know, who bears the cost burden of making sure this is fully integrated?Some of it's probably going to be legislated.
Some of it's going to probably be require certain partnerships in place to be able to do those things.I don't know that there's a straight line to it, I guess is what I'm saying.Uh I think it's probably more of a arc or a curve to make sure that number one, do I have the foundational piece in place to make sure I'm not doing any harm to patients and I'm not addingadditional workloads to the the the caregivers that are out there, right?The physicians and the nurses.So I think we have to take that.Uh I have this thing whenever I'm helping a client roll out that I call the heart of AI.Number one, is it human led?So in other words, everything that I'm putting in place, I want to make sure that that AI is based off of being human led.
Number two, is it embedded into the current workflow?No physician wants to come out of a screen.If If I'm inside of Epic, I got to come out of that and go into another tool, you're not going to get me to use it.It's just not going to happen to to be able to do the things.Number three, is there somebody truly accountable for it, right?Like, do I have if somebody needs to go turn that off, is there a button that somebody can go hit a kill switch uh and make sure that actually happens, right?Uh the fourth thing is is there a real true ROI back to the business?Are we doing it just because it's a nice new shiny object or is it really returning value?And the value that in my mind is am I saving patients' lives?Am I improving the value of what we're doing?
Am I giving time back to the physicians and what they're doing?We have three specific rollouts that we've done here in the last 12 months that, you know, one large research clinic that if I named you all would all know that's based out of Ohio, you know, that says thatuh with the same letter that's in both of their names, uh they've come back to us and specifically said to us, hey, leveraging this tool, we've saved 52 lives in the last six months.Like, they measured the value of that AI by the lives of patients.And so I think that's one of the main things.So is it really real?And then the last thing goes back to trust again, right?Is that do I have a trusted system?But let's answer the first part of your question.That trust is not do I just trust the AI?
Do I trust the foundational security that's in place?Is it going to violate any HIPAA rules?Is it going to be PHI in a place to where it can't operate?It's going to put financial risk to the company.Is it going to put a marketing risk to where the wrong type of information gets out and all of a sudden now you having to spend money to recover because the wrong thing was actually put out by AI?So it's a full circle arc that in my mind more than it is a straight line.But I think if you have the right mindset of why you're building it, the outcome you're trying to get to, I think it helps you get there.Hopefully.I know that's a long answer, but hopefully that helps.
I think we will see consolidation over time and I think we're starting to see an interesting place in time right now between some of the major competitors, if not major technology.For example, whenever Meta launched Muse and then we talked about a little bit is they um Amazon or AWS ended up blocking uh Meta agents to actually go make purchases on Amazon's marketplace.Took a week for this to happen.And so you would think that people want agents to go make purchases on behalf of consumers in order to drive up sales.Instead, what we're starting to see is we're starting to see technology players put um what is known as a moat, if not a fence up between their technology being able to interact with one another from an AI perspective.And so I do think that we see consolidation um from a Mercy Health perspective is um we we would love to be cutting edge or bleeding edge, but in reality, we're okay with just being a fast follower, right?
Because we there's another healthcare organization in the country that went very early on in 2023 and said among the open AI is we're doing X, Y, and Z, and it was on every single publication, and there's amazing stories.Now they're starting to consolidate from their early on investments back to some of their core strategic vendors with Epic and with Microsoft and with Workday.And for us is the the word enterprise experience is what really sticks out for us is we want to extend AI and experiences that our workforce are used to.We want to keep our knowledge workers, which are those individuals that are more strategic in nature.We want to extend AI and Microsoft because that's the ecosystem that they're used to.We want to keep our clinicians within side of Epic because that is where they are.Um now we may have certain certain ancillary services, if not other systems that we may extend AI, but we want to use our core investments to where we can extend AI very similar down how we do benefits.
Who worries about benefits day and day out other than your benefits team?Only whenever it comes to open enrollment.For AI, we want to be able to be able to extend AI to where the organization is used to.
6. Patient Access and Democratizing Care
The panel examines how AI can improve access, reduce friction, and help patients navigate care more easily. They also discuss rural access, social determinants, and the risk of creating a two-tier system.
Yeah, as an organization, I think you have to know yourself, and you just exemplified that very well.Many health systems I work with say they want to be on the cutting edge.That's not easy.You have to manage all these different inputs.You're doing a lot of experiments.So most organizations really want to be what I would call a fast follower.Do the things that have been really proven out, that have shown benefits.For example, a simple example is um how many of you use MyChart?That's a patient portal.That's where you can go to see your appointments and message your doctor.Okay, quite a few of you.I do, too.And my son was sick this week.He ends up he has pneumonia, but we were up in the night and I was worried about him and I I sent a message to my care team and it was the middle of the night, so they weren't
there.And I had a question about where I should go.And we've developed an AI called Emmy.That's the persona of the AI that patients can put a question into at any any time of the day, any day of the week.And they can get an answer that's private and secure and that's that's supervised by the health system.So it follows the policies and the procedures of like if you used it of Roper St.Francis, it's not just a Google search that, you know, could lead to just tells you to go to the ED.It tells you you can call the nurse triage line anytime of the day.And for patients, a lot of times this happens in the middle of the night, it's a crisis.So we did something simple.We started with billing questions.We didn't start with trying to triage patients to the emergency department or not.
We tried by answering questions, does my insurance cover this?How can I appeal this decision?And what we found is that cut down on the number of phone calls and messages that were sent to the billing office by half because patients can answer those themselves.And so I I think you can start in a place that is lower risk and still high value and you can build on that.And as an organization, you can follow what others are doing and you can see, you know, across all the health systems that are using AI for coding, they're seeing a reduction in denials because AI can make sure thatthe codes are correct.And so, you know, that's an area the the revenue cycle is an area where you really should be doing a lot.You should be taking advantage that's proven out.Other areas that are newer, maybe higher risk, you not only consider what's the risk of doing this.
That's usually what people are thinking about.Oh, if we make an AI triage, the risk is the AI could triage the patient incorrectly.What what I sometimes see is that you're not thinking about what is the opportunity.What about all the patients that could never access this triage without this tool?What if we could screen a million more patients for cancer?What if we could connect a hundred more patients with the food bank?You know, those those types of assessments I think are really important.What is the risk if we don't do this?What what opportunity are we missing?I talked to a member of the compliance team at one of our health systems.I'm not a lawyer.I actually got into healthcare because I worked in my community hospital's pharmacy and I thought doctors are so cool.I just knew I wanted to work in healthcare.
And so anyway, the compliance person says, yeah, compliance, it's not like you're compliant or not.It's actually how close to the edge of the line do you want to be.It's not just a one or a zero, but it's how forward thinking, how how much risk do you want to take to try something new?And for AI, one of the strategies I think that has helped in this case because you might be doing something new, there's not a playbook for this.What what procedures are we going to put in place to make sure that it is safe?And how are we going to assess these types of things?So we've created new ways of evaluating, new strategies of engineering these, of having separate AI that can supervise the other AI.So there are lots of ways to make sure translations and other example.If I get something translated to French and I don't speak French, how can I check that?
And this is an area where we need to continue to develop tools and this can also help to to build trust if we can say, you know, here's what we checked and we checked for all these things andthey all checked out, and we show our work just like we do in other applications where you're using tools, but you don't know how they work.You don't I don't know.Maybe some of you know how Google Maps works, but you probably trust it to get you from point A to point B.
Yeah, that's uh what what I would say is, you know, AI is moving so quickly, and if you don't jump in now, you're not developing the practices and the muscle to adjust to the change as well, right?And so if you're just waiting for, you know, it to be ready, it will never be ready to the point where you feel ready to adopt it.So you have to just go in eyes wide open with great partners to help you along the way um and to help you, you know, navigate what's happening as well, right?Because you're never going to feel ready as an organization.You're just going to have to rely on your partners to help you um with that as well.
I think one of the things about having great partners is this technology changes day in day out.Things that were relevant two weeks ago are sometimes relevant still.Sometimes we've beyond it.We're in a new era.We're in a new hemisphere.And I think with having great partners as such is it helps our our health system be able to be positioned to let the the bumps and the bruises go on through and also assess where we have seen pilotshave success and on how do we bring that back here to our own community health systems and then apply that success here, but be very niche as far as we will have our own internal use cases that arespecific to Roper, specific to broader BSMH to where we can apply.
Um Lydia, you scared me a little bit.You said AI supervising an AI supervising an AI, and I'm like, where we're going with that?Um but then you said we're never going to be ready, so I'm like, okay, well, maybe I have to accept a little fear.Uh, it's just going to be part of our journey, and then there's something that's going to be unknown.You know, one thing uh we talk about a lot is is is consumer centricity.I I uh uh I'm uh for the four five years ago I moved to DC and um until then I always lived in an area where I knew the docs, I practiced, I was able to access healthcare andmost of us who have the privilege of doing so get uh a very easy experience.You pick up the phone, you call somebody, it's like, hey, I need to see you tomorrow, can you see you today?
And typically you get in.And then when I did try to get a primary care, it took me 18 months.
And uh I had to call a board member that I knew from a prior work event to get me with his primary care I mean, it was not an easy.And then once you get into the system, it's not easy.
How is AI helping navigate uh the consumer journey and and how and we're helping uh you know, take that uh experience that some of us have been privileged because of access in a different way and and almost makeit everyone's experience.
You know, um and I share a personal experience.I grew up with most people would call a health desert.I'm originally from Mississippi, uh very small town, you know, you maybe 5,000 if you count the dogs, the cat, and the one giraffe.I mean, like that includes the one stop sign.And my father, who is very reluctant to healthcare, uh I'm able to leverage AI a lot of time to kind of keep track of certain things that's going on with his health records and keep up with it andI live about 8 hours from him, my sister is a little closer.But when I was a senior in high school, my mom went to her regular uh doctor care like she was supposed to, he sent her to a specialist, did it, and then she goes back to the same specialistabout 7 8 months later, and he asked you a question.
He says, oh, by the way, how's your cancer treatment going on?And she's like, what cancer treatment?He's like, well, the last time you saw me, uh, I told the doctor you had cancer.And come to find out, he it wasn't sent over in the original charts that he sent back to the primary care doctor.It was later caught and then it was faxed, but it was never taken off the fax machine and put into my wife's record.So my wife found out a diagnosis, my wife my mother found out a diagnosis almost seven and a half months later.My mother died my senior year of high school because it was something as simple as communication from one provider to another one.AI today fixed that.Even if that same scenario happened today, in a record between those two doctors, that record would have been updated instantly via AI.
And what now took seven months that cost a life and no longer a part of our would have been taught and and honestly, my mom's next primary care visit, it would have been updated.It would have already scheduled that visit.It would have already had those record and notes.AI has changed that completely.And in my mind, that's one of the reasons why I do what I do is because I don't want another senior in high school to experience what I went through.That's why I leverage AI to keep track of what's going on with my father as much as I do because I I'm not two doors down from it.And in my mind, AI is now giving me back time that I lost because it wasn't available to me in that time period when it was there.
I well, I was thinking one of the things AI allows us to do, you I you said I scared you about the AI AI AI, but that is something we think about.And there are some cases where letting AI experiment and research is really valuable.We developed a model.It kind of works like if you've used ChatGPT, ChatGPT predicts the next word like may the force be with you.And similar to that, our model uses a patient's record that we trained.It's deidentified to predict events, medical events, like the patient will be admitted to the hospital or they'll get a stroke or they will be diagnosed with cancer and we can use that that simulation to predict the likelihood ofthings.And that can be useful if you could change the outcome.If we could predict something sooner, then can we treat the patient?That could do things like cut down on this diagnostic odyssey, which it could be a misdiagnosis, it could be how long does it take to get the right diagnosis for the patient to get them a good treatment.
And we're starting to see health systems use AI to also keep an eye on patients and make sure that if a step didn't happen but should have, that we nudge that.And for example, if there's a a scan and a lung nodule gets detected on the scan and we'll probably hear more about this in the next panel, that AI can keep track of that and make sure that thatdoesn't get missed, that gets followed up on.The result is that at the health systems that are using this, early stage cancer more early stage cancer is being detected.So the rate of detecting cancer at the earlier stages can hopefully result in better outcomes for those patients.And I think we can also use this type of prediction to do things like predict more near-term risks, which we were talking about earlier.If you knew you were going to have a stroke in one year versus 10 years and I said all you have to do is quit smoking, how much more likely are you to make that change if you knew
it was going to happen really soon?So lots of different ways we can start to apply that, but we can have AI researchers that are looking for these insights that are connecting clinicians who treat a patient that they've never seen before with another clinician whohas treated a patient like that and make them aware of clinical trials or treatments that are not the standard of practice yet and and to help share information really across the whole world.
Yeah, I just I think about what the patient and uh physician relationship is.But also when I'm I looked at those posters over there where you're expanding uh your reach, that also requires bringing in new employees, right?And so like it's not just about the the care practice, but also the the whole underlying operations of your system as well.And so AI can play a really great role in helping in those operational elements, whether it's um getting nurse nursing hiring really quickly, right?You know, um getting um administrative support to do certain things.Uh you know, really building out that infrastructure uh for administrative excellence or operational excellence.All of that is really important in order for your health system to actually deliver the outcomes that you're looking for from a patient perspective.So it's not just AI in the realm of patients um uh and physician you know, care.
It's also AI in terms of overall operations of your of your health system, right?And so that's really important to also consider is it's not just that it's it's how AI can really help um augment the the entire system.
I think we've seen the overall business model change in the realm of healthcare just a little bit.And we see this a little bit with clinical AI assistance of like an UpToDate or an OpenEvidence or so forth is no longer business to business.It's almost like business to consumer or business to doc, right?Now as we think about business to consumer, they now we have personalized agents to consider and what what role do they play?And these agents are personalized to an individual.The more and more information that they find out about us as consumers, the more and more recommendations that they make.Like if you think about a chat GPT for health, if not a patient GPT from like a K Health or something along those lines, that allows me to upload my own personal records, whether that come from a GarminCirca or a Whoop.
Um um I then combine that with my own personal health data from my chart that stays with me and it knows all these other consumer things about me of my likes and my dislikes and so forth and Ithink as we think about more access more broadly is the more and more context which is the data that you feed into it that the agent will get to know about you or an a co-pilot like engine willget to know about you will really help determine what your way of access to seeing care is and making that much more personalized.I think as health systems as we look forward is our ability to have more digital presence and more and more digital presidents whether is that be in just generative search if not having more than just beyond the typicalwebsite and the more and more context that we show up I think is only going to help better our ability to provide better and easier access to care and it'll become more personalized as time continues on.
7. Ethics, Safety, and Human Oversight
The speakers debate the ethical risks of AI, including privacy, security, and overreliance on automated systems. They emphasize human accountability, guardrails, and the need to preserve human judgment and learning.
As I as I hear some of the comments and conversation I can't help but believe that at some point we're going to end up in that two-tier system like some people are going to have access to all thisand some people are not.How are you approaching that?How are you approaching democratizing and and ensuring that we don't end up in a place where only a subset of the Lydia your example right you had access to that AI and and maybe companies out there are goingto invest in that and some companies are going to have that and then there's going to be a whole lot of people that won't have that.How have you approached democratizing it and ensuring that we don't end up in a two-tier system?
Yeah um there are a couple of things that we're doing and by the way it is very different depending on uh the government in the country in which you operate in so be very clear I mean even forme I cover from Canada all the way down through southern part of Latin America and when we talk about the sovereignty around data and how we exchange and move data around from state to state it's okay for myphysician in Dallas to send my records to a physician in Tennessee you can't do that in Brazil right you can't do that in other nations so I think it's very different.Number two, at least from Microsoft, we're trying to make sure we partner with groups that actually look at the overall social determinant of health of patients holistically.You know, we're working with CMS right now and we are the base platform for them to try to make sure that there's an interoperability intersect that where this data can be exchanged not only for those who can commercially
afford it, but also that gets integrated into our government funded side of the house as well, which it didn't capture 100% of our population, but it at least captures the majority.The same thing when we start looking at DQMs, right?That's going to impact almost every provider that's sitting in this room right now.We're actually building that model with CMS and that is kind of part of an investment that we're doing because we're trying to make sure that we do that very thing that you're talking about, right?That we don't create a system that says the haves and the have nots.We want to make sure that it's available to everyone that we can and then we partner with partners that are sitting on this, you know, panel with me today to try to make sure that it's available in thethe most effective way that's possible.
I think providing access to AI, it's a real risk, right?Go back to that stat.2% of US households have pay for a AI subscription.How many of us want another subscription subscribe to?But as you think about just those general like generally available co-pilot like solutions, you have OpenAI ChatGPT, you have Grok, you have Claude from Anthropic, you have Gemini from Google, now Meta Muse and so forth.Um is it always going to be that way to where organizations if not I mean organizations are willing to pay for the cost of AI?The more and more that this gets used, it provides more context and data back to these large organizations.Is it always going to be like that?I don't know that it is, but only time will tell.But once again, we're we're very much in our changing time.I think one thing that I would always highly recommend for organizations to realize is us as tech lovers and technicians or technologists is sometimes we have to take a step outside of our own shoes and realize what it
is like um living in a new community uh that may not have the accessibility to funds to be to pay for some of these things, if not even to a computer.And that's a different look for some organizations if you live and breed this stuff day in day out.
We focus on partnering with FQHCs, which are federally qualified health centers.They tend to treat Medicaid Medicaid, Medicare, uninsured patients.And so we've worked really hard to extend Epic to them.We've also partnered on many of the rural transformation programs across the country to make sure that people have access to this technology.I also think that if we use AI to decrease the demand on the system overall, so that patients can resolve things themselves, you know, imagine if you didn't have to go into the doctor and you knew you didn'thave to and you could just take some Tylenol or something, that can prevent a lot of visits that are clogging up the system today.And if we can match patients better with the specialists they need, a lot of times I hear specialists complain, I just saw these patients.I can't help them.They have to go to neurosurgery, not oncology, for example.
And And if AI can do a better job of matching patients with the right resources, we can use AI to balance schedules in infusion center, for example, so that patients don't have to get you know, you don't getwaiting and sitting in the waiting room for a long time.So we use AI and other techniques, you don't need AI for everything, to manage the resources you have, make sure that you're making the most of those specialized resources, reduce the demand.Now you open up access for
all kinds of people and services that you couldn't do before, like more care coordination, more community outreach, partnering with community programs.
I'm going to be the downer here.And umyou know, all of this stuff costs money to be to be frank, right?And so to deliver something to people for free, right?What's the saying?If it's for free, you're the product.And so, I think that's that's something that we're all going to have to think about and and really consider when it comes to providing these services as well, right?To To a consumer base that might not have the resources to have a to isolate their data or or do something like that cuz they're going to have to be sending their information to these public services that willdo something with that data if they're providing something for free ultimately.And that's where the trust element comes in, right?Is like you have to build these trusted services in a way that can be delivered to the community community so that they can access it.
And it it's not Workday's place to do that, but it's also, I think, an overall community decision and community conversation that we all have to have in terms of who gets access uh to this and how we deliverthat access in a way that people trust and the way that we can deliver the the outcomes that we want.
But I do think that is why we need a safety net.We need mission-driven organizations who are going to keep this as a priority.I I totally agree with you that this costs money and we have to pay for it, but also we pay for shelters, we pay for all sorts of public services and what kind of place do we want tolive in?I mean, we you already provide a lot of charity care, you already do some of this.So I think also it's partially intention and willingness and when I see people make road maps in the AI world, we make road maps, it's like a list of all the things you're going to do.You kind of saw that in the marketing presentation and sometimes I meet with people and they say, we only want to do the ones that are going to make us more money or that are going to save money.
And that's important.You have to stay in business so that you can support your mission, but you also have to do the things that are going to reach more patients, that are going to fund things that might be loss leaders.But, you know, at the end of the day, I I do think that you can make it make sense and we're optimizing the techniques.So it's important to consider and I still think there's a way to do it.
Yeah, and I'll add one thing from Microsoft anyway.We are investing in communities that don't have access, right?We are actually putting technology into community centers.We're actually giving people access to AI.Along with that, we have a whole team that's being led by our medical director around the Rural Health Initiative with the United States government.So Microsoft is actually participating in those initiatives right now.One, to provide the technology.Two, how do we give access to it?And it's something that my team and I support, you know, because I know my having family living in that healthcare desert still today, I want to make sure that that's something available to them.So I think there are partners out there that are doing their part, that's investing in and I think that the question becomes is how do we unify the ecosystem so that the access can be utilized greater than what
And I think back to Lydia's point, I think you made a really, really good point is you have organizations that want to make hard ROI from a finance perspective, but can we blend that hard ROI in order tobe able to pay for some of the soft ROI?Soft ROI being improvement on the associates' lives, um improvement in the patients' lives, or whatever the use case may be, or if ROI may be.And as you think about just an overall health system, is per every single business unit, per every single service line, whether it be on the foundation side or on the marketing side or even finance, there's a digital transformationthat each one of those leaders will have on how we bring the workforce along with.But how do we extend the mission of healthcare, to extend healthcare to the broader community in those areas that may not seek access to care?
In some communities that we serve, we're the only We are the only hospital in town.And we we we know that we're there, and that's the reason why we're there, because there's not another option.And we have to blend that hard ROI for the soft ROI to get back to what we're what we're here to do, which is extending care for those areas that may not have access to care.And that's why many of us are in healthcare.
Yeah, I think great great thoughts and approaches.I do think that it has to be active.That left alone, we will have a two-tier system and left alone some people are going to accelerate that level of technology adoption while there's there's some that are staying behind.We're we're almost out of time, so last question.
I don't even know and you know, we can go for an hour talking about your what's keeping you up at night thinking about the ethical concerns when it comes to AI and clinical care.And you know, the top one or two things because I am pretty sure we we would have long lists.
Um simple thing, assumption that people make that use AI because one of the things as a as a Green Beret, I I spent 30 years as a Green Beret as an officer, and we go through training specifically aroundhow to make decisions when you're exhausted at the maximum level.And it takes a while to get to a point to do that.People who are exhausted and there's a lack of staff and eventually if you don't watch it where you're supposed to have a checkpoint of a human in AI, if you're exhausted, you're making an assumption that it works becauseit always works.And I think that assumption is a risk that we have to manage around.We don't want AI to be in place and replace humans, but we don't want exhaustion to get to a point to where they stop doing the check because we don't have the support that they need.
And I think that's one of the biggest things that I see that's in place when we're looking at nurses, we're looking at clinicians and thing is that I don't want you to assume the AI is always right ifthere's supposed to be a human check place there.That makes sense?
Absolutely.I think for me is two things.One is the human in the loop and human accountability.We are responsible for every single output that comes out of AI, and it is up to us to have that a human accountability of taking an action, verifying that action, and not just taking the first output that wereceived from an AI clinical evidence AI assistant or whether it be is we have to have those checks and balances system with those right guard rails and train on those guard rails to ensure that we are seeing thoseactions that we intend are carried out successfully in whatever it may be that we intend to do.I think the second thing too is AI for good use is always a thing as I think more broader across just all of humanity is AI for good use is always something that you have some bad actors out
there, but just stressing on the mission of healthcare and on the mission of for the greater good of the humanity is always something that I hope and intend that people have good use for AI.
I talked to our cyber security.Now new vulnerabilities are identified all the time.They're being exploited immediately after they're identified.I mean we see that so we've had to get a lot better and focus a lot more on security and privacy.The second thing is trying to make sure that people still know how to think and to learn.And you know you can laugh, but it's you know in in education for example it's been shown that if you can use AI to enhance learning and to enhance thinking, you can also use AI to outsource thinking andlearning.And I've worked a lot with universities who are developing their their curriculums and I will say I've been very encouraged by the students who are helping develop those curriculums that are they are very thoughtful and they see thevalue in understanding how to do research, but I do think it's something that needs a lot more attention.
Yeah, it's called cognitive surrender.That is something that you know that that is something I I I actually Yeah, I actually worry about a lot mainly because if you think about building your own intuition, right?A lot of a lot of what we a lot of what happens in the world you like there's no logic around it, right?Like you can't actually understand it and you have to build that intuition.Now if you surrender your cognition and your thinking and your memory to an AI to do then you don't build that intuition, and then you don't build the ability to be creative around all of these all of thesedifferent aspects as well.So that's the big thing that I worry the most about is like these things are smart, but they're not
they're not super intelligent, right?They're just token predictors.And if you surrender your own free will and your own cognition to that, like then you're going to lose the ability to do it over time as well.That that skill will atrophy, and that's the thing that I worry about the most.
You know, there was a recent article by a mathematician.They they released a position statement that that um you know, AI has been used to solve 500-year-old problems and some theorems that haven't been proven before, and the mathematician's perspective was that that's the worst thing that can happento mathematics because if you stop trying to solve the problem, you are not going to invent new tools, new techniques, new algorithms.And so it it was great to have unsolvable problems so that we can keep trying, and that's how knowledge progresses.And now that we have a tool that can do that for us, we'll stop trying.And once we stop trying, we stop reinventing in in uh our ability.
8. Second Panel Introductions
The moderator transitions to the clinical applications panel and introduces the physicians. Each clinician briefly describes their specialty and experience with AI and robotics.
Um I want to thank you all.I mean, we could go on and on.Uh we will move into the second part of the conversation where we are focused now on more practical clinical applications of AI, and specifically where we are today, what we're seeing today, and how we're approaching AI in theclinical uh space.But before I do that, I'm going to turn over to my colleagues and ask them to introduce themselves uh briefly, and then we'll start our conversation.
All right.Thanks for Thanks for having us.Um Jeff Paulson.I'm a cardiologist at Roper with Coastal Cardiology.We joined Roper as an employed physician practice about 3 and 1/2 years ago, and uh happy to be be I'm pretty bullish on AI and technology in healthcare and the healthcare space.So I'm glad to be able to talk with you all about it today.
I'm Elizabeth Klein.I'm one of the general thoracic surgeons at Roper.I've been here for 23 years.I'm part of Physician Partners.I use the robot almost every day and I've done over 3,000 robot operations.Many of them are navigational and AI and robotics have really changed the course of my practice and really the course of our patients.So it's all very exciting.
Hello.My name is Dr.George Kieo.I'm a hematology oncology physician.I've been in practice here since 2004.I serve as the director of the Roper St.Francis Cancer Center and AI has absolutely been more and more helpful.It seems every day.So glad to be here with everybody.
Hi.I'm George Curry.I'm a neurosurgeon.Going to talk about how neurosurgery and AI interface and how it's revolutionized neurosurgery over the last 20 years.
Well, delighted to be with all of you today and then we'll start where we started with the with the technology panel.
9. AI for Diagnosis and Risk Prediction
The clinicians explain how AI can combine discrete data from records, labs, and imaging to improve diagnosis and risk assessment. They highlight earlier detection, better interpretation, and more complete clinical context.
What's your perspective?Where can AI help with giving you as a clinician better insight, better patient outcomes, better analytics and data?
Yeah, I mean I I think one one thing that that most patients and maybe even many clinicians don't realize is that over the last 10 to 15 years, Epic and other other EMR systems have basically been calling data.Okay?So when you when you go to the physician, they check your vitals, you get labs drawn, there's information put into your chart on imaging and and data points from imaging sets.That's all data that is it's granular, it's discrete data, and it's data that a human being can't really process in its entirety and and put together the overall picture by culminating all of these data points.So the one that that just the tipping point here that's I think fascinating about AI is that it can pull all of this data together and start to amass interpretation of the data and help you understand as a
patient and clinician, how does that data reflect your risk and how does that shape how we take care of you in the long run?
I think what Jeff said really translates into our patients as well, not only finding people who are high risk for any number of risk factors, not just tobacco, but finding incidental lesions or pope that was an incidentally foundlesion and many of our patients are just like that.They come in for some other reason and if that small incidental lesion on an image isn't followed up on, it can become an incurable lesion.So for many of those reasons, AI is terrific for us as it's called from you know, databases like Epic and whatnot.
Yeah, I agree and I think back to medical school where they taught us about two big parts to medicine, diagnosis and treatment and they are two different things in a lot of ways and we've actually historically been betterat treatment and not as good about figuring out exactly what's going on, right?We all worry about these things and how do we put it all together?How do we take all the pieces and come up to the right answer?And we remember things that our professors told us like common things have uncommon presentations and those may be more common than the very rare things and having AI help you to sort through these things, I think is amajor opportunity for improving care just in terms of accelerating, getting to the right diagnosis so that we can then move on to proper treatment where we where we feel more comfortable, I would say.
Yeah, I would second what's been said.Coming from surgery, we're usually coming for more specific surgeries, diagnosis.
And that visit, that first visit they have a brain tumor, spinal condition is usually nerve wracking for them.It's good that AI is around.We have all the films that they had, you have all the history that they had.You can easily contact our colleagues, you know, for tumors and we already know what they're going to say.I think the biggest thing for neurosurgery is we're able to contact people who help us with the diagnosis neurooncologists and the neuroradiologists and the patient leaves us after that first visit, you know, we can calm them down.They're comfortable with the diagnosis treatment plan.
Yeah, I think what if I could say one more thing while about so in in cardiology so much of what we've focused on for years and years is dealing with illness and waiting for the patients to declare themselvesas sick, heart disease, congestive heart failure.We see you in the ER, you're already sick.And and some of this can also be true in oncology that when you have these these AI, this machine learning based models to look at the data before the patient becomes sick, you can actually shift the paradigm alittle bit away from managing disease as opposed to the predictability.Which patients do I think are going to or we are identifying that are going to get sick and we can actually prevent disease.We can actually and it sounds a little Pollyanna, but it's not.It the data's there.It's just that we have and we don't have the capacity on our own to pull all that together.
And so it's it's really a huge shift for us.
Jack, we had we had a board meeting.I was talking to your partner, Scott, and he was saying currently AI is reading EKGs way better than anyone else could read.And not only is reading them
it's giving you an ejection fraction
which normally you would need a an echocardiogram or some more advanced testing to get that.And so the ability of AI to read raw data, not the translation of the data.And same things happening in imaging where it's reading the pixels.It's not reading the image.It's reading the raw data from a cat scan or an MRI
and translating that into insights way earlier.
Yeah, it's it's pretty fascinating and and you can you can cut off my mic if I talk too much because it's this is all I find this very fascinating.But you know, an EKG is a very cheap We were talking about this out here earlier.An EKG is a very cheap modality and we've been relying on it to identify a change in condition in patients forever.Whereas really with machine learning, we're able to see we're able to predict.We're able to identify things that we never thought were possible with EKG.
I think we we're going to change your mic, George.It's not working.
You want to turn the other one off, I think.
Yeah, the the technology is amazing and you have your EKG on your watch now.
We do.Yeah, and we'll talk more about that as we get into some long-term plays here.
So that that gets us to the second question.
10. Daily Clinical Use Cases
The panel shares practical examples of AI in everyday practice, including note-taking, identifying high-risk patients, and speeding up diagnosis. They describe how AI helps them spend more time with patients and make better decisions.
So what are examples you're using you're seeing AI being applied every day in your practices?
Well, so so we George and I were talking about this a bit ago.So we use AI every day in all of our clinic visits.I'll just give you a little kind of backstory on me.So I was actually an I was an economics and finance guy coming out of undergrad.I worked in venture cap.I was a I was in private equity for a handful of years out of undergraduate.And it just it wasn't filling me up from a from a sort of soul standpoint.I just it wasn't it wasn't really where I saw myself long-term.So I decided to go back to school and I ended up going into into medical school and with the purpose of filling the soul by by addressing the human condition.And only to you know fast forward 15 years down the road spending four hours of my day writing notes.
And and so I think where I'm very excited about AI is how it's navigating this whole patient interaction.It's getting me and my colleagues back to humans taking care of humans, which is really why we went into this to begin with and it sounds it sounds cheesy, but that's the reality and and I'm I'm very excitedabout where this is taking us.I get to when I'm in clinic seeing a patient, I'm face to face.I got really good at typing blind and and you know watching and talking to patients and and making notes, but but I'm wholly listening to them.I'm wholly responding to them and I don't think you get the whole you don't you don't get enough out of the patient experience in the clinic if you if you aren't there with them.So.
I second that.I have to say that the AI that helps us just purely navigate our notes is changed my life and it has given me the opportunity to have a to be what I started long time ago is just sitand talk to the patient and not worry about typing at the same time.But I think importantly AI is helping us now or at least behind the scenes for me.I don't do it every day, but other people are doing it every day is they're finding our patients who are at risk for different diseases and they're getting us them in front of us.And you know you can argue that's going to take some maybe some human jobs away.I think all those things settle out.You know there's a lot of radiology reports that are read better by by an AI program, sadly.
However, it takes a human sometimes to put all that together.But anyway, that's really exciting.So makes my day better.I can I can be an interactive doctor with my patients and talk to them as Jeff said and touch them.And and then the second thing that I can do is I can get more people in front of me and find people earlier who have risks that sometimes they don't even know about.So great stuff.
I'll I'll share a patient example.So this was last week.I was meeting with the woman with her husband.She's had breast cancer that has spread into the bones.It has had a very good response over a prolonged period of time.But now we're starting to see signs that that what we're using in our first line therapy is is not working as well and using some technology looking for actionable or driver mutations of the cancer.We are able to identify those drivers, but but AI helps us take the next step.We say, okay, there's mutation X.But which exact point mutation is that?Well, for her, she had two and one was very significant and was actually associated with the very very strong response she had had for all these years.And the new mutation that was found was actually And you have to Only AI can help me answer this question in seconds in the clinic that this new mutation of say ESR1 a resistance mutation was actually not a
relevant mutation because of its position in this gene where the mutation was.And this is a level of detail that you could not access without a trip to the library back in the day or even a few hours or some time on the internet doing searches.It's allowing us to make much better choices for the care of our patients and and this is almost like for free.I mean, it's it's right there and it's so fast and it helps avoid pitfalls that you would never know that you might be making in the past.So so it the more we can implement this and help uh you know the better off we'll be and our patients will be.So a lot to be optimistic about.
Yes, I would agree with what's been said.Um the hardest patient is the cancer patient that comes back to you after a procedure and you're wondering what to do next and AI is helped us with those patients.Um you know MRI scans are now you can manipulate elements and molecules on the MRI scan.You can tell the extent of resection and what's left.You can do virtual tumor boards.You know the patient comes back and I can get online with Dr.Ki almost immediately and see what he's got up his sleeve for the patient.And so in the again it you don't have to go to the library, you don't have to go online, you just can talk to your colleagues and you can look at the MRI and talk to the neuroradiologist andyou can really make quicker decisions about patient care and treatment.
You know um you mentioned a lot of these technologies are helping improve the the uh time that physicians are spending with the patients.And if you're not familiar with it, we've turned AI in the practices so that it's listening while the physician is talking to the patient and then it's transcribing the note.And we have the ability in Epic to measure how much time does a physician spend after hours or beforehand.So 5:00 p.m.to 10:00 p.m.and then before 7:00 a.m.in that chart.I think the technical term is pajama time in in Epic.Um the average for primary care was more than 4 hours.And that's just time spent after hours.That's not just your normal work day.That's the extra time that you have to go in just to write, document, put things in there that you needed to do.And to a large extent it's been reduced dramatically to the point where you know there are almost a lot of providers that have eliminated that completely with the help of technology.
Still a burden of communication but not nearly as high as it was before and you heard from our panel reiterate that.I think the other one that that we've seen George we we saw a demo of a patient journey that Entropic had where the primary care saw they send the person to specialty, x-rays, CT scans, subspecialist, pathology test andit took eight months through that patient journey to get a diagnosis.And Entropic was able to get a diagnosis in eight minutes.Now they had to they had to use the data.So a physician ordered that data.But had we been able and I think that's the future to bring that technology into the physician workspace it's probably not going to be eight minutes but it's definitely not going to be eight months and we're going tobe able to enhance dramatically the communication between the providers and you know I just saw a paper which is like syndromes are on the rise because finally we can diagnose them.
Yeah.Yeah and in cardiology this is really something it's not just cardiology it's cardiology, kidney disease, metabolic syndrome.There's actually something called cardiac kidney metabolic syndrome that we're that as a as a clinician that is a very hard sort of it's sort of an amorphous concept that there's all these things that are tied to a similarinflammatory nitus like when we talk about coronary disease these days we're not talking about plaques in the arteries we're talking about the upstream inflammatory disease that leads to the plaques formation in the arteries but that's also tied toother syndromes like fatty liver disease, kidney syndromes, diabetes.These are all in in a patient in a patient focused management strategy, you need to be focusing on all of these things at the same time.Well, you you we're so siloed in healthcare that it's very hard to do that unless you have some collaboration and the AI engines help us do that.
Um so it's yeah it's it's definitely definitely enhancing the way we can even approach wellness in the long run.
11. Early Detection and Precision Medicine
The discussion focuses on AI’s role in shortening prediction windows, detecting disease earlier, and supporting precision medicine. The panel also considers wearable data and when to trigger more invasive interventions.
I think we we we mentioned many uh all of you mentioned a little bit about early detection of disease and then appropriate therapy.Where do you see that in your different specialties where AI is helping you with early detection and appropriate clinical care?
Yeah, so a lot of the models we've used historically in cardiovascular medicine have been pooled cohort models.What that means is we we look at how patients do over time and we look at their their different clinical factors like their lipids, their blood pressure, their diabetes status and we try to put together some predictive modelsto determine okay who's going to have over a 10 to 30 year window risk of having clinically significant outcomes.Well, what these AI models allow us to do is we can take all of that data and then also take some of the objective data from EKGs and we can shorten that predictability timeline.So we can in the newer models we can say not only do we have based on the latest algorithms an assessment of your 10 year risk, but we can actually assess what is your one year risk of having
a heart attack relative to your risk factors.And that's unprecedented.And they talked a little bit about that in the prior panel about the the short shortening and truncating the predictability windows and will that does that affect change because ultimately the end of the day we're about not necessarilymanaging disease, but preventing the sequela of disease.And that's to me the really fascinating part.
Wow.I don't think I had anything nearly that exciting.That's pretty cool.I think what AI does for us now at least in my world is that it it it it can see in your pixel reading things that are very subtle changes sometimes in findings on imaging that aren't always pickedup by humans but they are picked by AI for density and presentation and that really makes a difference in patients as well.So I think right now for at least the things that are that are either up and coming and sort of knocking on the door or already ready for us that that's probably the biggest thing I see now.
I think in oncology cancer is a very very very exciting space.I think it's one that we're still learning about and I we certainly don't have all the answers about how to use it yet but I have a feeling that there will be some that come to the fore morequickly than others certain types of cancer that and I can't quite pick what they might be but I just have a feeling we're going to find that this testing is going to be more successful for certain things thanit will be for others and this story is still being written but I would pay close attention and I I think how exactly we use it and how often and what its role is going forward uh we'll seebut probably probably bullish on all of this.
I think we're all seeing patients who come in with the like a full body scan.I mean some of y'all may have gotten you know full body scan that's early detection of tumors.Um a number of papers in the neurosurgery world comparing neuroradiologists reading scans versus AI reading scans and telling the difference between cancer benign disease stroke and other abnormalities of the brain.So I think that is getting to the point where cancers are detected early particularly in the brain the earliest time you can predict a cancer or diagnose a cancer the better the patient outcome is going to be.
You know um I I do not have diabetes so let's start there but I I decided at one point to put a glucose monitor just to see like how my blood sugars is is changing.And I was sharing with some colleagues who were sitting over on the break that I noticed 3:00 a.m.drop in my blood sugar that gives me headaches.On the days I wake up with a headache, that was consistently what was happening.Or I noticed that if I eat a fraction of a cookie, it's not even a cookie, on an empty stomach, my blood sugar spikes.Uh I couldn't help but think about precision medicine.At some point, personalized medicine, you know, um trying to tailor that individual therapy to that individual patient and and absorbing all that data uh will be a maybe an area of opportunity.Do you see that in your specialties and probably it'd be different for the surgical specialties versus the medical specialties?
Yeah, definitely in cardiology.I know in oncology that's sort of the holy grail is is identifying the specific and I won't steal your thunder, but uh but absolutely and and particularly when talking about wearables and sort of the the sort of democratizationof of healthcare.I mean, how many people are wearing Oura rings or or Apple watches that are acquiring all of this data?What what has been interesting is that we're all wearing these things, but it hasn't really tilted outcomes all that much just yet.And so that that actually really gets to that that notion of well, how do we assimilate all that data?Certainly one individual can't take all of your Oura ring data and and analyze all of your body temperature fluctuations to determine if there's some actionable target there.Uh but using AI to help assimilate that data and put together kind of a framework for us.
And it does it already.I mean, the Oura is already doing that.Apple's trying to do that with its with its watch data um identifying fluctuations or changes in your heart rate variability as a marker for unwellness.Um that's definitely going to be a play for all of us.And it and it actually frankly it it really expands our patient population immensely.
I could say a little more just about early detection is what and with these kinds of longitudinal data points maybe from a lot of different uh sources for for somebody we can answer a little better when are wereally dealing with cancer that needs treatment?What's the When are we hitting a triggerable event where we need to act?We need to take out part of someone's body.We need to go through an invasive procedure to see if we find cancer there or not and put them at risk.Uh when do we pull the trigger so to speak on such a move?And how do we really know that this really is becoming something that requires that kind of maneuver?And we're obviously not there yet uh but hopefully with maybe not aura rings or maybe with who knows we'll we'll be able to compile enough and and build validity and confidence in in getting to those points because it
eventually it really makes a big difference to someone's well-being when we say something's in the chest or the abdomen or the pelvis.And um what we have is non-traditional information about that.So stay tuned.
12. Robotics and AI in Surgery
The surgeons describe how robotics and AI have transformed operative care through smaller incisions, better visualization, and safer resections. They explain how these tools improve precision and outcomes in thoracic and neurosurgery.
So what's happening in the procedural world people are getting custom implants in the spine.So instead of having a long incision with multiple uh implants you can make a smaller incision with your own George Curry implants.So companies are making implants specific to the patient.So that's decreasing surgery time, decreasing more morbidity and having better outcomes.So taking AI a little further into the procedural world where are customizing individual implants for patients and that cuts down on surgical time and recovery time in the hospital.
And maybe this question is more on the surgery side, but where do you see AI assisted surgery and where does that field grow?I mean, we're reading a lot about robotics, advancement in robotics.Um, how do you see that space growing?
Wow.That's a great That's a great question and that's something I love to talk about.Um, not only has robotics made our operations better, and I've I'm old.I'm an old surgeon, uh, which means I trained with open surgery, big incisions and long instruments or, you know, standard instruments.And we've converted all of that now, or most of what I do to small incisions and, uh, a robot.And, uh, my patients now routinely after a long operation or major chest operation go home the next day.And not just was unheard of when I started my training, you know, 30 years ago.So I have to remember that all of these things, we still do the same operation on the inside.I think we do it better.Don't tell my teachers that, but um, uh, we do.We We do a very nice job and we're able to see things, but there are so many things, um, uh, that AI is a part of and it helps us build things like seeing lesions that are under the
surface of the lung, for instance.And there are serum markers that a patient can get and so they are they are, um, technician agnostic.I don't have to have I don't have to be there.Someone else doesn't have to be there putting dye or something to find a lesion.But there are, uh, serum markers or serum medications to give, uh, that will then light up a tumor so I can see it, uh, when I have no other marker on the surface of the lung to remove that.So that's been really terrific.Things like that, I can get an idea of where the edges of those of those lesions are.So if I my goal is to only take the lesion out, not have huge margins, I can do that a fair amount of accuracy.There are new technologies coming where I can see through the tissue.
I can know where important structures like nerves and major blood vessels are.I I know where to expect them because I've operated for a long time and I know what the textbooks show and whatnot.But it's really very fascinating now where you can give different reagents and they'll light up these structures.So as you're operating, you are a safer surgeon and you're a surgeon who oftentimes can do a little bit less and still get the same outcomes.So really terrific things that robotics does for me.Do I ever I'm not sure I'll ever going to see the day that a robot does an operation without a human?I'm not sure that will ever come.But but what the what robotics lets us do is really terrific and what the what things that AI have brought in and AI is kind of in this this peripheral ring out here, but it's fed all of this
technology that lets us see things sometimes earlier.Sometimes lets us take a patient with advanced disease and get systemic therapy like George gives and then we can remove the rest surgically.So really a lot of a lot of great things that the combination of robotics and AI do for us.
Yeah, pretty much robotics has revolutionized nerve surgery.Anybody that's had a nerve surgical operation has been part of the AI world.We don't hardly do one surgery without AI assisted technology.For brain tumors, smaller incisions, better resections.For spine surgery, Dr.Stewart, one of my colleagues presented a paper to one of our national councils, 500 straight spine surgeries without a misplaced screw in our program.And so that's just unheard of.Really AI has assisted in safe surgeries.So in the brain world, a complete resection or safer resection without interrupting pathways is the goal then like Elizabeth of their dies you can put in.So it can guide, you can take the microscope and aim it at the brain and see what's a tumor and what's not a tumor and get a safer section.So it everybody that's had neuro surgery in the last few years has been part of the AI world.
13. Adoption, Training, and Workforce Change
The panel discusses how different specialties and generations are adopting AI at different speeds. They also address concerns about burnout, identity, and how AI affects clinicians, nurses, and non-clinical staff.
A lot of positive advancement that the AI is bringing to the table.We heard earlier a question in about adoption.How do you feel the specialties that you're in are adopting this technology, embracing this technology in if you have any reflection on on training and and how are we preparing you know new physician entering the workforce into that?
Yeah, I think it's variable but I will say that that with the new physicians coming into the workforce they have different expectations than physicians that have been working for a while.They have a little bit of a different idea about work-life balance than I think a lot of us have.And so they're welcoming with open arms the idea of being able to go home and just go home and be with their families and which which you can argue is is better for patient care because they're refreshed andthey're not going to be angry tomorrow when they go to see their patients.But but I I think that it's it's definitely variable in terms of adoption.Although I will say that a lot of the the the educational engines like open evidence, which is one that basically it has grown faster.Open evidence is is a online tool for accessing information in regards to to healthcare topics.
You can ask anything.Like I use it a lot for George's patients because they're on all these nibs and nabs and these new medications that nobody nobody actually knows how to pronounce that do have some cardiovascular ramifications when you're on them.And so we we can look them up in in in 5 seconds I can I can do actually a picture of a patient's list of ridiculous medications and it will help me assimilate a risk factor profile in termsof cardiovascular outcomes.So the the adoption on that realm on the educational side been, has been strong.I think the workflow adoption is coming.It's, it's, it's evolving.Um, I'm a little bit of an earlier adopter on that front because I can see the value.But you got to train it.You got to be invested in it.And with employees, we were talking in the earlier panel about employees, non-clinician employees, um, I think that a lot of the younger folks coming up and getting hired, they're already familiar with it.
They're not, I think the fear is a big factor of adoption.They're not fearful of it.It's already running their lives.And so they're very comfortable with the transition when they come into the medical sphere.
I'm going to answer that twofold.Um, I think that using AI has really changed the way we're able to interface with patients.The, the listening, the transcribing and what not.So it makes our days shorter.And and then the other thing that it does, which is my other hat that I wear, so I'm thoracic surgeon and I also run the well-being, physician well-being and clinician well-being for the system, or at least for, letme back that up, for Roper.And AI has been really, there, there are some tools in AI which are really able to find not the people that you, that say, hey, I'm really getting tired here or this is too much work or I don'thave the right instruments, but you're able to see the flags that go up peripherally on people who are sort of the non-responders or the people you don't recognize as being burned out and who are less effective and less
efficient and maybe become unsafe or very costly or maybe just want to quit because they're done and that's not somebody that you really want to quit or would really want to quit in other circumstances.So AI, there are some really terrific tools in the well-being world for recognizing those people on our staff.And that's something that I really look forward to using and working with, too.So, thanks for letting me make that plug.
Yeah, those are those are all great comments.Thank you.I would just say that being early adopter is getting rewarded so rapidly.You're suddenly given things that you you just couldn't do if you hadn't taken the risk.And I speak as as someone who obviously didn't train up during these days.I feel like everyone who's coming up is much more familiar and their demands for AI and improvements are going to be very interesting.And I think it'll help accelerate it as well.So, a lot of a lot of optimism there.
Yeah, I'll take a little bit different tack.I came from a program that really emphasized doctor-patient connection.I think the newer surgeons coming out depending on the computer, you can click a button and it does it all for you.So, there's a little bit of danger of losing, you know, that one-on-one connection with the patient.So, sometimes AI because it's so good, it takes you away from the patient cuz you can just click a button and it puts it all in there.Whereas what the patient want is the direct connection.I think that's what they want.So, when I have a new colleague come on, I they're all good at the computers.My son is better at the computer than I am.You know, and so I want the new guys not only to be good at the computer, but keep that old school connection with the patient and not just depend on the computer.
And you know, our experience has been slightly variable between different clinical work groups.So, physicians tend to be early adopters and they're actually pushing us.We see physicians coming in with the solutions and then we're trying to figure out if if we can implement those solutions or not.We did see uh mixed feelings in the nursing population.And uh you mentioned fear earlier.I think a lot of it is that fear.It's that um fear of loss of identity, loss of control.Uh we saw a demo for a robot that actually gives medications, you know, goes from room to room in the hospital and gives medication.Uh it films your neck muscles, your your epiglottis and it can actually make sure that you took the pills and uh safety records 100% because you know it follows it it's all automated.
But then when we talked to nurses, they were like, well, that's part of who I am, right?Like like giving meds is part of my identity, part of my my professional practice.So we we see some resistance when it comes to the fear that this is going to replace me, where we don't see that as much in the physician workforce where it's going to augment what I do.It's going to but it's never replacing what I do.And you mentioned earlier, maybe one day, I don't know, but we're not I don't think we're going to be there.Uh so so that's the variability in in receiving.And then on the non-clinician side, definitely uh to your earlier comment, mixed feelings.You know, we see uh we have lawyers actually here in in Aroper that were doing uh early adopters of AI that were doing some research, legal research that would take several weeks and you know going into the original
papers and going back to cases that was done in minutes.And so uh in the legal space, we're we're hearing a lot of lawyers saying like, you know, what you used to do for three years in your part as a partner is now reduced to learn how to usethe technology.Um I think our obligation is to go back to the fear conversation.And um one of the things that we're paying attention to in the last few big negotiations uh between nursing unions, uh there's some in Kaiser, in California, in Minnesota uh I think it's Alaina, uh in Minnesota they havea state union.Uh use of AI was one of the negotiated items.And that's tells you how how concerned uh they were about it.And in in both cases it didn't make it to the to the final agreement, but uh it reflects that at least a subset of the clinical workforce has a concern and fear about whether it's losing job or or
losing control and identity.So something for us to keep in mind as we we implement some of these things.
14. Safeguards, Trust, and Shared Decisions
The final discussion centers on safeguards, accuracy, privacy, and preserving shared decision-making with patients. The speakers close by stressing that AI should enhance human care, not replace it, and that the evidence burden is only growing.
I know we have probably time for one more question before we take some Q and A.You mentioned a little earlier a worry.Uh what safeguards we have to have in place uh to make sure that some of these uh concerns that we all have don't materialize when it comes to improving patient outcomes to the use of AI.
Yeah, I I think it's it the answer to that really centers on something that we've done a terrible job of when we've adopted handheld technology, which is making sure that we see it as a tool and not asa crutch or not as something that devours our existence.And and we've done a bad job of that to be honest.And I think we as we enter into this phase of AI coming into every aspect of our lives, we have to be much more intentional about it.Um it's going to have much more power than we even understand.As in as you integrate machine learning, I think keeping the keeping the animal in the cage is going to be very hard to do.And from a healthcare standpoint, we just have to make sure that we're vetting everything that comes through.It always has to be a shared decision making.
It always having it always needs to be a face-to-face decision, a conversation.We have to keep the humanistic component.They mentioned that in the prior uh in the prior panel.The humanism, we need to use this as a as an enhancer of the humanistic quality of patient care, and not allow it to supplant what we do.And it gets back to what I said before.This is why we're here.We're here to take care of human beings, humans taking care of humans.That's the priority.We need to make sure that's that it's a tool that allows us to do that.And the minute it violates that sort of standard, that basic standard is the minute we have to basically pull it out of out of our use.And I don't know how to do that.I don't know how when we release it, how we're going to control it.
But at least we need to be mindful of of of the risk.
I agree with all of that.Um I also think too, and you probably said it as part of it, is I think we have to make sure a lot of that is accurate.And um I think at the beginning um we're going to be fine tuning a lot of the accuracy parts of it.And then I think sort of the one that makes us all cringe is to say, what about privacy?Like where's privacy gone now?It's like everything is there and um and findable, you know, discoverable.So I think we have to figure out a way to not only be sure that somebody's chart that's in your office or that you can access is is accurate, really reflects who they are and the testing that theyactually had done.Um but then what what things are off are out of bounds for us to find?
And um you know, that's I think it's a bigger question.
I think uh one of my main worries is does everyone remember the It's one of the car commercials where the guy says I'm going to take out my own appendix and I got this and you doctors stand back.Um I think there's a lot of challenge um as physicians to spell out risks more clearly than maybe we had to do in the past.And I see a lot of under emphasizing by the patient of the danger of the cancer and a lot of over emphasizing of side effects of drugs that are hypothetical even though they've been maybe proven to help improvecure or life expectancy, uh the conversation shifts.So you have to you have to know the patient is looking at these things and maybe they have a list of 11 herbs and spices they're using cuz they saw on the internet that this is definitely the way
to do it and they could just do it this way and we just have to it's not wrong.I mean it's not blame but but the partnership that you alluded to is crucial and and these questions are definitely going to be here and we just have to be ready to address them so that we can uhwalk with the patients uh through what they face as as well as we can.
So I had a patient come in last week she had a complicated spine problem.Her daughter was with her and we went through the whole thing and then she said what are you going to do and I said well I'm going to do X and she pulled out a piece of paper andsaid that's just what AI said you know so Right.So you have to be careful.I think it's a little bit intimidating and and stuff you know stick to your guns right AI says something it may say something completely different than what you say and so and then I think the other part ofit is when you're hiring folks and MDs and providers you know hire good folks you know I think that's going to be the biggest protection.
I I think Dr.Curry makes a good point though you know we always kind of roll our eyes when patients come in with with they did their own research and uh and I will say that AI actually has improved the quality oftheir research.I mean it it they're coming in with really valid and very very clear very educated points and I'm like you know what you're absolutely right let's go with this plan and uh usually hopefully usually it's in line withwhat I was going to recommend but uh but they've it's actually helped that experience.It's not it's not as much of like oh here can I just take these herbs and and uh can I get away with just taking uh
It is.It's a different experience.The shared decision-making model is definitely stronger than it's ever been.
Yeah.Uh if you have the opportunity to see Hippocratic, it's something.Uh the the ability to make diagnosis at I know on the break we're talking about uh AI in primary care and in in complex chronic medical condition syndromes is a great example of that.That's the type of thing that the technology right now is able to make connections in ways nearly impossible for us.It either takes a ton of time, ton of energy, or simply not not feasible.Um you know 10 years ago I I should probably look up the updated stat.It was you need 27 years to of continuous reading to read every paper that was published in 1 year.Uh I don't know where we are now.It'd be a lifetime.
Uh it's a we can't keep up with that amount of evidence that is coming out there and that's probably the right use of technology replacing us.That probably the wrong use of technology and and I don't know though that we're going to be able to prevent some of that.But thank you again for our panel.Thank you for your time.Um there's a lot more we can talk about this, but we are out of time.