The Hardest Part of AI Isn't the Technology. It's Adoption.

Doug Flora: Hi everybody, Doug Flora here. I'm your host today. If you haven't joined us before, this is where we are decoding community oncology. We have a special guest today we'll introduce in just a moment. If you haven't joined us before, this is Sanjay Juneja, my partner in crime, and Clint Taylor, who's our industry expert who knows how to build the things that we say we need.

And we're gonna get into some of that today. I'd say today's episode is gonna be around decoding community oncology since this is something that you do especially well. And our guests may not need introductions, but we'll do that deeply for you guys as we get into who he is, where he trained, what he did. I'm a community oncologist as well. And we want to speak to you guys about the problems that you're dealing with in your practice every day and how digital health, innovation, AI, and all the other fun stuff can help.

Really glad you're here. I'll hand it off to Sanjay. Yeah. Let's get

Sanjay Juneja: it I'm super excited because today joining us is Doctor. Manoj Kumar Bupathi, and he is the president of Rocky Mountain Cancer Center. He has 55 physicians that he's responsible for, all with their unique workflows and different ways they do things in different locations. And what I'm really interested in is talking about all the evidence that is out there, but translating it into practice, right? Because we have a lot of evidence, but how do we take evidence to actual action?

And there are some technological solutions that are available that can finally make that process so much faster than we can even conceive. And especially what we're used to. So that's why I'm super excited, Manoj, to have you. And you do research at Sarah Cannon, and you also love getting into the weeds about how do we make sure this stuff is actually relevant in clinic tomorrow. Thank you guys so

Manoj Bupathi: much for having me. It's really an honor, and truly, truly speechless, to

Clynt Taylor: be honest with you.

Manoj Bupathi: Such a fantastic effect on people. Yeah, we've been doing oncology decoding now for a year or a year, a little longer than a year, and that's really what we've been focused on is how do we take all this data that's being presented at conferences like ASCO, ASCO GU, ESMO, and how do we help our clinician partners figure out what there becomes next? And why does it come next? And how do we manage these side effects on a practical basis? We actually talk about our own side effects that we're managing, and we'll disagree.

Co host, Ben Garbercy, will disagree with what I'm doing. I'll disagree with what he's doing. That's the reality of where care is today. It's very different depending on what we're seeing in our local areas. But aside from that, it's also how are we going to make it better from an operation standpoint?

How are we going to better use our technologies that are evolving so fast? And how do we make our practice more streamlined so that we can provide better and easier care for our patients and improve our patient experience? We want them to be able to be connected to us, but at the same time, not have the physician burnout that we're seeing as well.

Sanjay Juneja: Yeah. We started today just in the pre talk with the question of what is the biggest contributor burnout? And we had differing opinions. And for me, it really is just having the stuff you need to do your job effectively. And by that, I mean having the data in front of you.

And, you know, it's always been challenging. We use fax machines in Southern Louisiana and most of our institutions. But now with natural language processing and LLMs and all of the things we're able to do, it's actually conceivable to be able to have all of that stuff in real time. And not just the data when it relates to the patient, but also everything that relates to guidelines or things that came up literally in the past week surfaced and contextualized for that specific setting. Clint, I know that's what excites you with all the work you're doing at Oncologic.

You know, tell us how you see that changing and what you're most excited about, about bringing physicians and patients when it comes to that technology aspect to speed things up?

Clynt Taylor: Well, really just looked just a few months ago. We all know what's happened in the last few months. The idea that we could use LLMs to go from what's kind of neat to something that's actually really practical and achievable, that's changed everything. And that's one of a few things I think that's really changed the whole dynamic. We're able to develop software really quickly.

Some of the hard stuff we still have to do, grind through it. But getting access to the data, making the data accessible in standardized format that really opens the opportunities to build on some of these real tools, that's what's super exciting to us right now.

Sanjay Juneja: I would love to know, Manoj, what is your biggest concern with adoption that you suspect may slow down the taking advantage of, so to speak, when it comes to these technologies, especially when you have 55 physicians and I assume 55 personalities?

Manoj Bupathi: Yeah. So that is actually probably the biggest thing is how do we get this used by everybody and be comfortable with it? Oftentimes, change is the biggest concern for everybody. Historically, when we implement something as novel as things like this, the implementation process has to be smooth. That's one of the biggest things that people will say is that you've started doing some new technology, but it takes me longer because it's not implemented right.

When you go back and think about electronic medical records when they first started, it was harder for everybody to do it. And there were so many complaints because it was taking more time to document in the chart, click in many different places, and to meet all the requirements that they need from a compliance standpoint. Now you're talking about streamlining that even more, but the idea here is that it's actually trying to make all those clicks go away. It's trying to gather all the information versus in the original time where we're making this radical change in healthcare, we were trying, it's all very fragmented because you have to click one place to go to the pathology report, one place to go to imaging, one place to document. Labs are different places.

And you look at any EMR, it's always like that. And finding it is incredibly hard. So, we have to prove to the partners saying that using technologies like AI in any sense will actually help you garner that information much faster. You don't have to spend time looking for a result. For example, if I'm looking for a patient who has bladder cancer, and I'm thinking about second line therapy, and I need to know, do they have FGFR?

Well, if I'm looking at the EMR, I have to go back and look at every single one of them and figure out where did I document or where is it stored for the results section. Or if I'm looking for HER2, then I got to go back and find the original path report and then tell my MA saying, Hey, I need you to find this path report to go do this.

Sanjay Juneja: Sometimes the document says, See attached report. And I'm like, thought this was a report. And it's a dead end. You get it.

Manoj Bupathi: Or you can't read it. And then you have to go find the source again. So that also becomes very hard. So using technology will actually help bring all that forward. The other thing is, in a true community oncology practice, a lot of physicians are seeing breast cancer, then they're going to a benign heme patient, and then they're going to something else.

And then if there's a patient that's progressed on two lines or three lines of therapy, sometimes there's a need to go look at NCCN guidelines. Say, hey, what do I actually need to do next? Because the therapies are changing so fast. But that's added time. And that's added time that you're taking away from spending with the patient, or even from documenting, or even answering a phone call, or whatever it may be.

So, if there's a way of surfacing that information upfront, it makes it much easier.

Sanjay Juneja: Yeah. I mean, it's technologically possible. The other side of that, Doug, not to get too personal, but kind of, you're a multi cancer survivor, thriver, and I'm sure you get excited about not only all of the data and elements that relate to your cancers being surfaced to the top and ensuring that you have everything that's up to date when it relates to prognostic or therapeutic value, but there's this other side where, you know, technology potentially will permit you to have guidance when it comes to your cancers for your profile specifically, either with real world data and things that are readily available, or just understanding what your germline status is and your demographic. How does that kind of sit with you when it relates to the management of active cancer? Doctor.

Yeah, I feel a

Doug Flora: little guilty. So for the audience who doesn't know, had kidney cancer cut out nine years ago. I'm fine. But I really got to look behind the yellow tape as a patient. And I thought I was the world's most empathetic oncologist.

We probably all think that. And it was different. And I felt guilt because here I am a super well connected medical oncologist. I go to Cleveland Clinic for my visit to see Doctor. Rinne.

And you know, there's such a gap in community oncology care where not everybody has subspecialized like you have, like I have. Our centers are both doing multi D care in the community, but these tools can kind of level that playing field. Everyone has complete access to information immediately. You're not searching through a fax machine to get your NGS results from a third party. You know, obviously Clint, you're doing this now.

Automation, it reminds me the James Clear quote. I'm sure you guys have all read Atomic Habits. You don't rise to the level of your dreams. You fall to level of your processes. The processes now are machine learning, LLM, instant immediate data.

And so what you guys have built, maybe talk a little bit about that for our audience so they know that these things even exist. Because as a patient, I want my doctor nudged, right? You can't know everything. There's 144,000 articles indexed to oncology and PubMed last year. None of us are that smart.

None of us have that kind of time. And so I love when my little computer says, Hey, are you aware this patient hasn't had matched specimens? There's no RNA on this patient. Which test do you want to run? And I start to look at all the XF, S, and I need help.

So maybe I'll put it over to Clint and talk a little bit about building tools that help the doctors. What were the pilots flying in the air?

Clynt Taylor: Yeah, we think of those as decision gaps. These are gaps that are just, they're just blind spots. I didn't think to look at this, or I didn't have time to look at this, or it never occurred to me that this was an option. We think of those as decision gaps. Those are things that AI is really good at because they can surface them in perfect context, meaning I don't have to try to remember all those things and have them logged in my head.

Now it's in a system that renders and presents just at the right time. So we also think of it in the sense that we've always used computers in a pull sense, Type something in, put a bunch of keystrokes in, wait for it to come back, click through a bunch of things. But if we can use AI today, which we can in more of a push sense, which says, which basically says, Doctor. Flora, here's something we think you probably need to know about because at this moment, today this patient's coming in and this is information that's new, probably relevant, you might want to see it. Not overwhelming you in the sense that we've had to in the past, which was just a bunch of alerts, but really in context contextualized.

Yes. Contextualized along the way. That's where most of us are getting really excited because it doesn't create more things for you to think about. It actually clears the air so that you're actually just thinking about that at the moment you need to. You're decoding?

Sanjay Juneja: Yeah. There we go. Decoding. Capable of Just say you're ahead

Doug Flora: of us.

Sanjay Juneja: Just say you're ahead of us. And, but see, so if someone's listening to this, they're like, that sounds amazing. I would love to know between both of you, what are the steps? If I'm sitting, you know, in my home hearing this podcast, and I said, I want that. Literally, what are the steps to make this happen?

What is the technological lift? What is the bandwidth? Do I need to get a full time, you know, an FTE just to be able to integrate these solutions? Do I need consensus? You know, can you walk us through the process that if somebody is motivated and said, that will make my life better, what do they do?

How do we, how do they start taking action on Monday?

Manoj Bupathi: Yeah, so I think there's a lot of compliance that needs to come in. We have to make sure that we're following the HIPAA guidelines. We have to make sure that patients are informed that we're actually using AI. So that's actually one thing that's been started now. A lot of practices are getting consent saying that there's AI that's being utilized in some shape or form, whether it be in the form of a patient transcription, recording a visit, and then transcribing it to the EMR, or even trying to utilize things where it's streamlining the chart in the background.

So patient consent information is really important. That's number one. Number two, from an FTE standpoint, you're probably going need to add one or two people at least, depending on the size of the practice and how much of a lift it's going to be. There's got to be some element of human check. So you have to have people that are currently in place in order to verify what it's doing is actually right.

I think one of the misconceptions is that AI is going to replace whatever the function of the job that individual is actually doing, and that's not the case. It's actually going to make that person's job easier so that you can get better information. So you're only going to add people to do more complex tasks, while the AI is going to do the more simplistic tasks and make your clinics more efficient. That's the way I look at it.

Sanjay Juneja: Yeah. And one thing that we talked about on previous episodes is it's not that you just have a designated person that says, Yo, when something arises, you're gonna solve this. There needs to be a very dedicated map or organization of what exactly that governance or that process is going to look like. What is the auditing frequency? How are we going to audit this?

Is it 10%? Is it 5%? Is it 1%? How do you shape those things when you don't even have the tool yet?

Manoj Bupathi: Yeah, so one of the things that we started looking at, and I'm going to use an example if that's Okay, is clinical trial matching and using EMR, right? So there are systems out there that'll read through the entire chart and figure out eligibility criteria for a patient in terms of saying this patient's eligible for trial or not eligible for X trial, right? And we let the agent do the background work, figure out who's eligible, not eligible. But then you still have your research staff that's going to go in and double check that, right? See, not everything in medicine is a black and white yes and no, right?

There's a lot of times when pathology reports are a little bit in the gray, and it needs interpretation to say that this actually meets criteria or does not meet criteria. So that's where your CRCs come in. That's where your physicians come in. That's where your physician assistants or NPs, anybody in the medical team will come in to help clarify that. But now they don't have to go through the age requirement.

They don't have to go through, is the blood count right? They don't have to go through all those simple things. It just has to go through and address the very key ideas. So now you're taking somebody that would have waited six days to, or seven days to ten days to figure out, are they even eligible, bringing it down to maybe a day. Now you can start looking at what is that option going to look like because now you still have that person where you ease their burden in figuring out their eligibility.

And so from a workflow standpoint, how we've looked at it is we're probably going to need to add somebody to double check what that eligibility looks like, and then you have your rest of the team to still continue on what they're doing. But you've shortened the process that it takes to enroll somebody on or present an option of a clinical trial.

Sanjay Juneja: Yeah. And it's imperative that you have that planned or in place prior to the implementation or integration of something. Clint, I'm curious, you know, I recall distantly now, in surgical suites, when they had all this crazy equipment for, you know, ophthalmologic things or fancy, devices, that the company would come in and they would kind of like regularly calibrate and assess and essentially just check for the fidelity of the instrument. Where does this fall when it comes to technology companies? Obviously on the clinician side and as President of Rocky Mountain, he has everything on his end that he wants to ensure is going correct.

How much can we rely, or does it vary company to company, on how much is happening on the back end once that's implemented? Is that an ongoing process? Or are they kind of off the shelf, out the box, here it is and, you know, I hope it works out? How does, what does that look like?

Clynt Taylor: Yeah. Well, as you know, with older software, we built a software package and that has to be kind of what you install and support. So it's kind of like, here it is, conform to it, there's some configuration, but mostly that's the way it works. We don't see that with the use of AI really at all. We see it quite differently.

In fact, we developed our own adoption methodology where we actually come in, we have to understand the workflow of the practice from a people standpoint. We think of ourselves a little bit differently. We definitely are a platform rather than just an application. Consequently, we have to think about end to end. What needs to come into this?

What needs to be a part of it? The earlier we can get involved with practices who are thinking about governance and thinking about where's the human in the loop? I think it's better because I think now you have machines learning along with people, both the supporting people and the people who are going to live in the practice, and that gets really baked together at the same time. Somebody told me yesterday, if you've been in one oncology practice and you've seen it, then you've seen one oncology practice, they're all going to be different, their workflows are different, They're dialed in differently. I think this implementation is going to look that way as well.

Sanjay Juneja: It doesn't mean it has to

Clynt Taylor: be long and drawn out. It just means it has to be very methodical and there has to be a process in place to end up with what you're looking for. Can I ask a question, please?

Manoj Bupathi: So when I have a practice of 55 doctors and you're talking about the uniqueness of one practice to the other practice, if I want my AI system or AI ops to conform to the individual physician, can I make that happen?

Clynt Taylor: Yeah. To the extent that you are willing to let them do that, it might be like, Hey, I don't want you to do this because that's wrong. Then you're going to have to create governance that says, as long as we all do this, then there's flexibility within each practice, each preference, I should say, for each oncologist. We did that even with an older technology where we still said a network has guidelines, a practice has guidelines, and then providers have preferences and we can build those in.

Manoj Bupathi: So that's my question because when I look at my partners and I look at me, right, how I write a note, how I think about streaming my process, and how I want to translate that information to the patient is very different compared to my partner, right? If I'm implementing a new system, the concern's going to be, you're going to change how I'm thinking about how I'm practicing. Now I've got to rewrite my note. That's a bigger problem to adapt, right? Because that's where one of the fears is going come is, now it's going to take me more time to adapt to what you want me to do.

So that's why I was asking the question, can I have the AI conform to what my partner needs? So that way it's an easier transition and it's easier to implement. Yeah, and it's going to be that way. Lived

Clynt Taylor: through the EMR, I lived through the Internet, I've lived through these different technologies. I absolutely believe this will be an easier adoption just because we have the context. They've adopted these other technologies. This is going to feel easier because it isn't going to require them to conform as much to a rigid, it's got to be done this way. I think it's going to be much easier to adopt.

That doesn't mean it's going to be like, you know, it's important. We have to think through the process. It will be easier than the others.

Doug Flora: But Manocha, would say for you too, you and I are both executive leaders over lots of different personalities and egos and problems. And I think even before you get to governance, you have to get to trust. And we talked about that with Emily for our first episode. She was kind of our conscience. Everybody has a persona here.

I would say you're the translator for us. Your gift seems to be taking the very complicated moving into this community academic hybrid, which is our new thing, right? You have multidisciplinary care. You've got lots and lots of phase one through three trials. You guys are automating everything that is automatable.

And that's how we're working at St. Elizabeth too. But what I found with physicians that work in my charge, it's the trust factor first. We are here to solve your problems. We're not here to sell you a solution.

And part of that is Clint, that's for you guys and for all of our tech builders who listen into this thing, start with the clinicians, start with understanding and solving the real problem because the second my doctor finds out you have to do 13 clicks, they're out. The second they find out they have to jump out of a medical record, maybe that's not their ideal. And so it's on us to maybe help them understand that these are tools to make their lives easier, to provide better outcomes for their patients, to do all the things that we say we care about, but we haven't enabled our systems to perform at the level that we know is necessary and possible. And know Sanjay and I talk about this all the time. We're tired of the gap.

We're tired of knowing the level of care that patients could receive and do in many places, and then what patients actually receive. So Clint, my enthusiasm for what you said is I love the idea that there's a new floor. Everybody has access to all the information all the time. Everybody can be screened for clinical trials. That's going help the numbers.

Everybody can be screened by mammogram or low dose CT that's overread by a tool that catches more dangers in their faintest footsteps. So I'm always the optimist on our panel. It doesn't mean there aren't problems, but I would say before you even worry about HIPAA and you worry about governance is you start to communicate the urgency. The fact that patients aren't getting the best care right now and compared to what, right? And so we're not talking about the care you get with Cliff Huddis at Sloan Kettering.

It's what is the survival rate for this patient with urogenitourinary cancer an hour and a half from an academic medical center this month? And I'm super confident that AI can deliver on that promise. You don't have to be better than Cliff. You have to get us all closer

Manoj Bupathi: to Cliff. So I think the key word is actually going be flexibility. And I say that because I think from a clinician standpoint, they need to understand that this technology is flexible. Historically, everything has been rigid. And here, this is not rigid.

This can adapt to how we're changing our practice. And you can train your agents to adapt to what you want versus the other way, it was technology comes in, we have to modify our behavior to exactly what you were saying earlier to make that work. And that's the real pain point. Doctors don't love that. No.

And that's the change problem that we have seen historically. But here, partners understand that this is a flexible technology, they're more apt to try it. And so that actually is a key message that I've been trying to tell my partners. It's like, what is the HARP? There is no HARP, right?

It may cost, but the upside is so much higher than the downside.

Sanjay Juneja: This is something I talk to

Doug Flora: my docs about all the time. We call it misuse. What is the risk if we don't? How many patients will die unnecessarily? How many patients will be admitted?

How many people have grade three toxicities and drop off trials if we don't explore these things? And so, I try to create this sense that this is no longer optional. It's table stakes. It's standard of care. There's absolutely no reason not to start to at least upscale your organizations.

If you're not ready to buy yet, train everybody. If you're a I'm a copilot shop. I can't use the other tools because of HIPAA and enterprise solutions. Train the whole team on how to use the tool that you've just given them that sits unused. That's to me a very low barrier to entry.

And you know, we both do this in our practices. My docs did not want ambient documentation. They pushed back. And I'm like, we're solving one of the most vexing problems in community oncology, saving you two point three three hours a day and you don't want it? And they just, they don't want the change.

It's not even about the tool.

Manoj Bupathi: It goes back to the note. It goes back to the fact that they're used to a certain format. They're used to a certain format of saying, I want my A and P up here. I want my imaging here and this is how it's going to be. And that's how I'm used to it, and that's where we're going to keep going forward with it.

But the moment you introduce ambient AI like this, where you're talking about there's some standardization, because that flexibility hasn't been built in yet. But if you start building in that flexibility, saying that now, well, we can modify it to fit the way you want it, there'll be more adoption.

Doug Flora: This is what he does for a living now. I mean, he builds agents that are cogent, clear, concise, but they are physician friendly. And that's why we talk about when you're building, if you get the clinicians in early, the product that you're going to deliver is, we're not going to blink. We're going oh, that's exactly what I need. How did you...

You read my mind. And no, you spent two weeks in our office and you watched over our shoulders and you saw the real pain points before you wrote the first line of code.

Sanjay Juneja: Yeah, I would say one of the biggest challenges is making an agent or a workflow that there's a consensus with physicians. Because we'll have these advisory boards, we'll have 15 physicians, and we'll say, Hey, what do think about this? And it's, I mean, every time it'll be fiftyfifty, you know, or sixtyforty in one direction or the other. But that's really what is malleable about these agentic workflows and AI is like, can actually have a true customization, you know, a shoe, so to speak, that fits, comfortably for a flat footed person like myself or somebody with like a nice, you're really getting personal.

Doug Flora: You're really getting personal for it. Sanjay is opening the kimono today.

Sanjay Juneja: But, but that's a unique thing that otherwise I would think has less friction than other pieces of technology that we've been used to in the past, where it's like, it's pretty binary. Like this is how it's going to work. It's EMR or not, or it's, you know, this interface or not. And, and there is actually a lot more wiggle room and shades of gray, so to speak, when it relates to this kind of technology.

Manoj Bupathi: We keep talking about the physician improvement from documentation and getting them used to this technology and everything else. Let's flip the side. From a community oncology practice, integrating AI, how do you foresee it'll improve the patient side, patient experience? That's often a key thing. Patients want to feel like they're connected to the medical team.

How do you think that this is going to help?

Sanjay Juneja: I have a little bit of a concern, to be honest. And I was talking about this last week. And there are, what I read, 300,000 to 350,000 health apps, digital apps that patients are able to download now. And a lot of these, you're able to upload your records, fairly easily and seamlessly and get all kinds of information, which I love. I'm all about patients being empowered with information.

But what's happening in these cases is it is removing the provider from the process. And you know, whether it's even having all the records or what is being looked up or what it's, what's being told. That's also a big concern of physicians and institutions is like, what is actually correct or not? Am I responsible for that or not? So I don't know what the right answer is because one, patients and foremost deserve to know what's going on and deserve to be educated if they seek in any capacity of what the treatment decisions are, you know, what they're looking at, is my care right?

I think that is 1000% the priority. The second piece is where does the physician fit in that? And I, and I think it's challenging and there are companies that are looking at pieces of technology that, that are bidirectional, so to speak, to where there there's a UI for the patient and the UI for the physician. And you can kind of see windows into each other's homes. I don't know what the right answer is.

And my fear is the bidirectional one is very much the minority compared to there are tools for physicians and there are tools for patients.

Clynt Taylor: I'm the odd guy now, right? Because I'm the only one who's not a doctor. But I will say that, to your point, patients are going to become increasingly empowered with more effective and accurate information than in the past. Should. So rather than the Google search that dumps a bunch of stuff out that has no context, the idea that I'm putting in my medical record in and now the information I'm getting back has much more context.

So I think first of all, we have to just come to the reality that is going to be the case when, and therefore those solutions you're talking about, which create this bidirectional approach, I think is really the right answer. Because if I go to my oncologist and say: Look at all I dumped down from online, And you're saying, Great. We have a great AI tool that allows you to interrogate more about what you're looking for in the context of your patient information, your payer information, things that we already know about you. I think there's a real opportunity to build trust for patients to make them feel like, okay, you're using all the latest technologies because I think that's what's going to happen next, is patients are going to think, I probably know as much more than you do because I'm now looking at all of the universe of information and you're still stuck in a EMR that I don't necessarily know has all the information that I...

Doug Flora: I think the timing's awesome though. And we've got a couple of younger physicians, me, so I will preach from on high now, because I've been in practice almost twenty five years. Your generation is ready for this. My legacy physicians who are older than me were very uncomfortable sharing partnership with the patients. It was more patriarchal.

It was more traditional. I'm going to do this for you. And I see my doctors that are, you know, we're hiring, probably hired 40 people in the last six years. They love the engagement. They want their patients to be deeply embedded in their care.

And that empowerment, I could tell you, is a patient feels good. When you are included in the discussions, when you are treated like a whole person, when you are listened to, it allays fear. It gives you great confidence and trust in your provider. And I think this group of docs that are your age, Sanjay, they get that. And so, I think I'm seeing your guys welcoming the technology because you recognize the burning platform is ADCs are here, biosimilars are here, really complicated care is here.

And the old way we used to do it when we had 200 drugs instead of 200 versions of lymphoma, it doesn't work anymore. Right? And this is why in community oncology, you're seeing so many of us subspecialize. I love the community docs that can do it all. I struggled.

I to pick a lane, a major and a minor.

Manoj Bupathi: So I actually foresee this as being very favorable. And the reason I say from a patient standpoint is I'm going write in my note medical jargon. I'm going to explain the medical jargon to the patient in a very simplistic form. They leave my office, they understand the simplistic form, they go online, they read my note, they think that I told them completely something different. Where do I expect AI to do?

I expect AI to take my medical jargon, translate it to them on their end, the patient version, which is the simplistic way of what we talked about. And they're able to explain it. Any patient that walks into my clinic, the basics are simple. You need to walk out understanding the basics, right? Where is it?

What is it? What stage is it? Is it curable or not curable? If you cannot understand those basics, then I didn't do a good enough job. But my note says none of that.

My note says stage X, whatever it is, presented this way. This is the treatment options that we talked about. This is the side effects. So when they read my note to what I talked to him about, there's an easy connection.

Doug Flora: Yeah. So this is the two hundred and forty first Google translate language. It could be medical speak. Maybe they can do the multimodal and they can look at our prescriptions as well.

Sanjay Juneja: It's truly, it's harmonization. You're harmonizing the relationship in a way that otherwise could actually foster mistrust. There is one question that I have to get to, because I don't know many people that would be better suited for this question, and it's a little controversial. And I even hesitate to ask this, but what do you think... We talked about all of the ways that AI is working in an internal capacity.

Tools, things for patients that help them understand things that make our job easier, but externally or more systemically, we are able to query data and outcome outcomes longitudinally on patients with multiple variables in a manner we've never been able to do before. What is that implication on how we think about evidence and evidence based treatments, because right now we essentially tabulate and catalog what's happening with a patient on an organized, you know, trial that somebody has to come in, you ask questions, and of course there's their own leakage with, with having paperwork than, than just knowing what's happening.

Doug Flora: And just linearly.

Sanjay Juneja: That is literally anyone listening. That's the way we do things right now. And we say, because of that study, where we went through these check boxes, this is what we found to be evidence. How does, how does treatment look now that we are actually able to capture all of those things and is real world data going to somehow, perhaps one day, trump trials that are more tabulated?

Manoj Bupathi: Yeah, I'm going to disagree a little bit that it's controversial.

Doug Flora: I like when people disagree with Sanjay. I do it all the time.

Manoj Bupathi: I don't think it's controversial. I actually think that's the way it should The reason I say that is because it takes a lot of bias out, and you are now using systems that can tabulate the information much more accurately. You can get a global picture much more accurately. We know we extrapolate thinking that because a drug works in a general population, that it works in every disease state across that population. Like, for example, if you have bladder cancer, you're saying EDP works, But does it really work in the Asian Southeast?

Does it really work in the Western population? Where are those differences?

Doug Flora: We don't

Manoj Bupathi: have any of these information.

Doug Flora: That's the misuse. Yes, the tests are validated and it's not my population. So they get the least of everything again, and the gap widens. I love that you brought

Clynt Taylor: that up.

Sanjay Juneja: It's like I'm telling the patient, Oh, by the way, asterisk, you weren't really represented in this evidence, but like, we're just thinking it applies to you.

Manoj Bupathi: Right. And so that's, so the real world data, I think actually is going get more powerful. Historically, there's always been a concern that real world data is not as accurate Or complete. Or complete. And if you're not and so do you really even look at real world data when you have so many gaps?

And so, I mean, I can say as being clinical trialist for so many years, I don't put a lot of emphasis on clinical real world data. I put more on the actual published data because that's much more cleaner and I know I'm not missing any information. But if we're able to use AI to capture all of it and everybody's able to do it, then the data becomes far more real. And so it actually may be that your population is more representative compared to the actual clinical trial because that's one big thing in community oncology is, it's always, well, you pick the best patients for going on this study, and that's not the patient that I see tomorrow. And so when you tell me that the toxicity is this well or tolerated this well, but I'm not seeing that, uptake's the not going be there.

Sanjay Juneja: Yeah. And I mean, that is the definition. When people say, We will go into a personalized treatment world, that's what they mean. When they say precision, it's not just biomarkers and getting NGS. We're talking to your cohort, to your demographic, to factors we can't even know to consider because like it's very humbling when you get the insights from AI on what actually ends up being relational that you didn't, we didn't even think to look at.

It was like a move. What was it? Move 42 or, you know, 37, you know, with AlphaGo and Google, like they did a move that they just, all the grandmasters said, Oh, like this doesn't make any sense. And sure enough, it did. So that's, you know, I get super excited about that.

And, and anyone listening, if nothing else, for our children and us and our family members, like that's the world that is almost immediately possible. But the biggest bottleneck is going to take, is going to be collaboration, like working together, less friction. Those are going to be to your point of like, you know, what is the cost of not using or what is the damage done? It is going to be completely contingent on how much do we come together, how much do we actually prioritize this and try to maximize its value? And that to me is my biggest concern is like, is what is that delay?

Know?

Doug Flora: Yeah. Got to land this plane. We're out of time guys.

Sanjay Juneja: These people have to go back

Doug Flora: to their lives and I suppose you have to as well, but you one of the things that we're dealing with now is, and I know you know this, Manoj, because you are a purchaser, right? We are getting inundated with lots of people that have great ideas that might solve problems for us. And I'll tell you what I'm struggling with is can't purchase 200 programs to solve 200 problems for 40 doctors. And so I know that the world is moving away from maybe isolated point solutions to these broader platforms that might go anywhere from screening to survivorship. I'm curious how you're approaching this as a buyer, as somebody who has to evaluate these tools, who meets with vendors all day long like I do.

How are you approaching the how you decipher where to get started in terms of the types of tools? Because I'm not looking for 200 different tools at the same time.

Manoj Bupathi: That's a super important question, actually. And it's just like anything else. I think we have to start small, because it's the fear factor, right? It's the fear factor and buy in of the remaining partners to make sure that whatever you're doing, we need to be successful. So I'd rather do small, aim at over 100%, and then say, we're going to go a lot bigger.

And then they have a lot more buy in. So I started looking at doing small steps. So we're going to start tackling one task at a time. And what is our most pain point? Our most pain point right now is how can we streamline clinical trials?

We have amazing new drugs that are in development, but we're not able to provide that to everybody for any number of reasons. And there's so many that you can think of. But if we are able to at least streamline that portion first, and at least know that all of our patients, every single one of them that's ever walked into any Rocky Mountain Cancer Center is going to get screened for a clinical trial, and I don't have to worry about putting in 50 more people to get that task done, that's an easy fix. So if I can solve that fast and then show my partners, now you have this.

Doug Flora: It's early wins. It's an early win.

Manoj Bupathi: And they also are going to be able to get the experience of using a new drug, which excites me, right? And if they get the use of a new drug, then they will continue to provide that better care for everybody. If we keep building on that, then we'll be able to go, instead of doing one more smaller change, we can just make a bigger interface change because they'll see that huge benefit.

Doug Flora: So you're Jim Collins fire bullets before cannonballs.

Sanjay Juneja: Yes. Yep. And I think that's such an important concept because what you're, you're not saying that you need a point solution first. You're saying, I'm going to put in the effort to have something that can grow once I've done that. I don't think people realize the amount of lift that is required, both legally and security wise and financially, to have your data normalized and harmonized in a manner where you could have an AI tool.

So whether it's a direct connection, whether it's health information exchanges, the first step is to have all that stuff normalized and harmonized in a manner that then the data elements can be used for a purpose. And for you, you said, I went through that entire process and then I'm doing clinical trials. But if you have a technology or a company that is able to take those same data elements, all of that lift that you put in, you can now have additional tools, whether it's clinical decision support, whether it's billing, whether it's prior authorizations, you're taking advantage of the fact that you put in the legwork previously. I think what some people make the mistake of is they'll do one company, and then they were like, wait, but I already have everything normalized and harmonized. How come I can't just do this other point solution?

You're having to do all of that contractual work and that IT lift again. That's an important question to ask, and there's no right answer. But what are the different things I can prop up when I'm able to take the steps, the laborer steps, to be able to actually make this possible?

Manoj Bupathi: That's a very fair point. It's gonna be the same things, whichever decision you go. So you wanna find a company where you can, or find a tool where you can at least start slow. So when you're expanding, it's just easier to expand.

Doug Flora: They grow with you. Yep. That's great. So for our audience, I should point out this is the McCormick Centre right behind us. We're at ASCO.

This is my twenty fourth year at ASCO. And I will tell you, walk in the halls, the vibe is different. People are leaning in. People are now genuinely curious about the tools that we're discussing today. And it's, you know, they're sober.

They're oncologists. There's some trepidation because it's so new, but I think we've definitely moved in oncology past the do I need to, to how do I. And that's an exciting thing for me. This is even different than ASCO last year for me. And I was speaking at a bunch of AI things.

You had a bunch of presentations. We've got a couple more this week that we're doing tonight and tomorrow. But people are now asking us, how do I work? How do I get this stuff started? So you being here as a builder, you having 55 doctors that you're trying to help engineer change gently, but aggressively, if you can do both at the same time, think it's a pretty fun thing to be here happening behind us with 45,000 of our friends.

Sanjay Juneja: That's an excellent point, Doug.

Doug Flora: I like, I, I realized that. I thought so too.

Sanjay Juneja: I thought you were forty four years here to ASCO,

Doug Flora: but Manoj, I was

Sanjay Juneja: wondering, I'm curious, have you, have you witnessed that as somebody that's like, you know, president of, of, a cancer center? Is there a shift from curiosity to more urgency or FOMO, as some kids say, fear of missing out, or just the fact that I am not optimized if I'm not using these things and I need to figure it out?

Manoj Bupathi: Yeah, it's actually the case, I think. I think some of these tools that we're talking about and seeing this year, we've been asking for something like this for years, and it's never been there, right? So it's never been there because it's always the same pain points, same pain points, and we're finally at a stage where there's many people that are now presenting at the same time. It's really exciting actually, because even to what you were saying, I didn't see this last year. Last year when I was here, I saw some AI, but this year it's way more.

And it's a lot of exciting tools that have come way far in a very short period of time. So it's really exciting to see what's going to happen in the next year as we adopt more of these tools, because these agents will get trained pretty fast and then you can really customize and see where we're growing at.

Doug Flora: Think it's better and better. I don't think people realize that. Recursive learning. It's iterative. As we're at ESCO, and we're going to go home, we like to leave our audiences with homework projects because they have so much free time.

They can't wait to do extra chores, But I think they need practical steps like, okay, they love what Sanjay said. They love what you said. They liked what I said. What do they do? They get home Tuesday.

They dig out of emails. They're starting to get things together. If you were going to another Rocky Mountain Cancer Center and you were telling them, how would I start today? What would you tell those docs now that they're leaving ASCO and they're actually really enthusiastic and want to get going?

Clynt Taylor: And this

Sanjay Juneja: is their homework. This is the homework. Is the homework project. They're to practice

Doug Flora: the speed of innovation.

Manoj Bupathi: How do

Doug Flora: they do this thing?

Manoj Bupathi: Yeah, I would tell them to just start playing around with AI, right? There's a lot of people that actually hear about AI, hear about AI, but then the question of what do I do next to what does that mean, right? Then there's a number of partners of mine, friends of mine that actually don't even, they're hesitant to use it because they're worried about whatever it may be, right? I would say, download Claude, download ChatGPT, Grok, whatever you want, Gemini. Start putting information in.

Just test it for yourself. See what you think. Before you get into the medicine side of it from a physician standpoint, just explore AI. Just download the app, put in simple questions, see what you get. And I think it'll be very reassuring to you that it's actually gonna be quite addictive because you're gonna keep testing that.

Doug Flora: I would say, add one thing though, have a conversation, right? Don't look at it like a search engine. Don't look at it like Google. I would even say for those of you that are really excited about this, don't even type. Because I'm finding typing is a filter.

And I leave a meeting and I use Plaud for everything. I record everything I do. And then when I debrief in the meeting, I had nuance like, Hey, the CEO seemed like he cringed when I said that, or so and so had a really uncomfortable look on her face and wasn't excited about it. And that nuance is what the large language model can turn into the magic. And you know, as I know it, there's dopamine there.

This is a quick hit. When you see the work that it does, you're like, I cannot believe that just came out of my head because it's better than anything I would have produced without the help.

Sanjay Juneja: And let me, I'm just going to give one also use case. If you're curious and of course, you know, feel comfortable linking your Gmail. I got on and I had, so many things for ASCO. And I was like, look, there was something on Saturday afternoon. I asked my LLM of choice.

There was something on Saturday afternoon. I know that it was a panel and Doug Flores in it. Tell me what it was. And it immediately surfaced it and said, based on your Gmail account, it seems like this is the grand rounds that we're doing, etcetera, etcetera. It said, this time, this is the contact.

I said, I need the contact of that person. So that's a use case of how you could just make things easier. And when I have a flight, I said, Hey, I'm flying to St. Louis. What's all the deets?

And it will just pull up the airline, pull up everything other than me having to type in, was it American? Was it Delta? You know, all of that stuff. So there's a little piece of homework that if you want an actual, you know, granular example that you can make happen. I don't even know what are details.

Details.

Doug Flora: I'm just kidding. I'm not okay. Guys, Sanjay is my little brother. Thank you all for joining us for that.

‍