Latest / CareTalk: Healthcare. Unfiltered. / AI and The Future Of Behavioral Health w/ Alon Joffe, Co-Founder & CEO, Eleos Health

Transcript

AI and The Future Of Behavioral Health w/ Alon Joffe, Co-Founder & CEO, Eleos Health · CareTalk: Healthcare. Unfiltered.

• 0:07
  John: Welcome to Care Talk, America's home for incisive debate on healthcare, business, and politics.
• 0:13
  John: Today, my guest is Alon Joffe, the chief executive and co-founder of Eleos.
• 0:18
  John: I'm John Driscoll, the chair of UConn Health.
• 0:22
  John: Welcome to the show, Alon.
• 0:25
  Alon: Thank you, John.
• 0:26
  Alon: Thank you for having me.
• 0:27
  Alon: Thrilled to be here.
• 0:28
  John: So, so tell, I mean, there's a lot of technology companies in healthcare, and we'll get into what your company does for vulnerable populations, but how did you come to found this company, and what motivated you to, uh, to found Eleos?
• 0:46
  Alon: Yeah, thanks.
• 0:47
  Alon: Actually, you know, we operate in this behavioral health product space, and we got into this, and also for me personally, because of friends and family members dealing with behavioral health conditions, right?
• 0:58
  Alon: We've seen them progressing and challenging through, um, through those, uh, through those conditions, and we've seen how for some of them it yielded into an amazing progress, and for some of them it didn't.
• 1:10
  Alon: And we asked a really simple question, which is, why?
• 1:13
  Alon: What happens in those conversations in the moment that the door is-- when you enter the room and the door is shut, what happens that yield to an improvement, yes or no?
• 1:24
  Alon: That's kind of where we started from.
• 1:26
  Alon: This is, I'm thinking back, John, to 2020, end of twen- end of 2019, pre-ChatGPT, before AI was cool, pre-COVID.
• 1:36
  Alon: And we really thought that in behavioral health, the conversation is the treatment, but what do we really know about these conversations?
• 1:42
  Alon: That's kind of how, how it started.
• 1:43
  John: But you start as a, as a-- You did military service in search and rescue.
• 1:48
  John: You were a venture capitalist.
• 1:50
  John: Why start a company?
• 1:53
  Alon: Yeah.
• 1:54
  Alon: Uh, for me, it runs in, in, I guess, in the family blood.
• 1:57
  Alon: Um, my grandfather opened the first, uh, private hospital in Israel back in the, you know, '60s, um, and my father ran it for 20 years.
• 2:08
  Alon: So I grew up in a healthcare entrepreneurship family, so maybe, you know, maybe I-- it, it's, it's in the blood somewhere.
• 2:16
  Alon: Um, and we really wanted to make an impact on the space that, that we care deeply about, that, you know, there's really a lot of vulnerable population that we thought we can help.
• 2:26
  Alon: And, uh, we thought this is also a market that is, unfortunately, the s- sometimes the stepbrother of traditional healthcare and is long behind in some areas.
• 2:37
  John: Did I read my notes correctly that it's your, it's your access to on-demand healthcare that was quite good in the military that sort of gave you a sense of what, what's better versus what's available today?
• 2:49
  John: Is that, is that a fair reflection?
• 2:51
  John: Uh, I mean, 'cause it's, it's sort of remarkable if that's the case.
• 2:57
  Alon: Yes.
• 2:57
  Alon: I think, you know, in the Air Force, it's very, very different, right?
• 3:01
  Alon: Um, and I think we had this hindsight, or we were well-positioned to potentially read the map and say, "Look, all these conversations, the technology is starting to get there in terms of a place that we can start to understand what's happening inside all these conversations," right?
• 3:20
  Alon: Today we call it AI, and we call it large language models.
• 3:23
  Alon: Back in the day, we called it some other names.
• 3:25
  Alon: But I think we had that foresight into what's gonna happen- And then of course, COVID came in, the taboo just, you know, spiked the demand, and it was the perfect storm into building this company at that time.
• 3:42
  John: And the, the wild thing is that demand has stayed high even as the- Correct … pandemic, as the viral pandemic has left us.
• 3:49
  John: And so a- as, as I understand it, you are going to deconstruct what works in human to human and try to embed it in technology pretty early on.
• 4:00
  John: Was that the vision?
• 4:02
  Alon: Yes.
• 4:03
  Alon: The vision was how can we, from the encounter, from the conversation, get into a place where we can guarantee clinical and financial outcomes?
• 4:15
  Alon: And of course, that is a lot to unpack.
• 4:17
  Alon: What is guarantee?
• 4:18
  Alon: Guarantee, yeah.
• 4:18
  Alon: What is… Yeah.
• 4:20
  Alon: What is, what is outcomes?
• 4:22
  Alon: Um, but that was the core vision.
• 4:27
  Alon: If in behavioral health, the conversation is the treatment, if we can deconstruct those conversations, we can potentially start to understand what's working for whom.
• 4:38
  Alon: And guess what, John?
• 4:39
  Alon: In that process, we also understood that the conversation is also the documentation, the compliance, the coding, the eligibility.
• 4:47
  Alon: All the downstream effect starts at the top of the, the waterfall from that first conversation.
• 4:56
  John: What constituted the data of these confidential conversations, and how do you measure, how do you define success?
• 5:04
  John: 'Cause behavioral health, I think i- in addition to being stigmatized and, and being the second cousin, to your point, it's also kinda hard.
• 5:14
  Alon: Yeah.
• 5:15
  Alon: Yeah.
• 5:16
  Alon: First of all, you understand that really the, the heroes are the, the social workers and the clinicians on the front line, which are really dealing with really tough conversations, sometimes eight a day.
• 5:27
  Alon: But when I go back to the origin story, this was a really difficult chicken and egg problem.
• 5:33
  Alon: No one has data.
• 5:34
  Alon: No one had… It's not like, you know, when you think about imaging, right?
• 5:38
  Alon: There is structured data sets that you can license and, and collaborate with different IDNs.
• 5:44
  Alon: Here, there was nothing.
• 5:46
  Alon: So how do you build models- It's
• 5:47
  John: not, it's not what does a cat look like.
• 5:50
  Alon: Exactly.
• 5:51
  Alon: So how do you build models if you don't have data?
• 5:53
  Alon: But how do you get access to data if there is no value that you can provide?
• 5:58
  Alon: We, uh, managed to solve it, uh, using kind of more linguistic approaches.
• 6:05
  Alon: I won't get into all the technical details, but we managed to solve it and actually provide value to clinicians.
• 6:11
  Alon: And what you saw was really interesting, John.
• 6:13
  Alon: You saw that if the cl- the clinician has such a big pain point on the administrative side, and it was such a big barrier for them to stay at their job and actually provide care, that if you manage to reduce that even by 20%, that's where we started, today we're at, like, 90%, even by 20% them and the patients are on board And you can start to generate that flywheel
• 6:44
  John: So you embedded yourself in workflow by automating and simplifying administrative tasks.
• 6:50
  John: Is that basically- Exactly … correct?
• 6:52
  Alon: Exactly.
• 6:53
  John: And then, and then how do you engage in the process of a, of a, of the complex, vulnerable conversation?
• 6:59
  John: Like, what does your technology do?
• 7:04
  Alon: Yes, I think that's where you need to draw the line between a lot of technology that we see today potentially in the scribing space, and what does it mean to… when you apply them into a very specific field like behavioral health, like addictions, right, like, like home health and hospice.
• 7:23
  Alon: You need very different approach in terms of the models you work with, the data retention.
• 7:29
  Alon: How do you even know what is a clinical technique?
• 7:33
  Alon: John just asked me how am I feeling today.
• 7:35
  Alon: In every other setting, it's how am I feeling today, but in behavioral health- And then you stop listening… it might be a CBT technique.
• 7:41
  Alon: Exactly.
• 7:41
  Alon: Yeah.
• 7:42
  Alon: Exactly.
• 7:42
  Alon: In this section, it might be a s- a, a CBT technique, might be a clinical technique, right?
• 7:48
  Alon: Um, and so you really need very different approach into how you in-integrate in, into the workflow.
• 7:55
  Alon: How do you d-deal with this sensitive data?
• 7:58
  Alon: Think about potential suicidal and homicidal ideations, right?
• 8:04
  Alon: How do you even treat that data?
• 8:06
  Alon: It's a very different tech stack of what you need to build, and I'll just give one last pointer on that.
• 8:14
  Alon: The most complex, probably, setting for a technology like this is in behavioral health.
• 8:20
  Alon: Imagine 20 people sitting in the room, talking for three and a half hours.
• 8:25
  Alon: That's group sessions.
• 8:28
  Alon: You need to decipher each person and compose that narrative across three and a half hours, okay?
• 8:37
  Alon: So you… when you're building for a specific end market, it's really different.
• 8:41
  John: But to your point, a scribe is simply capturing information.
• 8:45
  John: You know, the cognitive behavioral therapy technique is a form of, uh, you know, sort of call and response inquiry that's, that's structured.
• 8:52
  John: Um, you're, you're getting all this information.
• 8:56
  John: What are you applying to kind of create the logic and the learning around what works and what doesn't?
• 9:04
  Alon: A, you go through clinical studies.
• 9:08
  Alon: And we've published north of 12 peer-reviewed clinical studies, including randomized control trials.
• 9:13
  Alon: Like, you need to put in the work to show that it's actually effective and accurate and safe.
• 9:20
  Alon: Second is you work with clinicians on the field.
• 9:25
  Alon: Today, we work with north of 50,000 clinicians across 41 states, and you get the feedback from them, right?
• 9:32
  Alon: How much is this accurate?
• 9:34
  Alon: How much did, how much s- time did it save me?
• 9:36
  Alon: How much cognitive load did it reduce?
• 9:38
  Alon: By the way, I think what ChatGPT did to, to the world, it did a worldwide market education.
• 9:45
  Alon: Now everyone is an AI expert.
• 9:47
  Alon: Everyone is expecting AI.
• 9:50
  Alon: So now we're seeing AI trickling into other parts.
• 9:55
  Alon: Let's talk about compliance.
• 9:56
  Alon: Let's talk about clinical insights.
• 9:58
  Alon: Let's talk about eligibility, which is a huge problem right now in, in Medicaid.
• 10:02
  John: If I hear what you said c- correctly, you're leveraging the, the, you're by-- you're, you're reducing the clinical load, you're engaging in workflow, but then you're actually calling back and saying, and if creating a feedback loop where it's led by the clinicians, not the technologists.
• 10:17
  John: Exactly.
• 10:18
  John: The technologist is- Exactly … a tool for the clini- for the clinical leadership.
• 10:23
  John: And so it's a, it's ideally, I think, a learning system.
• 10:27
  Alon: Correct.
• 10:28
  Alon: Absolutely.
• 10:29
  Alon: And a- at the end of the
• 10:31
  John: day- And does that help the clinicians get better at the clinical part?
• 10:37
  Alon: I, I can tell you a really quick story about a real clinician from New York I met a month ago.
• 10:44
  Alon: She is dealing with a patient with severe mental health conditions.
• 10:49
  Alon: That patient was not willing to look at her eyes, did not have a job, and barely came to sessions What she's done is, for the first time, she dropped her notebook and her pen, right?
• 11:04
  Alon: And she stand up-- she stood up, she walked besides him, she sat on the floor, and looked from the bottom up into his eyes, and created that human connection, right?
• 11:16
  Alon: And, and what happened, that's what she told me, her name is Sarah.
• 11:21
  Alon: What happened is now she managed to create a connection where there wasn't one, a therapeutic relationship where there wasn't one, and now he found a job, he's coming to sessions, his symptoms are improving.
• 11:33
  Alon: That's 99% because Sarah is an amazing clinician.
• 11:37
  Alon: But I think it's a real good example of how 1% of how technology can unlock her potential in a way, and allow her to be the best clinician that she is, right?
• 11:48
  John: Was that because… And will, and will you then popu- propagate that information out to other clinicians so that there's can-- they can kinda learn as well?
• 11:59
  Alon: You, you, exactly.
• 12:00
  Alon: You share that knowledge.
• 12:02
  Alon: Also, Sarah, as a clinician, now understand that she can treat maybe slightly different than what she, than the, the chains of administrative burden that she was tied to, right?
• 12:13
  Alon: Um, and, and you're seeing clinical improvement.
• 12:19
  Alon: You're seeing clinical improvement in, in, in those very vulnerable populations.
• 12:25
  John: And be-before we-- I'd love to get into Medicare and Medicaid, but before we go there, so how does, how do people get access to your technology, and who pays for it?
• 12:36
  Alon: We-- Elias today is what we call a system of action, meaning we're deploying AI into multiple workflows within, um, an enterprise provider organization.
• 12:48
  Alon: So we work with big organizations that employ hundreds, sometimes thousands of clinicians.
• 12:54
  Alon: Um, these are mostly community behavioral health centers, addiction centers, home health and hospice agencies, and so we, we contract directly with them, right?
• 13:03
  Alon: Today, we don't, we don't work with, like, private practice because we feel like this, that's where most of the volume is, and that's where most of the impact is.
• 13:12
  John: So it's an enterprise sale to- Correct … mental health, you know, largely l-larger enterprises.
• 13:19
  John: Is it a, is it a software license?
• 13:21
  John: Is that what you're kind of at a classic- Yeah … SaaS subscription model?
• 13:24
  Alon: Exactly.
• 13:24
  Alon: Mm-hmm.
• 13:26
  Alon: Exactly.
• 13:26
  John: And, and, and but let-let's get into a little bit of the, the, the vulnerable population stuff.
• 13:32
  John: A lot of those enterprises that you're selling to are serving Medicare, the, the federal program in the United States for the elderly, and Medicaid, the poor.
• 13:41
  John: What's-- As, uh, as we're going through, you know, restructuring, you know, health plans coming in and out of the Medicare managed care program because of V28 and covering different geographies with different kind of benefits and, and Medicaid, a lot of people, uh, you know, falling off of coverage and falling off of the exchange.
• 13:59
  John: What's going on in your practices?
• 14:01
  John: What are you seeing happening to vulnerable populations right now?
• 14:07
  Alon: John, I think right now it's probably the most, um I would say turbulent times for these population.
• 14:16
  Alon: And let me just give you like three quick headlines.
• 14:19
  Alon: One is we're seeing, you know, Vice President Vance announce, uh, $1.3 billion suspension of Medicaid payments in California.
• 14:29
  Alon: We're seeing fees for Medicare enrollments for, uh, new home healthcare and hospice providers across the country, right?
• 14:37
  Alon: There's a lot of focus, uh, from Dr.
• 14:37
  Alon: Oz, from CMS on fraud, waste, and abuse in Medicaid and Medicare right now.
• 14:44
  Alon: And I'm, I'm not getting in, in this show in whether it's right or wrong, that's besides the point, but we're just seeing a lot of focus on that area.
• 14:54
  Alon: We're now seeing AI getting deployed to basically understand moving from what they call, "Hey, we used to pay and then we chase you," to now we're auditing first and then paying later.
• 15:06
  Alon: That's a huge mind shift in how money's flowing through Medicaid and Medicare, right?
• 15:14
  Alon: And so I think for these populations- Well, and it's a massive
• 15:16
  John: potential cash flow hit too.
• 15:18
  Alon: Exactly.
• 15:19
  Alon: And remember, these are organizations that already have very thin margins, right?
• 15:24
  Alon: Um, so we're seeing a huge shift, and for these organizations, it's about change.
• 15:34
  Alon: If they don't change now… That's what I'm hearing from CEOs.
• 15:37
  Alon: Their, their mindset is, "If I don't change now, if I don't rethink my workflows and how I operate, it just won't be sustainable." Right?
• 15:46
  Alon: "It won't be sustainable."
• 15:49
  John: How at Eleos are you trying to help support them in this change?
• 15:52
  John: I mean, it's obviously disrupting a lot of the core provi- I mean, the, the, the, the many of the largest providers, if not all of the largest providers in behavioral h- health, health, substance abuse disorder, hospice, they're all large enterprises
• 16:09
  Alon: Correct.
• 16:10
  Alon: One example is, you know, now with HR1, um, they-they've introduced two work requirements per year, right?
• 16:20
  Alon: That means that-
• 16:21
  John: The, the, the HR1 being what the president somewhat, uh, quixotically calls the big beautiful bill that- Correct … introduced a lot of structural changes to Medicaid, hence, uh, one of which is the work requirements.
• 16:35
  Alon: Correct.
• 16:35
  Alon: So the result is many patients might lose their eligibility.
• 16:42
  Alon: So here is a classic example of how AI can help, is instead of the processes today, which are really manual, picking up the phone, calling, clicking on different portals, et cetera.
• 16:57
  Alon: If you can shift left that, if you can move that upstream, you can uncover a lot of eligibility.
• 17:03
  Alon: And by the way, if now Alon on a, on a, on a session said that, you know, "I lost my… I qui- I left my job, I quit my work," right?
• 17:14
  Alon: That is a flag that will go into eligibility workflow.
• 17:18
  Alon: I can tell you that one organization in Texas we work with, we just uncovered for them over two hundred thousand dollars, uh, worth of, you know, eligibility a year.
• 17:29
  Alon: That's, that's two thousand six hundred sessions But they can now provide 2,600 sessions that they can now provide to their population
• 17:38
  John: And they're not high mar- a high margin institutions.
• 17:41
  John: These are, these are- No … these are or- institutions that are organized.
• 17:44
  John: So effectively, Ilan, as, I think what Eleos is trying to be is the operating system-
• 17:52
  Alon: Correct
• 17:52
  John: increasingly for at least part of the workflow.
• 17:56
  Alon: Correct.
• 17:56
  Alon: Every, you know, one area that, that is very sacred, and, and we don't intend to touch, that's the core clinical interaction between the patient and the clinician.
• 18:07
  Alon: But when you think about where healthcare is, healthcare is probably the only, one of the only sectors in our economy that with, with time became less and less productive, and less and less effective, right?
• 18:21
  Alon: Although there was huge introductions of new technologies.
• 18:25
  Alon: And so we believe that AI and Eleos can help in everything that's happening around this encounter, and streamline a lot of these workflows.
• 18:35
  Alon: And at the end of the day, allow for more time to care and more access to care for a population that need it now more than ever.
• 18:45
  John: So is, is the right way to look at Eleos is it's, is it's a workflow optimization tool that increasingly is starting to get into some of the things that drive revenues and the ability to bill?
• 18:58
  John: Yes.
• 18:58
  John: Eleos is the AI operating
• 19:00
  Alon: system that moves all of these manual workflows into the AI era, and in the process it improves outcomes, it improves access, it improves margin and financial outcomes for these organization, making them sustainable in this time of change, and really allow clinicians to focus on their patients.
• 19:19
  John: And are you wor- are you building it off of one of the frontier models in particular, like Gemini or Claude, or?
• 19:26
  Alon: It's, uh, it's always a combination.
• 19:28
  Alon: It's always a combination.
• 19:29
  Alon: We are probably one of the only companies that work multi-model, meaning we work from the audio directly.
• 19:37
  Alon: Uh, because in behavioral health there's a lot of emotion and a lot of content in, in, in the audio that, that you lose with text, right?
• 19:43
  Alon: Um, but today when, when you're deployed at such a large scale, um, it's, it's a con- you need to build an infrastructure that allows you to constantly adopt the best models.
• 19:56
  Alon: We also, by the way, John, I have to tell you, there's models that we developed in-house back in even 2020 that's still in production, that are really good, that you get really high margins on.
• 20:07
  Alon: Um, and not, not everything you need to, you know, run through all the frontiers models, because you also want this to be, um, economically viable- Well, it's expensive … for the providers.
• 20:18
  John: Yeah.
• 20:19
  Alon: No.
• 20:19
  Alon: Exactly.
• 20:19
  Alon: All those tokens
• 20:20
  John: start to add up.
• 20:22
  Alon: Exactly.
• 20:22
  Alon: And if you're a community-based nonprofit somewhere, you know- It's not like you're now, uh, uh, you don't necessarily wanna pay for all these tokens, right?
• 20:33
  Alon: You-- At the end of the day, you want the outcome that will move the needle for the organization.
• 20:38
  Alon: It doesn't really matter if you use this token or that token.
• 20:41
  Alon: S-
• 20:42
  John: so Alon, last question.
• 20:44
  John: You look into the future five years from now, where do you want-- what role do you want Eleos to play, and, and how is it playing it?
• 20:52
  Alon: I-- We want Eleos to be able to support communities to provide care to everyone, right?
• 20:59
  Alon: And I think the way to do this is you need to strip away all the administrative burden that is shackling them down.
• 21:07
  Alon: You need to enable them to do much more with less And if we are able to play a small part in that story, that means millions and millions of patients will get access to care that otherwise they wouldn't.
• 21:25
  Alon: And that's from an impact perspective, John, our role in this story, especially to get AI and to trickle down into the most, you know, sensitive and vulnerable s- um, levels of our society that sometimes, let's be honest, it doesn't get there, right?
• 21:44
  Alon: So we want-- I think that's, that's kind of our vision, and what we're working towards is allow these communities to care for every patient in need within their areas.
• 21:54
  John: That's it.
• 21:54
  John: That vision of going from a scarcity and pain to abundance and healing is a, is a really exciting one, Alon.
• 22:02
  John: So that's it for Care Talk.
• 22:04
  John: I'm John Driscoll, the chairman of the UConn Health System.
• 22:07
  John: If you liked what you heard or you didn't, we'd love you to subscribe on your favorite service.
• 22:12
  John: And Alon, thank you for giving us a little bit of hope for one of the hardest problems in healthcare, which is taking care of vulnerable populations with behavioral health.
• 22:21
  Alon: Thank you, John.
• 22:22
  Alon: Appreciate it.