theCUBE + NYSE Wired: MedTech Unplugged, the Future of AI in Healthcare & Life Sciences | Salonia Brown Mount Sinai Health System & Akash Magoon, Adonis

Salonia Brown Mount Sinai Health System & Akash Magoon, Adonis

This conversation examines the application of artificial intelligence to the healthcare revenue cycle, focusing on automating denials, voice agents and operational readiness. It explores practical use cases for accounts payable, claims denials and revenue operations, and addresses interoperability with major electronic health record systems. Salonia Brown of Mount Sinai Health System, system vice president of revenue cycle, and Akash Magoon of Adonis, co-founder and chief executive officer, join Gemma Allen of theCUBE and the theCUBE team at the NYSE Wired studio for a MedTech Unplugged discussion. Brown explains revenue cycle responsibilities and the need for governance and operational owners; they emphasize dedicated operational ownership for successful deployments. Magoon outlines Adonis's AI-native denial automation, analytics and voice AI that engages payers and supports payer outreach. Key takeaways emphasize operational readiness and measurable return on investment. Brown recommends clear ownership, governance and an initial three- to six-month runway with continuous iteration. Magoon highlights targeted vertical automation that reduces manual work on denied claims and employs voice agents to interact with payers, delivering efficiency and organizational support. Demonstrating cost savings to chief financial officers remains essential for adoption. This discussion provides actionable guidance for revenue cycle leaders, health technology vendors and finance executives seeking measurable ROI from AI deployments in denial management and medical billing.

Share this session

Clips
More from theCUBE + NYSE Wired: MedTech Unplugged, the Future of AI in Healthcare & Life Sciences

Gemma Allen

Host, theCUBE + NYSE Wired

theCUBE

Salonia Brown

System Vice President of Revenue Cycle

Mount Sinai Health System

Akash Magoon

Co-Founder & CEO

Adonis

play_circle_outline Healthcare revenue cycle as vital hospital cash flow: billing, reimbursement, and collections.

play_circle_outline Predictive Analytics for Payer Behavior and Claim Optimization Across Dozens of Insurers and Hundreds of Contracts

play_circle_outline Change management: tribal knowledge, human-in-the-loop governance, ownership, and training.

play_circle_outline Adonis AI-Native Voice Agents Automating Healthcare Claims: Denial Detection/Appeals, Secure Data Retention, AP/AR, Agent Interactions, 80/20

play_circle_outline Vertical focus strategy enables deep productization and network effects across hospitals.

Info

Salonia Brown Mount Sinai Health System & Akash Magoon, Adonis

Gemma Allen

Host, theCUBE + NYSE Wired theCUBE

HOST

Salonia Brown

System Vice President of Revenue Cycle Mount Sinai Health System

Akash Magoon

Co-Founder & CEO Adonis

search

Gemma Allen

>> Palo Alto Studio Connection, Silicon Valley and Wall Street.

Akash Magoon

>> I'm John Furrier, hosting theCUBE here with Dave Vellante, my co-host.

Gemma Allen

>> Welcome back to theCUBE studio here at the New York Stock Exchange. I'm Gemma Allen, co-host of NYSE Wired, MedTech Unplugged. And today we are talking about how healthcare has an AI race happening in what might be the least glamorous place imaginable— accounts payable. Payers are automating, providers are automating, and everyone is trying to figure out how to get paid faster with fewer people chasing more complicated claims. Today we're going to unpack exactly what happens when AI moves from the demo room into the business of actually getting healthcare paid for. Joining me now to discuss that is Salonia Brown, System VP of Revenue Cycle at Mount Sinai, and Akash Magoon, co-founder and CEO of Adonis. Welcome,

Salonia Brown

>> guys.Thank you. Thank you for having us.

Gemma Allen

>> So I know I made a joke there about it being less glamorous than most of the world of AI, but it's certainly not a space we think about much, right, when we think about what is happening underneath the hood of AI across enterprises and organizations in America. But healthcare is complicated, and may I use an Irish term, convoluted space, okay? Especially as a European trying to understand the system here in the US. So, Salonia, I'm gonna start with you.

Salonia Brown

>> Sure.

Gemma Allen

>> Help me understand the world of the Mount Sinai revenue and billing and payment like what, help me understand what it is that you're dealing with.

Salonia Brown

>> Yeah. So first off, nothing that we do is isolated to Mount Sinai, right? So this is just healthcare in general. So within revenue cycle, it is probably the largest, most misunderstood department within the healthcare landscape, right? And when you think about revenue cycle and what we do, we actually provide cash flow for the entire organization. So all the money that we collect and the reimbursement from payers that funds everything from clinical care all the way up to philanthropy and, you know, charitable contributions as well. So it's vital and it's very important that what we do is efficient and it's sustainable. So when you think about it, in healthcare it really runs off of clinical excellence. However, unfortunately, the light bill runs off of cash.

Gemma Allen

>> Of course.

Salonia Brown

>> So it's vitally important that we collect what's due to us and what's contractually obligated. From our payers as well.

Gemma Allen

>> So Akash, talk to me about the decision to solve for this using an AI-native platform, right? Where did this lightbulb moment come from for you and the team at Adonis? help me understand what it is that you decided to set out to build and why.

Akash Magoon

>> Yeah, so me and my brother are co-founders at Adonis, and in our past lives when we were running our first company, we were working very closely with the insurance companies and we learned about how some of the biggest insurance companies in the country are using AI to automate the denial of claims. And so when patients are going to Mount Sinai, for example, and Sal's team is submitting claims to UnitedHealthcare, many of the insurance companies are incentivized to try to slow down the payment of cash and really propagate this game of cat and mouse. And so what we realized 4 years ago was that AI is in a really advantageous place to really understand the rules of the insurance companies and the changing rules of the insurance companies to help revenue cycle teams like Sal's understand what's going on and how you can go and actually submit claims in a much better way to get paid a lot faster.

Gemma Allen

>> So let's say you're working on Sal's team at Mount Sinai and you have 100 claims that you're chasing, right? Some of them could be $500,000, some of them could be $5,000, but it's not necessarily true that the $500,000 is actually more probabilistically valuable than the $5,000, right? Because the likelihood of maybe getting paid out on that might be lower. Am I correct in saying that there's an intelligence and understanding of where the time is best spent?

Akash Magoon

>> Of course.

Gemma Allen

>> Maybe help bring that to life for me a little bit.

Akash Magoon

>> I can start and then— Yeah, sure. To double-click into that even further, it becomes more complicated because there's dozens of insurance companies with hundreds of contracts. So every hospital will have a unique contract in place with the insurance company. And so whether or not that $500,000 claim gets paid, is determined on the unique contract that's in place, how the insurance company might adjudicate or deny that claim. And so what we've realized is that if you can build an analytics intelligence layer to help teams understand those changing behaviors, you can then submit claims in a way that you can get paid a lot faster. And so our reason for being as a company is we seek to partner with as many medical practices and hospitals across the country to help them understand, that playing field. And give them an upper hand, whereas payers for the longest time have always had the upper hand.

Gemma Allen

>> So I can't say I know a huge amount about healthcare and operations at a hospital, but in a previous life I did do some work with public sector at Microsoft. Right. So I understand that these companies and entities and organizations have evolved in complex ways because they haven't necessarily had the same level of IT spend that a Fortune 100 company has. Right. Things, you know, people kind of grow up and solutions grow up around problems in some respect. How easy is it to now inject AI into those workflows, those challenges? Like how off the shelf is a solution like this? Help us understand it.

Salonia Brown

>> So that's where you get into a bit of a pitfall. It is very easy to just inject AI. The problem with that is if you don't have your workflows down pat, if you don't have operational owners, when something kicks out, when you use AI, if there's no ownership over that or accountability, then what you're in essence doing is kind of digitizing dysfunction. So to answer your question, it's very simple and easy to inject AI, but to do it the right way, making sure there's governance, making sure that the economics make sense, right? So what are you trying to get out of the implementation of AI, that takes a little bit of time. So I would say anywhere from 3 to 6 months and then it's continuously evolving too. So as, you know, new use cases come up, you want to make sure that you have dotted all your i's and crossed all your t's as kickouts come up. So things that inevitably you did not plan for, you want to make sure that you have that covered as well. So I think a good 3 to 6 month runway to start. And then it's going to be continuously watching it evolving, ensuring that you have owners and you're taking it up to the next notch and next level as well. Yeah.

Gemma Allen

>> I guess talk about the readiness side of it, right? We know that the intelligence opportunity is there. You mentioned you might have one person who deals with 7 or 8 different insurance firms, 7 or 8 different types of policies or governance plans in place. How do you think about the structured data element to the ingestion?

Akash Magoon

>> Like, how do you bring all that to life so the readiness is there, I guess, Yeah, I would say you don't have to spend that much time online to realize that a lot of companies are building a lot really quickly with very few engineers using AI. And the actual building the solution is sometimes relatively straightforward. I think what has been the most challenging but also the most exciting part about partnering with hospitals is that teams like Sal's are several hundred people, if not thousands of people that have been doing work the same way for decades. And when you're bringing in a new AI workflow, the secret sauce isn't necessarily just building it, but figuring out where you're going to embed it into the human capital sort of chain of who's doing what and how our teams are operating with one another. And so we spent a good bulk of time trying to figure out, okay, how do we build something, but how do we actually deploy it and help organizations see ROI from that? Much of that comes with meeting teams where they are. And so hospitals have been using the same softwares for decades. Epic is an example of one, Cerner is an example of another. And so we spend a lot of time thinking about how do we interoperate with those existing systems and then go and automate from there.

Gemma Allen

>> And talk a little bit about context, right? Because we know that a lot of information and knowledge lives in folks' heads, right? Especially in organizations like hospitals where people have probably been there a long time and they— tribal knowledge is So

Gemma Allen

>> real.

Gemma Allen

>> Yep.And I think sometimes, especially in the world of September 2026, we underestimate the power of tribal knowledge too, right? And how difficult it is to un-tribal the

Salonia Brown

>> knowledge.Right, right,

Gemma Allen

>> right.Talk a little bit about that, Sal. You go to roll out an AI-native system like this, you have a team, I'm sure it's very large. I'm sure some folks are enthusiastic, some maybe not. Help me understand how you figure that part

Salonia Brown

>> out.Yeah, so I just want to touch on something that Akash just mentioned. So you can have one payer with 50 different product lines, right? And all those product lines are treated totally different. They all have different rules. They have different edits, different complexities as well. So if you have one person who's working on 50 different variations, it can become problematic. So that's where, with the injection of technology, basically it enables us to be able to meet the payers where they are. Meet those demands, the requirements as well. And then when we started this journey, we made sure that we involved the teams, right? Because the teams are going to be working very, very closely with the technology. They have to trust it, right? We do not fundamentally use the mindset of set it and forget it, right? So autonomy still needs governance, right? So it's not immune from things that are errors. It's not immune from things that you didn't plan for. So we heavily engage and involve our operational teams to say, hey, what are the more complex things that we need the human judgment, the critical judgment versus those things that really you can automate and you don't need folks to do that. And they've been very appreciative of the automation. They've been very appreciative of, hey, this administrative burden, the red tape, right? No longer needing to do that. They actually want to work on more you know, higher yielding activities. And happy to say that's the journey we're on.

Gemma Allen

>> Well, that sounds fantastic from the perspective of your staff. I'm sure there's a world of exceptions, right? This world of understanding what you can and can't automate and what actually does need a human eye and maybe a human conversation. Okay. How do you think about that from the perspective of like a product build, right? You're obviously trying to build a product that is very multifunctional in terms of the governance layer, the compliance layer, all of the intelligent methods that go into understanding where a human can and cannot intervene. What about the security side of it? talk us through, Akash, what sorts of— what you hear around things like data retention, this world of agent talking to agent versus agent talking to human.

Akash Magoon

>> Of course.

Gemma Allen

>> What are you hearing? what are the challenges you see in the world of healthcare?

Akash Magoon

>> Yeah. To give the audience some context, in the world of claims, for every 100 claims that are submitted to an insurance company, Let's just say 80 of them are approved on first pass. So that's the happy path of claims. And then 20 of the 100 will get denied. And that's where teams are spending the bulk of their time is actually going and investigating why those 20 claims are denied, doing research on them, creating a rebuttal, and then going and submitting an appeal. And our reason for being as a company and why we were so excited about partnering with hospitals like Mount Sinai is we want to take those 20 claims that go through the unhappy path and reduce the human involvement in them. And historically, teams are touching every single claim in a very manual way. And our belief is that the human experts that are at Mount Sinai, for example, and across other hospitals throughout the country, they should be spending their time on that 1% of claims that require clinical judgment, require human guidance. And so we have been spending the last 4 years building AI automations that allow us to reappeal those claims in a way that require very little human touch. One example of that is insurance companies, and Sal can attest to this, insurance companies will gate a lot of things behind a phone call wall. So for those 20% of claims that are denied, you actually won't get that much information as to why. You have to call 1-800-Aetna, wait on hold for 30 minutes and facilitate a conversation. One of the things that we've built is a voice AI layer that will actually call insurance companies. We do so tens of thousands of times now on a weekly basis on behalf of our partners. And we'll actually host phone calls with the insurance companies with our voice AI. And that's one step in the puzzle that now is not on the plate of our customers or our partners. And so that's an example of where our AI agents sometimes are talking to an agent, a human agent on the other side. But now the insurance companies are getting smarter as well, and they're bringing AI agents on their side to have conversations with our agents. And so it's almost this game of cat and mouse, which has been pretty cool to see.

Salonia Brown

>> Listening to those calls, it's kind of scary, but it's amazing at the same time. One of the other things that I wanted to piggyback off of is some of the payers now, let's say you have 100 claims where you need to follow up on, they'll say, hey, we'll only do the first 3, and then you have to hang up, call back, do another 3, hang up, call back, do another 3. So by using technology, it enables us to deploy massive amounts of agents to make calls and figure out claim status and take the accounts to the next level.

Gemma Allen

>> In this world where agents are calling agents and algorithms are essentially fighting algorithms, right? What is the true, battle? How do you win that race? You think it's like constant iterations, like it's fully recursive. You're trying to understand all the time how policies and procedures and governance and compliance is changing on both ends of that phone line, right? Yeah, how recursive is that process?

Akash Magoon

>> I guess quick tangent. I recently saw a video where an agent was calling a Chinese restaurant agent to order takeout.

Gemma Allen

>> Yeah.

Akash Magoon

>> And they came up with their own language on the fly to make the ordering process a lot simpler. But jokes aside, I would say the insurers have many different levers to simplify things and no one actually wants to make the phone call. Being able to get the same information via API for example, or via their MCP server is much more straightforward. But the insurers know that insisting on the phone call is another friction step in the process. And so our goal is that, as our AI agents for voice calls get a lot better, for example, the insurer will just realize, hey, we don't really wanna play this game anymore. We'll just open up API endpoints. And so there's a couple different areas of revenue cycle where the insurance companies have built APIs and that makes things a lot simpler. And so our goal is that over time, we can get the insurance company to realize that it's in their best interest to play ball with us as well.

Gemma Allen

>> So I want to talk about the world of healthcare broadly, especially in 2026, September, where we hear so much hype around AI spend, the AI readiness race. Just before you guys came on the show, we had a full conversation on quantum encryption and what the world could look like in 2029 if we don't figure that out.

Akash Magoon

>> Right.

Gemma Allen

>> It is a scary and fascinating time out there. A lot is happening in healthcare. It's a legacy industry though, right? And like we spoke about earlier, you mentioned Epic and the world of CMS and feeding into those systems. We know that it hasn't necessarily had the same, I guess, speed and scale that Fortune 100s have had historically. What do you think, Sal, about the overall kind of spend on AI and the investment thesis around this world where are we going to head towards one orchestrator? What could it look like 10 years from now? Do you think the healthcare industry is kind of like rolling their eyes at this? What do you think is actually truly happening underneath the hood?

Salonia Brown

>> Yeah, so the healthcare industry is not rolling their eyes at it whatsoever. Think about it like, you remember when flat-screen TVs first came out and for like a Vizio 32-inch, it was like $7,000, right? However, as the technology got better, more efficient, cheaper, you can now get a TV for $200, $250. So that's where I think AI is headed. So right now it's very expensive. However, as it becomes more efficient, as it becomes cheaper, it's going to be more embedded in the technology and day-to-day operations that we handle today. When we talk about exceptions-based RCM model, historically what we've done is said, hey, here's a pile, right? In essence, a work queue. Sift through the work queue to find problems. In the future, what we're moving towards, that's no longer going to happen. The problems are going to find us. So those exceptions are going to go to the humans to then work and to make sure that we catch them further upstream. And that higher yielding work is what the humans really get pleasure out of and what they are really most satisfied with. But healthcare has definitely taken notice, especially within operations as well. I think agentic AI is here to stay and we have to figure out a way to create a sort of harmony or synchronicity with operations, the clinical side, the payers as well. We have AI agents. Payers also have AI agents as well. They talk. Some of them are— the calls are very successful. But this is only the beginning. Yeah.

Gemma Allen

>> And Akash, to you, I want to ask you, we have a lot of founders listening to the show. Some folks talk about this idea that they can be everything to everyone, right? In the world you're in, which I actually truly, really admire, it's more about a very vertically focused opportunity, right? You're not trying to be, you know, an Anthropic. You are very much operating in a niche space. How do you think about that from the competitive dynamics that are happening around healthcare and other industries where there is this, I guess, kind of looming existential conversation. Are we going to have one company manage everything 20 years from now?

Akash Magoon

>> Yeah, I would say over the course of building the company, we had the opportunity to think about do we want to go wide or do we want to go deep? And I think there's several companies in healthcare AI that are going a lot wider. And as a result, they have taken a very forward deployed approach where they still have a lot of humans servicing those clients, and it's very hard to productize things. And so we took the approach of, um, finding our swim lane and going deep on it. And so we see ourselves as a backend revenue cycle company, so everything that happens post-appointment effectively. And so we help teams on denials and AR management, and that allows us to go really deep on one area but also allows us to find overlapping learning lessons across all of our clients and then bring them back to all of our clients. And so many of our customers really appreciate the network effect that we have where the next hospital that we onboard is going to have a much better experience because we've partnered with, other hospitals throughout the country. And so being able to do that has allowed us to grow a lot faster. And so there's this old saying that fortune favors the focused, and we really see that being true in our category where, as Sal can attest, revenue cycle is a very humbling space. If you try to boil the ocean, it's going to take you down with it. But if you can stay focused and help achieve an ROI in one area, that'll allow you to develop the political capital, then expand over

Gemma Allen

>> time.Fortune favors the focused. I'm definitely going to use that. Okay, so folks, last question. September, back to school week. We're like 3 months out to Christmas. What does between now and 2027 look like? Where is the focus on your end? And also for you, Akash, and the team at Adonis, what's on the kind of top hit list between now and January 2027?

Salonia Brown

>> Yeah, so I would say continuing to be more efficient. Continuing to look at our use cases, continuing to add more use cases, looking at operations holistically, right? Where are there potential bottlenecks? Where— what are things that we can automate? What are things that we need to take a step further than automating and really adding some agent components, but really looking at all of our KPIs, making sure they're as crisp as can be and allowing us to continue to deliver outstanding care that Mount Sinai is known for.

Akash Magoon

>> For us, it's something similar to what you said a few moments ago, which is every organization spending a lot of money on AI. And I agree with Sal that there's going to be a Moore's Law where AI is going to be cheaper over time, just like the Vizio TV. For us, it's all about— between now and the end of the year, it's telling a louder story around even though AI is expensive at the moment, it's still fairly ROI positive and being able to tell that story and show the impact to a hospital CFO, for example, is going to be really important. For example, our CFO was yelling at me this morning about we're spending 50%— 50% of our spend is equivalent to our headcount spend. And so we're spending a lot of money. But if you can actually show that it's valuable and it's able to make each team member much more efficient, then our CFO is really happy about spending that money. And so similarly for our partners and our hospital customers, being able to show that same ROI story is something that we want to be able to improve upon between now and the end of the year?

Gemma Allen

>> Well, one phrase I used to use a lot when I was in product marketing at Microsoft was you have to spend money to make money. I think that was true 10, 15 years ago and it's true today. So today, thank you so much for joining us on NYSE Wired.

Akash Magoon

>> Thank you for having us. Thank you so much.

Gemma Allen

>> I'm Gemma Allen here at theCUBE Studio at the New York Stock Exchange. This is MedTech Unplugged, one of our programs with NYSE Wired. Thanks for watching.

Salonia Brown Mount Sinai Health System & Akash Magoon, Adonis

search

Gemma Allen

>> Palo Alto Studio Connection, Silicon Valley and Wall Street.

Akash Magoon

>> I'm John Furrier, hosting theCUBE here with Dave Vellante, my co-host.

Gemma Allen

>> Welcome back to theCUBE studio here at the New York Stock Exchange. I'm Gemma Allen, co-host of NYSE Wired, MedTech Unplugged. And today we are talking about how healthcare has an AI race happening in what might be the least glamorous place imaginable— accounts payable. Payers are automating, providers are automating, and everyone is trying to figure out how to get paid faster with fewer people chasing more complicated claims. Today we're going to unpack exactly what happens when AI moves from the demo room into the business of actually getting healthcare paid for. Joining me now to discuss that is Salonia Brown, System VP of Revenue Cycle at Mount Sinai, and Akash Magoon, co-founder and CEO of Adonis. Welcome,

Salonia Brown

>> guys.Thank you. Thank you for having us.

Gemma Allen

>> So I know I made a joke there about it being less glamorous than most of the world of AI, but it's certainly not a space we think about much, right, when we think about what is happening underneath the hood of AI across enterprises and organizations in America. But healthcare is complicated, and may I use an Irish term, convoluted space, okay? Especially as a European trying to understand the system here in the US. So, Salonia, I'm gonna start with you.

Salonia Brown

>> Sure.

Gemma Allen

>> Help me understand the world of the Mount Sinai revenue and billing and payment like what, help me understand what it is that you're dealing with.

Salonia Brown

>> Yeah. So first off, nothing that we do is isolated to Mount Sinai, right? So this is just healthcare in general. So within revenue cycle, it is probably the largest, most misunderstood department within the healthcare landscape, right? And when you think about revenue cycle and what we do, we actually provide cash flow for the entire organization. So all the money that we collect and the reimbursement from payers that funds everything from clinical care all the way up to philanthropy and, you know, charitable contributions as well. So it's vital and it's very important that what we do is efficient and it's sustainable. So when you think about it, in healthcare it really runs off of clinical excellence. However, unfortunately, the light bill runs off of cash.

Gemma Allen

>> Of course.

Salonia Brown

>> So it's vitally important that we collect what's due to us and what's contractually obligated. From our payers as well.

Gemma Allen

>> So Akash, talk to me about the decision to solve for this using an AI-native platform, right? Where did this lightbulb moment come from for you and the team at Adonis? help me understand what it is that you decided to set out to build and why.

Akash Magoon

>> Yeah, so me and my brother are co-founders at Adonis, and in our past lives when we were running our first company, we were working very closely with the insurance companies and we learned about how some of the biggest insurance companies in the country are using AI to automate the denial of claims. And so when patients are going to Mount Sinai, for example, and Sal's team is submitting claims to UnitedHealthcare, many of the insurance companies are incentivized to try to slow down the payment of cash and really propagate this game of cat and mouse. And so what we realized 4 years ago was that AI is in a really advantageous place to really understand the rules of the insurance companies and the changing rules of the insurance companies to help revenue cycle teams like Sal's understand what's going on and how you can go and actually submit claims in a much better way to get paid a lot faster.

Gemma Allen

>> So let's say you're working on Sal's team at Mount Sinai and you have 100 claims that you're chasing, right? Some of them could be $500,000, some of them could be $5,000, but it's not necessarily true that the $500,000 is actually more probabilistically valuable than the $5,000, right? Because the likelihood of maybe getting paid out on that might be lower. Am I correct in saying that there's an intelligence and understanding of where the time is best spent?

Akash Magoon

>> Of course.

Gemma Allen

>> Maybe help bring that to life for me a little bit.

Akash Magoon

>> I can start and then— Yeah, sure. To double-click into that even further, it becomes more complicated because there's dozens of insurance companies with hundreds of contracts. So every hospital will have a unique contract in place with the insurance company. And so whether or not that $500,000 claim gets paid, is determined on the unique contract that's in place, how the insurance company might adjudicate or deny that claim. And so what we've realized is that if you can build an analytics intelligence layer to help teams understand those changing behaviors, you can then submit claims in a way that you can get paid a lot faster. And so our reason for being as a company is we seek to partner with as many medical practices and hospitals across the country to help them understand, that playing field. And give them an upper hand, whereas payers for the longest time have always had the upper hand.

Gemma Allen

>> So I can't say I know a huge amount about healthcare and operations at a hospital, but in a previous life I did do some work with public sector at Microsoft. Right. So I understand that these companies and entities and organizations have evolved in complex ways because they haven't necessarily had the same level of IT spend that a Fortune 100 company has. Right. Things, you know, people kind of grow up and solutions grow up around problems in some respect. How easy is it to now inject AI into those workflows, those challenges? Like how off the shelf is a solution like this? Help us understand it.

Salonia Brown

>> So that's where you get into a bit of a pitfall. It is very easy to just inject AI. The problem with that is if you don't have your workflows down pat, if you don't have operational owners, when something kicks out, when you use AI, if there's no ownership over that or accountability, then what you're in essence doing is kind of digitizing dysfunction. So to answer your question, it's very simple and easy to inject AI, but to do it the right way, making sure there's governance, making sure that the economics make sense, right? So what are you trying to get out of the implementation of AI, that takes a little bit of time. So I would say anywhere from 3 to 6 months and then it's continuously evolving too. So as, you know, new use cases come up, you want to make sure that you have dotted all your i's and crossed all your t's as kickouts come up. So things that inevitably you did not plan for, you want to make sure that you have that covered as well. So I think a good 3 to 6 month runway to start. And then it's going to be continuously watching it evolving, ensuring that you have owners and you're taking it up to the next notch and next level as well. Yeah.

Gemma Allen

>> I guess talk about the readiness side of it, right? We know that the intelligence opportunity is there. You mentioned you might have one person who deals with 7 or 8 different insurance firms, 7 or 8 different types of policies or governance plans in place. How do you think about the structured data element to the ingestion?

Akash Magoon

>> Like, how do you bring all that to life so the readiness is there, I guess, Yeah, I would say you don't have to spend that much time online to realize that a lot of companies are building a lot really quickly with very few engineers using AI. And the actual building the solution is sometimes relatively straightforward. I think what has been the most challenging but also the most exciting part about partnering with hospitals is that teams like Sal's are several hundred people, if not thousands of people that have been doing work the same way for decades. And when you're bringing in a new AI workflow, the secret sauce isn't necessarily just building it, but figuring out where you're going to embed it into the human capital sort of chain of who's doing what and how our teams are operating with one another. And so we spent a good bulk of time trying to figure out, okay, how do we build something, but how do we actually deploy it and help organizations see ROI from that? Much of that comes with meeting teams where they are. And so hospitals have been using the same softwares for decades. Epic is an example of one, Cerner is an example of another. And so we spend a lot of time thinking about how do we interoperate with those existing systems and then go and automate from there.

Gemma Allen

>> And talk a little bit about context, right? Because we know that a lot of information and knowledge lives in folks' heads, right? Especially in organizations like hospitals where people have probably been there a long time and they— tribal knowledge is So

Gemma Allen

>> real.

Gemma Allen

>> Yep.And I think sometimes, especially in the world of September 2026, we underestimate the power of tribal knowledge too, right? And how difficult it is to un-tribal the

Salonia Brown

>> knowledge.Right, right,

Gemma Allen

>> right.Talk a little bit about that, Sal. You go to roll out an AI-native system like this, you have a team, I'm sure it's very large. I'm sure some folks are enthusiastic, some maybe not. Help me understand how you figure that part

Salonia Brown

>> out.Yeah, so I just want to touch on something that Akash just mentioned. So you can have one payer with 50 different product lines, right? And all those product lines are treated totally different. They all have different rules. They have different edits, different complexities as well. So if you have one person who's working on 50 different variations, it can become problematic. So that's where, with the injection of technology, basically it enables us to be able to meet the payers where they are. Meet those demands, the requirements as well. And then when we started this journey, we made sure that we involved the teams, right? Because the teams are going to be working very, very closely with the technology. They have to trust it, right? We do not fundamentally use the mindset of set it and forget it, right? So autonomy still needs governance, right? So it's not immune from things that are errors. It's not immune from things that you didn't plan for. So we heavily engage and involve our operational teams to say, hey, what are the more complex things that we need the human judgment, the critical judgment versus those things that really you can automate and you don't need folks to do that. And they've been very appreciative of the automation. They've been very appreciative of, hey, this administrative burden, the red tape, right? No longer needing to do that. They actually want to work on more you know, higher yielding activities. And happy to say that's the journey we're on.

Gemma Allen

>> Well, that sounds fantastic from the perspective of your staff. I'm sure there's a world of exceptions, right? This world of understanding what you can and can't automate and what actually does need a human eye and maybe a human conversation. Okay. How do you think about that from the perspective of like a product build, right? You're obviously trying to build a product that is very multifunctional in terms of the governance layer, the compliance layer, all of the intelligent methods that go into understanding where a human can and cannot intervene. What about the security side of it? talk us through, Akash, what sorts of— what you hear around things like data retention, this world of agent talking to agent versus agent talking to human.

Akash Magoon

>> Of course.

Gemma Allen

>> What are you hearing? what are the challenges you see in the world of healthcare?

Akash Magoon

>> Yeah. To give the audience some context, in the world of claims, for every 100 claims that are submitted to an insurance company, Let's just say 80 of them are approved on first pass. So that's the happy path of claims. And then 20 of the 100 will get denied. And that's where teams are spending the bulk of their time is actually going and investigating why those 20 claims are denied, doing research on them, creating a rebuttal, and then going and submitting an appeal. And our reason for being as a company and why we were so excited about partnering with hospitals like Mount Sinai is we want to take those 20 claims that go through the unhappy path and reduce the human involvement in them. And historically, teams are touching every single claim in a very manual way. And our belief is that the human experts that are at Mount Sinai, for example, and across other hospitals throughout the country, they should be spending their time on that 1% of claims that require clinical judgment, require human guidance. And so we have been spending the last 4 years building AI automations that allow us to reappeal those claims in a way that require very little human touch. One example of that is insurance companies, and Sal can attest to this, insurance companies will gate a lot of things behind a phone call wall. So for those 20% of claims that are denied, you actually won't get that much information as to why. You have to call 1-800-Aetna, wait on hold for 30 minutes and facilitate a conversation. One of the things that we've built is a voice AI layer that will actually call insurance companies. We do so tens of thousands of times now on a weekly basis on behalf of our partners. And we'll actually host phone calls with the insurance companies with our voice AI. And that's one step in the puzzle that now is not on the plate of our customers or our partners. And so that's an example of where our AI agents sometimes are talking to an agent, a human agent on the other side. But now the insurance companies are getting smarter as well, and they're bringing AI agents on their side to have conversations with our agents. And so it's almost this game of cat and mouse, which has been pretty cool to see.

Salonia Brown

>> Listening to those calls, it's kind of scary, but it's amazing at the same time. One of the other things that I wanted to piggyback off of is some of the payers now, let's say you have 100 claims where you need to follow up on, they'll say, hey, we'll only do the first 3, and then you have to hang up, call back, do another 3, hang up, call back, do another 3. So by using technology, it enables us to deploy massive amounts of agents to make calls and figure out claim status and take the accounts to the next level.

Gemma Allen

>> In this world where agents are calling agents and algorithms are essentially fighting algorithms, right? What is the true, battle? How do you win that race? You think it's like constant iterations, like it's fully recursive. You're trying to understand all the time how policies and procedures and governance and compliance is changing on both ends of that phone line, right? Yeah, how recursive is that process?

Akash Magoon

>> I guess quick tangent. I recently saw a video where an agent was calling a Chinese restaurant agent to order takeout.

Gemma Allen

>> Yeah.

Akash Magoon

>> And they came up with their own language on the fly to make the ordering process a lot simpler. But jokes aside, I would say the insurers have many different levers to simplify things and no one actually wants to make the phone call. Being able to get the same information via API for example, or via their MCP server is much more straightforward. But the insurers know that insisting on the phone call is another friction step in the process. And so our goal is that, as our AI agents for voice calls get a lot better, for example, the insurer will just realize, hey, we don't really wanna play this game anymore. We'll just open up API endpoints. And so there's a couple different areas of revenue cycle where the insurance companies have built APIs and that makes things a lot simpler. And so our goal is that over time, we can get the insurance company to realize that it's in their best interest to play ball with us as well.

Gemma Allen

>> So I want to talk about the world of healthcare broadly, especially in 2026, September, where we hear so much hype around AI spend, the AI readiness race. Just before you guys came on the show, we had a full conversation on quantum encryption and what the world could look like in 2029 if we don't figure that out.

Akash Magoon

>> Right.

Gemma Allen

>> It is a scary and fascinating time out there. A lot is happening in healthcare. It's a legacy industry though, right? And like we spoke about earlier, you mentioned Epic and the world of CMS and feeding into those systems. We know that it hasn't necessarily had the same, I guess, speed and scale that Fortune 100s have had historically. What do you think, Sal, about the overall kind of spend on AI and the investment thesis around this world where are we going to head towards one orchestrator? What could it look like 10 years from now? Do you think the healthcare industry is kind of like rolling their eyes at this? What do you think is actually truly happening underneath the hood?

Salonia Brown

>> Yeah, so the healthcare industry is not rolling their eyes at it whatsoever. Think about it like, you remember when flat-screen TVs first came out and for like a Vizio 32-inch, it was like $7,000, right? However, as the technology got better, more efficient, cheaper, you can now get a TV for $200, $250. So that's where I think AI is headed. So right now it's very expensive. However, as it becomes more efficient, as it becomes cheaper, it's going to be more embedded in the technology and day-to-day operations that we handle today. When we talk about exceptions-based RCM model, historically what we've done is said, hey, here's a pile, right? In essence, a work queue. Sift through the work queue to find problems. In the future, what we're moving towards, that's no longer going to happen. The problems are going to find us. So those exceptions are going to go to the humans to then work and to make sure that we catch them further upstream. And that higher yielding work is what the humans really get pleasure out of and what they are really most satisfied with. But healthcare has definitely taken notice, especially within operations as well. I think agentic AI is here to stay and we have to figure out a way to create a sort of harmony or synchronicity with operations, the clinical side, the payers as well. We have AI agents. Payers also have AI agents as well. They talk. Some of them are— the calls are very successful. But this is only the beginning. Yeah.

Gemma Allen

>> And Akash, to you, I want to ask you, we have a lot of founders listening to the show. Some folks talk about this idea that they can be everything to everyone, right? In the world you're in, which I actually truly, really admire, it's more about a very vertically focused opportunity, right? You're not trying to be, you know, an Anthropic. You are very much operating in a niche space. How do you think about that from the competitive dynamics that are happening around healthcare and other industries where there is this, I guess, kind of looming existential conversation. Are we going to have one company manage everything 20 years from now?

Akash Magoon

>> Yeah, I would say over the course of building the company, we had the opportunity to think about do we want to go wide or do we want to go deep? And I think there's several companies in healthcare AI that are going a lot wider. And as a result, they have taken a very forward deployed approach where they still have a lot of humans servicing those clients, and it's very hard to productize things. And so we took the approach of, um, finding our swim lane and going deep on it. And so we see ourselves as a backend revenue cycle company, so everything that happens post-appointment effectively. And so we help teams on denials and AR management, and that allows us to go really deep on one area but also allows us to find overlapping learning lessons across all of our clients and then bring them back to all of our clients. And so many of our customers really appreciate the network effect that we have where the next hospital that we onboard is going to have a much better experience because we've partnered with, other hospitals throughout the country. And so being able to do that has allowed us to grow a lot faster. And so there's this old saying that fortune favors the focused, and we really see that being true in our category where, as Sal can attest, revenue cycle is a very humbling space. If you try to boil the ocean, it's going to take you down with it. But if you can stay focused and help achieve an ROI in one area, that'll allow you to develop the political capital, then expand over

Gemma Allen

>> time.Fortune favors the focused. I'm definitely going to use that. Okay, so folks, last question. September, back to school week. We're like 3 months out to Christmas. What does between now and 2027 look like? Where is the focus on your end? And also for you, Akash, and the team at Adonis, what's on the kind of top hit list between now and January 2027?

Salonia Brown

>> Yeah, so I would say continuing to be more efficient. Continuing to look at our use cases, continuing to add more use cases, looking at operations holistically, right? Where are there potential bottlenecks? Where— what are things that we can automate? What are things that we need to take a step further than automating and really adding some agent components, but really looking at all of our KPIs, making sure they're as crisp as can be and allowing us to continue to deliver outstanding care that Mount Sinai is known for.

Akash Magoon

>> For us, it's something similar to what you said a few moments ago, which is every organization spending a lot of money on AI. And I agree with Sal that there's going to be a Moore's Law where AI is going to be cheaper over time, just like the Vizio TV. For us, it's all about— between now and the end of the year, it's telling a louder story around even though AI is expensive at the moment, it's still fairly ROI positive and being able to tell that story and show the impact to a hospital CFO, for example, is going to be really important. For example, our CFO was yelling at me this morning about we're spending 50%— 50% of our spend is equivalent to our headcount spend. And so we're spending a lot of money. But if you can actually show that it's valuable and it's able to make each team member much more efficient, then our CFO is really happy about spending that money. And so similarly for our partners and our hospital customers, being able to show that same ROI story is something that we want to be able to improve upon between now and the end of the year?

Gemma Allen

>> Well, one phrase I used to use a lot when I was in product marketing at Microsoft was you have to spend money to make money. I think that was true 10, 15 years ago and it's true today. So today, thank you so much for joining us on NYSE Wired.

Akash Magoon

>> Thank you for having us. Thank you so much.

Gemma Allen

>> I'm Gemma Allen here at theCUBE Studio at the New York Stock Exchange. This is MedTech Unplugged, one of our programs with NYSE Wired. Thanks for watching.