Great. Welcome, everybody. My name's Scott Schoenhaus. I am the Healthcare IT, or Healthcare Technology, Analyst at KeyBanc. Pleasure to have here Veradigm Management Team, along with the newly acquired ScienceIO Management Team here. So first, I'll introduce everybody. We have Dr. Shih-Yin Ho, Interim CEO and Board Member; Lee Westerfield, Interim CFO; Jenny Gelinas, VP of Investor Relations; and finally, but not least, Will Manidis. I hope I got that pronunciation right. The co-founder of ScienceIO, which was recently acquired by Veradigm. So I'd like for you guys to introduce yourselves, maybe some background, maybe background on each of the companies for investors that are new or re-approaching your story, given all the changes. And then I'll start in the fireside chat, and then we can open it up to questions from investors.
Sure. Thank you, Scott. Thank you for including us today. So I'll get started. So my name is Shih-Yin Ho. I'm the Interim CEO, and I've been on the board of Veradigm for over a year now. My background is I am a physician by training, but I've actually spent the last 25 years building companies in the health information tech space, including EHRs from 20 years ago and all the way through evidence-based medicine, health information technology around comparative effectiveness, a lot of pharma work, and real-world evidence. I am pleased to be here today. We are really excited about what has happened. As the representative for Veradigm, it's important to know that Veradigm is actually a really exciting and very differentiated health information tech company in that we actually service three of the end markets of physicians, physician practices, payers, and life sciences.
At this moment in time, we have taken a very interesting and very big leap in a direction that actually is a natural extension of a long-term strategy of harnessing the power of the data that we collect in the organization. So really excited to talk a little bit more about that. But I'll let Lee give his introduction and also the others.
Yeah, Will and Jenny. Good afternoon. It's Lee Westerfield, Interim CFO. I joined in the executive ranks back in December, but I think there's probably three things I'd really want to highlight at this moment. One is that this company, as I've come aboard and sort of just health-checked it, as it were, that we have not filed, which is very disappointing. We have been able to validate the strength and profitability. We have net cash. We have profitability up and down the ranks of our business units, and we've provided guidance for 2024 to that effect as well. I think that's important because it makes us a rare exception of a reasonably healthy company going into a delist. Most are troubled; we're economically on solid footing. But by way of background, gosh, I've had a two-part career.
I've sat on Wall Street as an equity analyst for quite a number of years in the 1990s and 2000s, covering internet and media sectors. It gave me a good taste for advancing change, especially when it comes to business model dynamics in new technologies. I developed that into a fun career as a CFO for a number of different scaling technology companies, SaaS companies, over the last decade. Just profiling this particular one in the health technology space, alongside the investment that has been made in the acquisition of ScienceIO, we bring an existing repository of data, health data, and apply that with machine learning and AI. That presents a striking opportunity for, in this particular case, at a time in industry life, for healthcare to take advanced intelligence into a marketplace that's changing as well. Pretty exciting point.
That's been a recurring theme in my career.
Will, do you want to introduce yourself and a little bit maybe of background?
Yeah. Totally. It's good to meet you all. I'm Will, leading AI initiatives here. By background, I spent the bulk of my career at Foundation Medicine. We were a cancer diagnostics company that was commercializing the Flatiron Health data set. Built that into a great real-world evidence business, but the fundamental blocker of being able to scale that thing was clinical data was so incredibly difficult to work with, right? We had teams of hundreds of physicians and nurses sitting in an office, medical record on one screen, spreadsheet on the other, right? I didn't think we could build a fair and equitable healthcare system if the way we worked with data was click farms of humans. So we started ScienceIO in 2019 to build the world's safest language models to work with healthcare data. Scaled the business over a handful of years.
Then as we were looking out into the market to try to understand where the best healthcare data in the world was, we found it inside of Veradigm. It's our strong belief that models are a reflection of the data they're trained on. Veradigm's data is massive. It's super diverse, and it's really unique among its peers. Ultimately, it made the most sense to join Veradigm. Inside of Veradigm, we're focused on infusing these kind of intelligence products across the Veradigm product portfolio to switch from the products we have today to really being a health intelligence provider that serves our end segments better.
Great. Really quickly, my name is Jenny Gelinas. I started with the company back in 2017. I was heading up corporate FP&A. Then in 2021, I switched over and started also doing investor relations. Now I do that on a full-time basis, so.
Great. Thank you, everyone. So I guess let's kick off with the recent acquisition. What was the strategy? How are you thinking about the ScienceIO assets, applications for the vast datasets that you have? And Will just talked about it. Obviously, for people that are new to the story, Veradigm formerly was Allscripts, has a huge foothold in the ambulatory EHR market with one of the most robust datasets. Maybe just talk about the strategy of acquiring ScienceIO's technology and application on this vast dataset.
So one of the I'll take that question. So this is actually a really exciting moment. Part of it is because it's exactly as you said, we have been in the business of supporting physician practices for many decades, actually for at least a couple of decades. And we have, over time, gained a lot of a very rich clinical dataset. But that rich clinical dataset has never been necessarily structured in a way that could be useful for research or for life sciences or even to support our physicians and also our payers. And so if you think about the way our organization is set up, when I mentioned earlier that it's basically we have three end markets, we also basically represent that ecosystem.
Because we service that ecosystem, we needed a way to basically sort of liberate or structure that data in a manner that would actually be supportive of care, of closing payer gaps, and more importantly, be able to support the research enterprise around real-world evidence. But if you were to do it the way it has been done in the past, you would have needed to hire an army of people to abstract data because the abstraction has always been an extremely expensive, labor-intensive kind of cost. And our goal was not to go in that direction at all. Our goal was to kind of look for other solutions.
The timing sort of worked out where we basically came across ScienceIO as one of, as a part of a larger perspective around finding the right kind of tooling and the right kind of perspective on how to abstract data. Now, what's important here is that we acquired ScienceIO because it would allow us to basically own the large language model capabilities. That's really important because I know that a number of our competitors are working with rented, have made the choice to basically partner. We made the choice to acquire. We made the choice to acquire because it's extremely important on two fronts. Number one, for large language models that are specific to healthcare, you really want to control the inputs. And in controlling the inputs, you control the quality of what comes out of it.
You control the learning in a way that you make sure that with the right kind of human oversight, you will end up with much more accurate information and a much more accurate approach to looking at research and also a much more appropriate approach to actually developing AI. We felt very strongly that that was an important way to do that because we were sitting on the largest, one of the largest and richest clinical datasets. We had the data to train a model, but we would not have been able to gain those benefits had we basically rented a model. It was important for us to basically own it. What was really exciting is that when we met with ScienceIO, Will and I kind of connected immediately on the mission, on a very shared mission.
We had the mission to always continually be the best for our customers and basically always make good choices around how we think about technology and how it could basically help us as an organization. And Will's perspective also, and I'll let him speak a little bit more to it, also was this perspective that you cannot do healthcare specifically because healthcare is so idiosyncratic and has such peculiarities around its data. You couldn't treat it like any other data. And you had to be very careful and train on just very, very specific information so that you could ensure that it wouldn't affect a person down the line in their health, nor would it actually propagate something that was false through research. And it was that sort of thinking that brought the two of us together. And again, we had the data.
Will and his team had, in essence, the tooling, the thinking, the technology. We felt that it was important to bring it together. In a proprietary fashion, we have the ability to spin out products in a much faster way and also support our customers in our current product suite as well. I think Will could probably say a few words a little bit more to what I was saying.
Yeah. I would just say the last 10 years of healthcare has been a deep skepticism towards technology. And we're seeing that change now, right? Providers are begging for AI solutions to the point where they're often forward-running the ability to do these safely, right? Many large provider-focused organizations are taking large language models off the shelf that have never seen healthcare data and deploying them on patients where the bias of the data that those models are trained on is going to directly harm the patients, right? The progress we've seen in large language models in the past year has been staggering. But part of the reason we've made so much progress is we've fed these models junk data, right? Social media data, internet scrape data. The Pile, which you can Google, it's just a massive set of tasks that everyone's training on. It's just a mess.
We take a very different view. Providers need these tools. They're going to demand these tools. We better provide it to them safely, cheaply, and in ways that touch the products they already use. We think it's a very different tack of smaller models trained specifically on high-quality healthcare data. It's something we don't see anyone else doing. It's something that deeply unifies how we think about this at Veradigm.
Great. So I understand from a provider's perspective and the clinical needs and obviously the importance of accurate datasets to drive better clinical outcomes. But maybe Dr. Ho and Lee, can we talk about the solutions for extracting data using ScienceIO for the payer and the life science customers as well?
Sure. Yin, would you prefer?
Sure. Sure. Happy to. So, on the payer side, if you think about it, as value-based care continues to dominate more of the way we think about closing gaps around what is actually happening and what is actually being paid for, this is actually very heavily dependent on data and very heavily dependent on information being structured pretty quickly as well. So, I think what it's important is that you can improve the quality of gap closure when you actually are training models on what to look for. And it ends up being a labor-saving approach to doing something that today we do with actually most people will do with a lot of specialized labor and understanding what they're looking at. And I think that that's an important aspect of what you can do on the payer side. On the life science side, actually, it's very exciting here.
And that's partially because if you think about it, what are we dealing with when we are developing therapeutics and when we are expanding therapeutics and when we are looking for subpopulations that will be affected by it? We're looking for additional data that can support it. We don't always want to run a trial. We also want to be able to see if we can track what might be happening. And this is why this is why the FDA has been pushing really hard for the fact that not only is real-world data and real-world evidence appropriate as an additional sort of arm to a lot of research, but that it should also reflect the diversity of the patient population in order to really understand subpopulations. And that is actually where this golden moment is for us because we have actually quite a footprint across the United States.
We do not only have ambulatory EHRs inside of places, what we would call centers of excellence, or really very, very heavily concentrated urban areas. We also have EHRs in areas that are a little less urban, a little bit more suburban, a little bit more rural, and also with a greater diversity, which basically means we have an understanding of what is going on. And if we can basically take that and start to structure that information, we will be able to cut data in such a way that it will be possible to understand a disease condition by subpopulation, by demographics, by social determinants of health, and start to be able to enrich in some ways the way we've been thinking about research. And so this is actually very exciting for the biopharmaceutical space and for the life sciences.
This is the kind of data that we're looking for. It also allows us, by having these tools, to create data products, not just simply streams of data, but data products that are scoped in a way that meets what our customers are looking for, particularly if it's consumable in a research perspective. I think and Will also, similarly with when Foundation Medicine also had a similar perspective around what happens when you bring novel data to the market.
Yeah. I think we think about this as unlocking real-world evidence for everything that's not oncology, right? Real-world evidence has been an incredible success in oncology because it's one of the only conditions that drives so much cost on an individual patient basis that you can actually pay these huge farms of annotators to get it done. At Veradigm, we think about building these really high-quality, highly scalable data assets in every disease area because we can fundamentally drive the cost of abstraction so low with technology.
That makes perfect sense. So what have been the biggest surprises, both positive and negative or challenges maybe, over the first initial weeks, months of integration? Or what has been the most exciting application so far from both of your perspectives, I guess, since you really have just this is the nascent beginning of the part of the acquisition and the partnership?
Yeah. I would say I am pleasantly surprised at how educated the buyers are, particularly in pharma. As we're preselling data products in pharma, we thought this would be an LLM education game. Here's what models do. Here's how they work. And that's not at all what we're hearing. We're hearing incredibly educated buyers say, "I want models today. I want models that are not trained on social media data. And I want models that I can audit from all the way from training to production to inference." That is not something you would have heard in pharma even 12 weeks ago, right? But you have such demand that they are actually learning deeply about these technology solutions and being an educated buyer on the other end, right? That's great for us. It means the burden of proof is higher.
But we're the only ones in the world right now that can actually provide that full end-to-end audit trail from training to inference. So that's been a very pleasant surprise.
I think the other pleasant surprise is that there are so many places where we could be utilizing generative AI technology within even our own product suites and the way our customers are interacting and wanting to have their learning or have us basically gain more of a user experience around their preferences. I think that that's also been there are multiple places. And I think that I think when in the beginning, we were thinking it would only be a few places we would start. I think at this point in time, we are now starting to have to think about prioritization and pace because of the fact that the demand is much higher than we might have realized. But at the same time, that just makes it extremely exciting.
Yeah. And Dr. Ng, that kind of leads me to a follow-up question. And maybe this is one for you and Lee. But historically, I always thought of your business as kind of low single-digit growth on the provider side. There was some relative degree of churn in that business given there's just many more players in that field. And then the higher growth from payer and life sciences. Clearly, you have now more products to offer. So there should be more cross-selling opportunities given your ScienceIO acquisition. But then on the payer side, is it unreasonable to assume that we should see less churn given the augmentation of ScienceIO into your platform and basically creating more value for your provider clients?
Well, I think that potentially exists if you're looking from a business model point of view. I think that would be premature to say so in 2024 because we're developing the products and not yet deployed. But in principle and in the future, potentially so. I think I'd really be focusing on some other initiatives that we're undertaking in the provider business to improve margin and retention. They come under a banner headline vNext. But in principle, enhancements to the existing array of products, more on the Veradigm Pro side, but in collection, improved user interface functionality. Also, we acquired a business, Koha Health, adding to our revenue cycle management array. This collection of items I would put together as additional, I would say, to the existing product set for retention and for margin.
Maybe one of the examples about how you sort of keep a business going is you also make sure that you sort of meet them from a user experience perspective. One of the reasons Will and I are in Phoenix this week is because we're actually at our Veradigm user conference, user experience conference. And one of the things that's really been important if you think about how you can utilize generative AI is this area of the ambient scribe or ambient AI. It's kind of known as both. And what it really is, is at the end of the day, if you think about a physician and a patient, the encounter today oftentimes involves a computer in the middle where someone is entering information. Or sometimes you have somebody else entering the information. And it creates a lot of time and hassle.
It also sort of detracts from the relationship between the patient and the physician. If you start employing techniques like an Ambient AI or an Ambient Scribe, what you're simply doing is now capturing the information in a voice file. In doing that, you take the voice file and convert it almost immediately into a structured note. If you put a learning model in the mix behind it, what you start to do is improve that experience over and over and over again until a point that basically, from a user's point of view, it works almost as seamlessly as if they had basically been dictating it or seamlessly as if it just magically appeared. That's important because that basically says that we can use these types of tools to support a user experience.
A user experience-oriented way of developing software or developing products or basically even improving products does increase sort of the delight amongst customers. From our perspective, hopefully, it also increases everyone's happiness with us as well with our products. So I.
And ultimately, the value that you can place upon your products.
Absolutely. We save a lot of time. We make it an easier process. We make it an all of the different processes become sort of accretive to one another. And in the meantime, we're also actually structuring data even better.
Right. If we can sell. Yeah. Just keep getting better and better and better.
Exactly.
Great. Let's talk about you did set 2024 guidance recently. So what factors are driving your revenue outlook and profitability assumptions for this year, Lee and Dr. Ho?
Sure. Ng, do you mind if I take that, or do you?
Yes. Yes. It's yours.
Okay. Sure. Well, for 2024, very importantly, the business that's generating revenue, the businesses are legacy businesses, and they're resilient. We see some changes in the rate of churn. We also see some other growth areas. But I will say that the beneficial effect and margin impact, the revenue effect and the margin impact of our business from artificial intelligence and the applications that are in development, go-to-market in future, those will impact 2025 and 2026. So I think a really important thing for this audience to have in mind is that this year represents a year of investment. And I'll describe the areas of investment in just a moment that anticipate revenue growth accelerating and margin expansion as well in the years of 2025 and 2026 and conjecturally beyond that too. So why don't I focus on this year's investments?
There are really two areas that are crucial, one in growth and profitability initiatives and the second in foundational initiatives. In growth and profitability, we're here arrayed as funneled to talk about the biggest area of that investment. That's the advancements in our artificial intelligence. So we have some work to be done this year in that area we've been talking about. The other areas I mentioned a few moments ago is enhancements and shoring up our resilience in EHR. Those are areas where we are focusing on, again, as I mentioned before, under the banner of vNext, which we'll refer to in future communications. Foundation. So let me speak openly. We've missed our deadline and have been delisted primarily because we were unable to present financials and file financials in a timely manner.
A key point there, we will be implementing new systems, ERP, revenue recognition elements, really company-wide system information improvements. That is a large project, a foundational project for us in an area of investment to improve business intelligence through the company and consistency in our filings, so internal and external productivity and transparency. That investment this year is a really big initiative. It will take place actually over more than one year, but predominantly this year. So this year is a year of investment. It will impact our margins in EBITDA, particularly. Then, as I say, we'll build back in 2025 and 2026 in both revenue acceleration and margin. I should say we expect to in those future years.
Great. Well, Dr. Ng, I'm not sure if you wanted to add to that, but we have several questions pouring in here. We have 10 minutes left. So if you don't mind, I'd like to address some of these investor questions.
We'll take one. That's fine.
Okay. So the first one is, could we see $1 billion revenues in the next two years?
What's the question?
I'm sorry. Could you repeat the question, Scott?
Yeah. Could we see $1 billion revenues in the next two years?
$1B? $1 billion?
Yeah. I think that's what they meant. I hate to put that I'm just reading the questions as they come in. $1 billion in revenues.
Well, if the question is cumulative for the period 2024 plus 2025, I would think that is a reasonable assumption. The addition of 2024 revenue and 2025 revenue, would that be in excess of $1 billion? Our guidance calls for more than $600 million in revenue this year. So I would anticipate more than $1 billion over 24 months. Again, we haven't put guidance out for 2025. So just adding two numbers together. But I think the question might be, will we have more than $1 billion worth of revenue in one year? Again, we haven't spoken to future years that much. But 10% CAGR does not get us to $1 billion per se.
Got it. Yeah. I'll move on to the next question. Where and when was large language model architecture developed? And how does that compare to the current state of the art in the industry? I agree that the data is the moat. And architecture is commoditized. But if the architecture doesn't work on your data, product timelines could be pushed out. Or even worse, customers won't like the quality. Maybe that's a question for Will and Dr. Ng.
Yeah. I would pull apart a couple of different threads here. I would say architecture has been a commodity for a handful of years now. Large language models are all built and constructed in the same way. What is not commodity at all is the things that happen in pretraining, right, data preparation, data curation, and dataset selection onwards. Various alignment techniques are also not commodity in the slightest, right? Things like DPO or PPO are very much still open questions in the literature. We've developed a large set of novel techniques that allow us to both train smaller models that are more aligned with provider preferences that we don't see in the literature whatsoever.
So I think if you think about kind of building differentiation in the large language model space over time, the things that will commodify are basic transformer architecture, the ability to procure billions of dollars of GPUs, and the ability to get bigger, right? That's the approach you see all the cloud providers and foundation model providers taking. What will not commodify is having profoundly novel data that is nonpublic, that others are not training on, in addition to access to workflows like we have through our provider life sciences and payer products that allow you to give rapid feedback and alignment to the models. So we have a bunch of novel techniques. We've been developing them over years. The vast majority of our most interesting work has occurred in the last year as technology compounds really rapidly. And given the Veradigm dataset, we can compound that faster.
Got it. One last question. Then we'll end it as we're coming up on time. This is a question. Feel free to answer it however you want, Dr. Ng and Lee. I'm just going to ask it as it comes in. When can we see Veradigm back on the NASDAQ?
Here's who we've got, may I?
Yeah. Please, Lee.
Thanks. Well, we've said publicly on this particular topic is that we're working diligently and collaboratively with our auditor. There's a lot of material that needs to be worked through. The review by the auditor is underway, active. I have a call in not too many minutes with our auditor on an array of topics associated with the periods that are being audited. There is a lot of material. It's quite the prior history of 2022 and backwards by a couple of years to 2020. And then there's the 2023 audit. And all of that's still underway. As to when, reliably, and I'll say this in a forthright manner to you, I would not pin a time and a date on that one because I think there's too far an array, short-term, long-term, medium-term potential timings.
But I will assure this part that we'll be coming forward with regularity to give you business updates and financial updates as one should as a public company. And so I know that's short of doing formal filings. And I really, really wish that I would be bringing you filings in a timely manner. But we'll be communicating with you in a timely manner.
Great. I think it's important to remember we are still a public company. And then the process of relisting is very mechanical once you have to pass restatements. And so I think that that's important to sort of recognize that the process of getting relisted after becoming compliant with restatements is simply going for a shareholder meeting and then basically working our way onto the exchange again.
I think that's helpful to understand too. Well, great. We'll end it there. Thank you guys so much for participating and getting us up to speed on ScienceIO. The integration and the potential opportunities are very exciting. Thank you, Will. Thank you, Dr. Ng. Thank you, Lee. Thank you, Jenny, for all attending.
Thank you, Scott. Thank you everyone. Thanks, Scott.
Thank you everyone.