Epic’s AI Bet, AMA’s New Guardrails, and athenahealth Leans into Interop

Plus fresh funding rounds and 4 Questions with Commons Clinic’s SVP of Care Services & Strategy.

AI has eaten healthcare and all the news - and funding - this week is centered around its march into the sector. All eyes were on the Epic UGM, where the company shared that it had more than 200 AI features in development. One of those - an AI assistant for clinicians called Art - can draft clinical notes. Art, its highly anticipated ambient scribing feature, was created in partnership with Microsoft (not many people in my network saw that one coming). In my opinion, by far the most important part of the announcement was Cosmos, its dataset built from tens of millions of de-identified patient records that simulates patient health timelines. It’s well worth skimming the actual paper. Yesterday, we hosted a webinar for Second Opinion subscribers to unpack all the Epic UGM news and announcements. If you missed it, you can [watch the full recording here].

While I’m intrigued by all of the AI in development for healthcare, I’m also wondering if we’ve gone too far in investing in software tools that sell into payers and providers. In my mind, the companies most poised to take advantage of generative AI are actually the clinic businesses that own their own stack. I am very bullish on AI-native care delivery, and I’m already starting to see companies reap the benefits of enormous efficiencies. That changes the unit economics of clinic businesses, and makes them far more scalable. Even more fascinating are the AI applications I’m starting to see in Life Sciences, and I truly believe we’ll experience an explosion of new drugs in the next few decades. 

Over the weekend, I plan to head to the UK for a week with the family (wish me luck and please send tips regarding two kids under 5 on a long flight). For next week, keep your eye out for a deep dive into the pros and cons of using SMS messaging for care delivery, and not just patient engagement. I truly believe a big company will be created here. Enjoy the rest of your summer and thank you so very much for reading. 

News Roundup: What you shouldn’t miss from the last week

With Annalisa Merelli 

Komodo Health launches first healthcare-native AI engine

The news: Marmot is the first AI engine to be native to healthcare, meaning it has health and pharma domain expertise encoded in the model, which can be applied to healthcare delivery and discovery of new drugs and treatments.

What’s different here?: Speed. There are plenty of AI models that can be applied to healthcare. Komodo Health is betting that tailoring its model to healthcare will make discovery and application much faster.

Who’s on board: Alnylam Pharmaceuticals has already agreed to embed Marmot in its operations to develop custom AI agents that can support its operations.

Our POV: The press release is a bit vague, noting that Alnylam Pharmaceuticals will embed Marmot into its functions. The devil is in the details. What functions exactly? How? What is the pricing model? I hope Komodo produces a more thorough case study on how this is going over the next six months, as I’d be intrigued to follow the methodology, the impact and the outcomes.

Ambiance is now on Epic

The news: Ambiance, an AI platform for visit documentation, is now part of the suite of software recommended by Epic in its toolbox, alongside Suki and Commure, which are recommended for voice-recognition note-taking. 

The plug: Ambiance’s tools are now available on Haiku, Epic’s app, so physicians don’t need to leave it to be able to record notes or get documentation immediately available in their portal. This integration greatly expands the reach of Ambiance’s AI services. 

What Ambiance said: "Clinicians now have those workflows available in the Haiku experience, which is a more ideal end-to-end workflow, so you never have to leave the Epic ecosystem," Ambiance co-founder Nikhil Buduma told Fierce Healthcare.

Our POV: All the large ambient scribing companies are now embedding more deeply with Epic, which is absolutely essential so the workflow is easy for providers. Epic is also releasing its own ambient scribe with Microsoft. I’ll be making it a point to watch how this shakes out over the next six months to see how providers react. 

CHAI is launching a new partnership to boost AI safety-net adoption

The news: The Coalition for Health AI (CHAI), a nonprofit that provides guidelines for AI adoption in healthcare, is partnering with the National Association of Community Health Centers (NACHC) to encourage safety-net health centers to adopt AI.

The deal: The partnership will develop specific standards to support low-resource providers, and provide training and education. 

The reach: This partnership would speed up adoption of AI in what is the country’s largest network of primary care providers, Healthcare Dive reports.  

Our POV: Public health experts are deeply concerned about AI algorithms leveraged by payers to deny care to Americans in need. This piece in Stat paints a particularly troubling picture. CHAI is taking this step at just the right time as these algorithms start to move from R&D to the real world. 

New AI toolkit from the AMA

The news: The American Medical Association (AMA), in collaboration with law and consulting firm Manatt Health, has released a new toolkit for use of AI in medicine. The toolkit aims to support organizations in drafting sound AI policy.

What’s exciting: “The transformative potential of AI to enhance diagnostic accuracy, personalize treatments, reduce administrative and documentation burden, and speed up advances in biomedical science,” says the AMA.

What’s concerning: Per the report: “AI's potential to worsen bias, increase privacy risks, introduce new liability issues, and offer seemingly convincing yet ultimately incorrect conclusions or recommendations that could affect patient care.”

Our POV: What stood out is really how many clinicians now are using AI in their practice, up a lot from 2024. It’s worth skimming through the report itself for insights on how they’re using AI, but to cut to the chase it’s really documentation and post-discharge instructions. Translation and summaries of medical research are also growing use cases. 

AI for clinical decision making from athenahealth

The news: athenahealth, a provider of EHR software for clinicians, is launching a new model on its platform athenaOne which will both ingest and contextualize data to support clinical decision-making. The model will introduce interoperability, helping real-time support among various healthcare players to more effectively share clinical decisions.

What athenahealth said: “Rapid advances in AI are now allowing us to reimagine the clinician and practice experience, solve previously unsolvable problems and bring back the human side of healthcare,” said athenahealth CEO Bob Segert.

What else is available: The company platform already includes real-time data exchange and longitudinal patient records, and will soon launch more services, including an AI assistant. 

Our POV: Athena has been out in front when it comes to interoperability. After years where it seemed like one step forward and two steps back, it’ll be fascinating to see more data get ingested into clinical systems and made vastly more relevant for clinical decision-making. And yet, despite all the attention on interoperability from the Trump Administration, big hurdles remain. 

 Interlude

"Finding ways to recruit more world-class talent as fast as possible is really, really important for us. If this is the legacy you want to leave, if you want to be a part of a company where every single day you can feel really good that you improved patient care, then we're going to resonate more than those horizontal technology companies." Abridge CEO and co-founder, Dr. Shiv Rao about what the company is looking for in its acquisition campaign. 

Deals and funding

THL Partners buys majority stakes in Headlands Research: The deal, reportedly worth $600 million, gives the private equity firm control over a US-based network of clinical trial sites, a sector in continuous growth.

Eight Sleep raises $100 million: The sleep tech startup raised this new funding round from investors including HSG, Valor Equity Partners, Founders Fund, Y Combinator and Formula 1 pilots. This brings the total capital raised by the startup to an estimated $260 million.

$21 million for Arintra: The AI medical coding platform raised funds to support its growth as it expands across health systems. The round was led by Peak XV Partners. 

DevelopHealth raised $14.3 million: The platform completed a series A funding round to support its AI model to automate prior authorization and benefits verification. The round, led by Wing Venture Capital, brings the company’s total capital to $17.6 million.

Cascala Health raises $8.6 million for AI in post-hospital care: The company’s co-investors include Flare and Eniac Ventures, and its goal is to use AI to summarize patient information and create an actionable transition plan. 

Watercooler (or whatever replaced it) chat

STAT published its annual list of the highest paid CEOs in healthcare, with a breakdown of their compensation. Would you have guessed the highest earner? 

4 Questions with the SVP, Care Services & Strategy of Commons Clinic

Benjamin Schwartz, MD, MBA of Commons Clinic

Digital health has been around for about 15 years, but there’s not much evidence on improving outcomes. Patients aren’t much healthier and costs are still going up. As a surgeon working in digital health, what do you think needs to happen next to fix that?

BS: The optimist would argue that prices would be even higher and patients even less healthy if it wasn't for digital health! Reality is obviously much more complicated. It's difficult to fix problems unless you understand and address the fundamental drivers of costs and overall health. So much of health is tied to lifestyle which is in turn tied to socioeconomic factors. We need to spend more time digging into the "why" — lack of access, social drivers of health, deterioration of the doctor-patient relationship — to make sure we're solving the right problems. The cost issue is driven by incumbency and the unintended consequences of healthcare policy, namely consolidation. The solution to the first problem is to have a willingness to go into the "dirty" areas of healthcare. We need healthcare companies that incorporate tech, not tech masquerading as healthcare. The answer to the second problem is not to wait around for sweeping reform. (We're already seeing that with direct contracting, employer consortiums, and entrepreneurial healthcare providers willing to embrace alternative models.) In summary, fix real problems, don't wait for policy changes.

You made the leap full time into Commons Clinic in an SVP role for care services and strategy. Can you talk more about what that role means? How is it different from say Chief Medical Officer (CMO)?

BS: SVP of Care Services & Strategy captures the dual nature of the role. To be honest, I enjoy all aspects of healthcare innovation, so it's a little from Column A (care delivery) and a little from Column B (strategy). My role as the first physician on the exec side is to serve as the bridge between the physicians/frontline care side of things and the executive/strategic side of things. Having spent 17 years in practice, I understand the day-to-day challenges and what the friction points are. At the same time, having developed an interest in healthcare innovation, the business of healthcare, and care model design, I also have a better understanding of operations and strategy. What Commons is doing is pretty ambitious and being able to bridge that gap goes beyond what a traditional CMO role might be. One day I might be providing feedback on our in-house AI tool, the next day joining our CEO on a call with investors or payer execs, and the following day discussing quality metrics and dashboards. Commons is both a brick-and-mortar care delivery company and healthcare startup. The role isn't quite a traditional health system CMO or tech-first CMO. It's intentionally different, and so are we!

Any thoughts on the recent policy changes we’re seeing from the Trump administration and the impact on digital health?

BS: I'm intrigued by "Make Health Tech Great Again." Pledging to meaningfully improve patient access to data, "kill the clipboard," leverage tech for chronic care management, etc. is a simple, straightforward step. (Commons signed the pledge.) Of course, talk is cheap. This gets back to Question 1 — was health tech ever great in improving outcomes and reducing costs? Many of the companies present at the announcement have been at it for a while and haven't made much headway. Meanwhile, if all the companies that signed on to data sharing follow through, that will be a huge coup for patients and for companies building in health tech. Fragmentation and lack of interoperability slow progress and lessen impact. This makes digital health solutions less effective. They either have to spend a lot of money integrating with different EMRs or risk being siloed. It seems like there's still a lot to be sorted out (for instance, participation is voluntary). That said, MHTGA is at least a step in the right direction. It's not clear yet how big that step is. There's plenty of opportunity for HHS and CMS to engage innovators, but it's a two-way street. It's not enough to sign a pledge, there has to be accountability.

What’s getting you most excited right now as a surgeon in terms of cool stuff that’s coming down the pike? Is it medical devices, ambient AI, VR or something else?

BS: AI. It's everywhere now, and it's hard to separate hype from reality. But AI will have a tremendous impact on how we deliver care so long as we stay grounded. In surgery, there's a lot to be excited about with AI. One of the most important questions is knowing when to operate. A technically proficient surgery done on the wrong patient may result in a poor outcome. We lack robust tools to guide those decisions. We rely on experience which is imperfect. Between 5-20% of knee replacement patients are dissatisfied with the outcome of their surgery, which is a pretty high number. It's a complex, poorly understood problem. The promise of AI is to tailor care to the individual in a way that's much more sophisticated. AI will also play a role in the OR. Historically, we were trained to replace all knees the same way. That works most of the time, but not all knees are the same. Knee anatomy and biomechanics are complex, but we've never had a great way to account for that complexity. Using AI to determine "when to cut" and "how to cut" may be the key to better outcomes. 

Want to support Second Opinion?

• 🌟 Leave a review for the Second Opinion Podcast
• 📧 Share this email with other friends in the healthcare space!
• 💵 Become a paid subscriber!
• 📢 Become a sponsor