A digital health nerd could have accurately termed summer 2025 the “Summer of Scribes” (though that coinage doesn't, necessarily, roll off the tongue). Then, Nabla raised $70 million in Series C funding; Ambience brought home $243 million. OpenEvidence secured $210 million in August. Finally, Abridge amassed a whopping $300 million in Series E, closing out an electric summer for ambient scribes.
A year later, much of that excitement has subsided, and, as model capabilities have improved, company products appear increasingly similar forcing, strategic convergence. The goal for each: expanding their suite of operations away from ambient scribing or chatbot services towards a holistic model servicing all stages of the care experience. Abridge has undercut OpenEvidence by partnering with Wolters Kluwer’s UpToDate; OpenEvidence wants to wipe Doximity with EHR-integration and patient take-homes; and Doximity keeps likening OpenEvidence to OpenAI , coyly intimating which role Doximity plays in the race (while, of course, Anthropic and OpenAI spar over healthcare themselves).
Differentiation is a struggle
The difficulty for each is differentiation. At a keynote address in June, Abridge announced its plan to provide support for physicians at all stages of service. Its Care Signals model, debuted with Kaiser Permanente in August, flags a patient’s chronic health conditions, identifies whether they’re discussed, cues physicians for appropriate assessments, and generates ICD-10 compliant documentation for post-visit. Its differentiator, as a Kaiser Permanente physician put it: “elegance.” That is, its model quality and its UX’s … well, “elegance.”
That “elegance” looks different for different startups. OpenEvidence, for example, focuses on diagnostic accuracy and supremacy and doggedly discredits evidence to the contrary.
For the health startups, the movement is towards EHR integrations, hospital contracts, and RCM management. The industry focus on a sort-of “holistic AI,” however, is bidirectional.
Cue Epic. Donning a straw hat, studded jeans, and cowboy boots, CEO Judy Faulkner ascended the rostrum at Epic’s UGM keynote in front of 10,000 baseball-jersey wearing, glow-stick toting attendees. Nothing about Epic’s approach, now under scrutiny by a coalition of state attorney generals for alleged anticompetitive practices, is “elegant.” Nor are its models: Faulkner announced Ergo AI, an AI infrastructure crowding out startups: clinicians can search medical literature with Epic’s Art AI, directly challenging OpenEvidence and Doximity; patients can converse with Emmie, undercutting Abridge’s pre-visit models; administrators can manage prior-auth with Penny.
Epic is betting on itself
The chase: Epic is selling better access to proprietary data; better native integrations; better access to chart summaries; better access to Cosmos data; better integration for EHR auto draft orders; better response with “Epic Rangers” IT experts embedded into hospital AI teams. Its strategy isn’t elegance, or accuracy, or eye popping trial data. Rather, it’s about capturing as much market share as possible: bundle the scribe, the prior-auth tool, Art, Emmie, and Ergo under one package, and Epic can render startups extraneous. Last August, for example, Epic, formerly Abridge's largest partner and shareholder, announced its plans to launch its own clinical scribe and sold off its shares in Abridge.
Patient risk enveloped by financial competition
Both the startups and Epic have identified the patient visit – pre, during, and post – as the terrain of this emerging pitched battle. The battle is over accuracy versus the ease of integration; it’s a battle of incumbency at the point-of-care. The startups are racing toward the EHR and the EHR is racing toward the startups. Read plainly: the destination they share is the patient.
It’s a battle over the patient experience, the physician, and the provider-patient interaction, the basic unit of healthcare. But that battle is monetary, not diagnostic. A terminated Mayo Clinic employee alleged that she was fired for escalating concerns about faulty AI because they “compromised Mayo’s competitive advantage.” That is the wrong paradigm: ECRI, which tracks patient safety across health systems, found that 9% of health systems had confirmed that AI errors affected care decisions directly. Already, patients in California are suing Sutter Health and Memorial Healthcare, alleging an ambient AI platform recorded their clinical conversations without their consent.
Now, as AI tools, previously ambient or integrated into EHR and RCM, converge upon the point-of-care, that percentage will, inevitably, increase. Healthcare systems must cut through the noise – the straw hats, the funding bonanzas, and the barbs – and prioritize patient safety, above all.
References
[3] Epic's UGM: https://www.statnews.com/2026/08/17/ugm-epic-confronts-challenges-succession-ai-strategy-antitrust/
[4] OpenEvidence: https://www.openevidence.com/announcements/openevidence-launches-patient-take-homes
[5] Doximity: https://www.healthtechnerds.com/p/weekly-health-tech-reads-8-9-26
[6] Joshua Liu's fabulous analysis: https://www.linkedin.com/posts/joshuapliu_judy-faulkner-made-big-announcements-at-epic-activity-7495684885588869120-zG57?utm_source=social_share_send&utm_medium=member_desktop_web&rcm=ACoAAENDH2gBcTFJAeqc0DJMfTvYke5r2nz61wI
.png)