b.well: One of The Most Important Health AI Companies That Isn't Building AI
In January, we covered OpenAI's ChatGPT Health launch and the thorny question of what happens when LLMs start interpreting your medical records. What we didn't spend enough time on was the company quietly powering the data connectivity behind that launch: b.well Connected Health.
Since then, b.well has had a particularly busy quarter, something most partnership-hungry health-tech companies dream about. In February, they launched bailey — a white-label AI health assistant that organizations can embed directly into their own apps, handling everything from provider search to medication management to benefits navigation. In March, Samsung expanded a two-year partnership to turn Galaxy phones into portable health records, complete with QR code-based data sharing at the point of care — aligned with CMS's "Kill the Clipboard" initiative. Days later, Perplexity announced b.well as the infrastructure behind its own health data integration, bringing citation-backed AI search into contact with actual patient records.
Step back further and the pattern sharpens more. Google partnered with b.well last October. OpenAI in January. Samsung and Perplexity in March. Four of the biggest names in consumer technology, all independently arriving at the same conclusion: we don't want to build health data connectivity ourselves.
While the spotlight often stays fixed on the AI models themselves, b.well is making a compelling case that it's the connective tissue holding the whole thing together. Their platform aggregates clinical, claims, pharmacy, and wearable data from over 2.4 million providers and 350+ health plans, cleans it through a proprietary refinement process, and delivers it in AI-ready format. It's high-value and unglamorous. It is, however, the thing that determines whether your health AI interaction is grounded in reality or just a very articulate guess.
Let's go for a plumbing analogy - there's a strong case for the pipework mattering more than the faucet. But there's a flip side. The same company now acts as health data broker for four competing tech platforms — a level of concentration that's drawn remarkably little scrutiny. b.well's consumer consent model gives users per-platform control, and data separation is a stated priority. But when one infrastructure provider determines the quality and completeness of health data flowing into OpenAI, Google, Samsung, and Perplexity alike, the stakes extend well beyond any single partnership. Getting the plumbing right matters. So does asking who controls the pipes.
The Novo-Hims Saga: A Year of Breakups, Lawsuits, and Inevitable Reunion
The Novo Nordisk and Hims & Hers relationship has been one of healthcare's more dramatic on-again, off-again partnerships. A quick recap: in February 2025, the FDA declared the semaglutide shortage resolved, pulling the legal rug out from under the compounding market that Hims had built a significant portion of its weight-loss business on. By April, the two companies partnered anyway, offering branded Wegovy on the Hims platform at $599 a month. By June, Novo terminated the deal after less than two months, citing concerns about Hims continuing to sell compounded semaglutide alongside the branded product. Hims stock dropped 35%. An FDA warning letter followed in September over misleading marketing of compounded products.
Then came 2026. In February, Hims announced a compounded oral semaglutide pill. Novo sued. HHS referred Hims to the Department of Justice. The pill was pulled within days. Barely four weeks later, a settlement was reached. Novo dropped the lawsuit. Hims became an authorized distributor of branded Wegovy and Ozempic, agreed to stop advertising compounded semaglutide, and its stock surged 46% in a week.
What's notable is how little the final deal differs from the original April 2025 partnership - branded Wegovy on the Hims platform - except for one thing: Hims has now agreed to stop selling compounded alternatives alongside it. That was always the sticking point. Novo wasn't opposed to Hims as a distribution channel; it was opposed to Hims running a competing product through the same front door.
The less comfortable question is what got left behind. Compounded semaglutide was available through some providers for as little as $50 a month. Post-settlement, branded self-pay pricing sits between $149 and $349. That's dramatically cheaper than the old list price of $1,200+, but it's still a significant monthly expense, particularly when fewer than one in four large employers cover GLP-1s for weight loss.
The compounding market was messy and drew legitimate regulatory scrutiny. It was also, for many patients, the only affordable option. As the market consolidates around branded products and authorized platforms, it's worth asking whether affordability got negotiated out of the room. It’s a cleaner market now, sure, but the self-pay market is inherently blind to those it prices out.
ElevenLabs: The $11 Billion Voice That Healthcare Didn't Build, But Is Now Building For Healthcare
Let’s get this out of the way - ElevenLabs is not a healthcare company. It's a generative voice AI company that went from founding to $11 billion valuation in four years, primarily off the back of creative and enterprise use cases. Founded in 2022 by two Polish engineers, it built its reputation on generative voice AI for creators, gaming studios, and enterprise customers. Four years and $781 million in funding later, it's valued at an eyewatering $11 billion, having more than tripled in twelve months, with $330 million in annual recurring revenue and clients including Meta, Salesforce, and Epic Games. It is, by any measure, one of the fastest-scaling AI companies we have.
It also has a quietly substantial healthcare footprint. Its voice agents integrate with Epic, Cerner, and Athenahealth, offering HIPAA-compliant patient triage, appointment scheduling, and prescription management across 32 languages. Healthcare systems are using them to handle after-hours calls, route urgent cases, and automate the kind of high-volume administrative work that burns out front-desk staff. None of this is the company's headline product - that's sort of what makes it worth looking into.
But the more interesting story might be what hasn't been applied to healthcare yet. ElevenLabs' dubbing technology can translate a video into 29 languages while preserving the original speaker's voice, tone, and pacing. Now apply that to a surgeon's post-operative instructions, or a discharge education video, or a clinical trial consent walkthrough — instantly localised, still in the clinician's own voice, without re-recording a thing. For diverse patient populations, that's an access unlock that currently either costs a fortune in human translation or, more commonly, just doesn't happen.
Then there's voice cloning. The headline application is the company's "1 Million Voices" initiative, which involves restoring speech for people who've lost it to ALS, stroke, or cancer, using as little as a voicemail or an old home video. That alone would justify attention. But the same technology could power automated follow-up calls or medication reminders in a patient's own clinician's voice, rather than a generic system tone. Trust and adherence are partly relational: hearing "your" doctor matters, even when the message is routine. The flip side is obvious: voice technology this convincing in a clinical context carries real risk. If a patient can't tell whether they're hearing their actual clinician or an AI reproduction, that's a trust problem dressed up as a feature.