The edge in AI and health won't be the model but who understands the real problems of patients and scientists

By American Enterprise Institute (AEI) · November 11, 2026.

Researcher David Shaywitz, an AEI adjunct fellow, publishes a commentary, originally appearing in the Timmerman Report, after taking part in several forums of investors, executives and analysts on what AI will do in healthcare and the life sciences. His thesis: in the midst of a technological “frenzy phase” (the term is economist Carlota Perez's), the most valuable voices are those that combine enthusiasm with humility in the face of the complexity of biology and healthcare.

As examples, Shaywitz cites several conversations. At the Deerfield/GV AI Forum he moderated a session with Daphne Koller (founder and CEO of insitro) and Rosana Kapeller (GV adviser and founding chief scientific officer of Nimbus, whose TYK2 program was sold to Takeda in 2023 for $4 billion upfront and up to $2 billion more in milestones, according to the author). Both agreed that the decisive asset of future companies will be the quality of the team, its R&D experience and its ability to work together. Koller rejected, according to the author, the idea of a “pipeline in a person”: she prefers diverse teams in which biologists and engineers respect each other.

The author also highlights Nate Gross, head of health at OpenAI and co-founder of Doximity. He recounts that Doximity wanted to eliminate the fax, but physicians pointed out that secure messaging was of little use if their colleagues were not on the platform; the company ended up letting users send and sign faxes from their phones. Gross, according to Shaywitz, talks about AI as a “super-assistant” that accompanies patients between appointments: understanding and remembering instructions, navigating a fragmented system and adapting the medical plan to their real lives.

The piece contrasts that approach with the predictions of the frontier labs. Shaywitz quotes a line attributed to Anthropic's Dario Amodei, who believes it is possible to cure most human diseases in about 5-10 years, and another from OpenAI's Sam Altman, who speaks of cures “at an unprecedented rate.” He also mentions a Wall Street Journal report according to which pharmaceutical CEOs do not believe Silicon Valley's promises. That is the view of the author and of that outlet; we have not verified the report.

Our take. We share the horizon: in the long run, AI can bring us closer to eradicating diseases and extending healthy lives. But Shaywitz's piece underscores something often forgotten: the bottleneck is not only computational. The human data we need most are the scarcest, and biology remains messy. Promising cures within a decade helps attract capital and attention; building the teams, data and clinical trust that make them possible takes much longer.

There is also a competitive consequence. If models become cheaper and accessible to everyone, as is already happening with the open frontier, the difference will be made by culture and domain knowledge: teams where engineers and clinicians have equal status, rather than a sidelined “squad of computer scientists,” as the author describes happening at some big pharmaceutical companies. That is good news for those who know how to listen to patients and doctors, and bad news for those who think a good model is enough.

One caveat is also in order. This is an opinion piece, based on the author's conversations and observations, not a data-driven study; and Gross's vision of a “super-assistant” is, for now, an aspiration, not a proven product. But it points in the right direction: the value of AI in health will probably show up less in grandiose announcements and more in closing the everyday gaps in care, between one appointment and the next. In the short term, that transition will demand patience and realism; in the long term, it is the path toward the health abundance we advocate.