Physicians Building With AI Suggest the Next Phase Is Bottom-Up, Not Vendor-Led
Anthropic’s profile of physicians building with Claude highlights a growing movement of clinician-developers shaping AI tools from inside care settings. The significance lies less in one model than in the broader shift toward doctors becoming workflow designers rather than just end users.
One of the most interesting developments in healthcare AI is the rise of physicians who are not merely evaluating tools, but actively building them. Anthropic's account of doctors using Claude suggests a change in where product intelligence lives. Instead of waiting for vendors to infer workflow needs, clinicians are increasingly prototyping solutions around the real bottlenecks they encounter every day.
This matters because many healthcare AI failures have been failures of translation. Technologies often look strong in demos but weak in practice because they are layered onto workflows without enough domain-specific nuance. Clinician-builders can narrow that gap by embedding operational realities—handoffs, documentation burdens, edge cases, time pressure—into the design process from the start.
Still, the bottom-up model introduces its own challenges. Homegrown tools may proliferate faster than governance structures, validation protocols, and cybersecurity review can keep up. What begins as productive experimentation can become another form of shadow AI if organizations do not create safe pathways for clinician-led development.
Even so, the direction is important. Healthcare AI is maturing from procurement of finished products toward co-development inside delivery systems. The winners may not simply be the vendors with the best models, but the organizations that best enable clinicians to turn local expertise into governed, scalable software.