From the stage: DocSF26

The Agentic Shift Meets Clinical Reality

DocSF26 ran under the theme "The Agentic Shift: Where Intelligent Systems Meet Clinical Reality in Musculoskeletal Care." Over two days in San Francisco, surgeons, investors, medtech leaders and startups kept returning to one question. Healthcare has no shortage of AI demos. Why do so few of them become part of the daily work?

XY.AI attended this event in October. Here are three lessons that stood out.

AI is moving from answering to acting

Fabrizio Billi of UCLA gave the room a simple model. Generative AI is a brilliant intern who has read every textbook. Agentic AI is the same intern with a login and a pager: it books the physical therapy visit, chases the prior authorization and calls the patient.

Peter Mulford, Chief AI Officer at BTS, explained why that matters now. Agents can work for hours without a person checking in, yet most organizations use a fraction of what AI can already do. His advice: be impressed by what AI can do, but never mystified by how it does it.

Pilots are easy. Production is the hard part.

Ken Trauner of Kaiser Permanente called the stretch between pilot and full rollout the "valley of death." Pilot users are early adopters who accept risk. The teams who roll a tool out to millions of members look for edge cases, support and economics.

Todd Gottula, CEO of Clarify Health, described how his company crossed. Clarify found that even obvious insights often went unused, so it stopped delivering reports and started doing the work itself. Hospital for Special Surgery saw the same pattern with ambient scribes: clinicians who used one for more than half a day kept using it, and a year in most reported about an 80% cut in documentation effort.

Buyers want the problem understood first

Keith Lavery of Zimmer Biomet told founders exactly what gets a pitch through the door. Show a shared understanding of the problem, then a solution that fits how the company already thinks about it. Every idea is tested against one filter: does it relieve pain and improve quality of life for patients?

Trauner added that projects fail when a vendor misses a stakeholder anywhere in the chain. DocSF founder Stefano Bini gave the sharpest example. At Kaiser, surgery is a cost center. At most health systems, it is a profit center. A pitch that doesn't change between the two has missed the point.

What this means for healthcare operators

The agentic shift is real, but the winners will be the teams that run AI on the Tuesday after the demo, when the queue has to clear whether or not anyone is watching. That takes three things the conference kept circling: agents that act, rules that keep them predictable, and a clear way to hand work back to a person when something looks wrong.

Start with a boring, high-volume task that already has a written process. Track how often staff accept the agent's work without editing it. As that number climbs, more work can run on its own.

That is the work XY does: we build and run AI agents for healthcare operations, from first review to full autonomy, with people in control. If you have a workflow stuck between pilot and production, we'd like to hear about it. Or, if you're unsure how to even get things underway, let's have a call to help. Book a consultative conversation with XY.