Strip away the hype cycle and a quieter fact emerges: artificial intelligence is already standard equipment in American medicine. Not in trials. In Tuesday clinic.
The numbers come from Bessemer Venture Partners’ State of Health AI report: 67 percent of clinicians now use AI tools daily, over 90 percent weekly, and 92 percent of provider health systems were deploying, implementing or piloting AI scribes as of last measurement. As a physician, I find the scribe number the least surprising statistic in the report. Documentation is the part of medicine clinicians most want taken off their hands, and the part AI is best at today. For how the technology actually works, see our physician’s explainer on AI medical scribes.
The money is following the stethoscope
Health AI captured 55 percent of all health tech funding in 2025, up from 37 percent a year earlier, across 527 venture deals totaling roughly $14 billion. Average deal size rose 42 percent to $29.3 million. And the revenue clock in healthcare is doing what it did in the broader market, where AI startups compressed the road to $100 million ARR: health AI companies now reach $100 million in under five years, against ten or more for traditional healthcare software.
Bessemer’s standout archetype, what it calls X Factor companies, combines 6 to 10x annual growth with software-like margins above 70 percent. In healthcare, a sector where software margins historically went to die, that profile is genuinely new.
A clinical caution
One distinction matters more than any funding statistic: scribes and administrative agents succeed because they carry low clinical risk. The harder frontier, diagnostic and treatment decisions, still demands the evidence standard medicine applies to any intervention. Adoption curves for admin tools tell us little about how fast clinical AI should move, and conflating the two is how the sector will earn its next backlash. That frontier is now moving: Aidoc’s brain-bleed triage now drafts full radiology reports, pushing AI from flagging findings toward the diagnostic write-up itself.
What to watch
Watch whether the scribe wedge expands into orders, coding and prior authorization, where the ROI case is strongest, and whether any of the X Factor cohort files to go public in 2027 on the strength of these curves. The same overlay approach is reaching hospital wards abroad, where an NHS trust is piloting an AI infection-risk score on top of its existing electronic record. The category’s confidence is already visible in its marketing, as AI scribes begin advertising like pharma.
AI and business tech news, verified by a physician who reads the filings. One email a week, no noise.
