Bairo GonzalezLeandro · Martinez
Start your guided journey The story

In the world · Artificial intelligence · 13 April 2026

AI in health in 2026: what the AI Index measured

The AI Index 2026 saw doctors spend up to 83% less time on notes thanks to AI, but only 5% of clinical AI studies use real patient data.

bairogonzalez.com team, drawing on Bairo's story · Published

The AI Index 2026, Stanford University's annual report published in April 2026, recorded that tools that automatically write clinical notes from the consultation spread in 2025: in several hospital systems, doctors reported up to 83% less time spent writing notes and less burnout. The same report issues a warning: in a review of more than 500 clinical AI studies, only 5% used real clinical data.

What happened

The AI Index is published by Stanford's Institute for Human-Centered AI (HAI). In the 2026 edition, health appears with two sides, according to the report's highlights:

  • What already works day to day. Tools that turn the consultation conversation into a clinical note saw broad adoption in 2025. Doctors at several hospitals reported up to 83% less time on paperwork and a significant reduction in burnout.
  • What is still a promise. Research on digital twins, computational representations of a patient, went from almost zero in 2015 to 372 publications in 2025.
  • What remains to be proven. In a review of more than 500 clinical AI studies, almost half tested the systems with exam-style questions rather than patient data. Only 5% used real clinical data.

IEEE Spectrum, the magazine of the Institute of Electrical and Electronics Engineers, highlighted another trend in the same report: the number of publications on AI in drug discovery more than doubled in two years, and research on biomedical AI that reads medical images together with text is growing.

Why it matters

The numbers show where AI helps today and where it has not yet proven what it promises. The most solid gain is in a task the patient never even sees: taking hours of typing off the doctor and giving that time back to listening. It is backstage help, and perhaps that is why it works well.

The 5% warning is just as important. A system that gets exam questions right does not necessarily get it right in front of a real patient, with an incomplete history, confusing symptoms and fear. That is why the right question, when facing any AI health tool, is not "is it intelligent?" but "was it tested with real people, and who is accountable for the decision?".

In Brazil, that second question already has an answer in regulation. CFM Resolution 2,454/2026, in force since August 26, 2026, treats AI as support and leaves the final decision with the physician, according to the Federal Council of Medicine.

In Bairo's view

For Bairo Leandro Gonzalez Martinez, the 83% of time given back is the best picture of what he means when he calls AI an extension of the human. In his reading, the machine does well the work that keeps the professional away from the patient, and the professional goes back to doing what only they can do: look, listen and decide.

He also reads the 5% as a warning for those who build. For Bairo, in health AI must be a clinical assistant, never a substitute, and every tool needs validation by a professional. That is why he bets on niche AIs: an assistant that handles one bounded step, such as organizing the history before the consultation, is easier to test with real data and to audit than a system that tries to do everything. CLAIN, the agent orchestrator he founded, follows this design, with the decision always going back to a person. The goal of 142 niche AIs in 142 days, announced on October 1, 2026, is a goal.

Where this meets the ecosystem

  • Hospital Triage, under development, proposes triage with a clinical assistant that collects information and follows the daily history and, when needed, schedules a teleconsultation with a doctor or nurse, who validates what the agent organized.
  • Xperienc Global Labs proposes international teleconsultation as a clinical second opinion, not a diagnostic report, and private records in the user's dashboard.
  • Hospital Granja Julieta is a partner of Xperienc and a distinct entity. Bairo is the founder of Xperienc, not of the hospital.

None of these ventures promises a clinical outcome. What they propose is to use AI where it has already shown it helps and to test carefully where proof is still lacking.

Sources

Read also

More on this topic

In the world: Artificial intelligence