01 /Physician Adoption

The numbers are striking. The American Medical Association's 2026 Physician Survey on Augmented Intelligence, published March 12, found that 81% of physicians now use AI in their daily practice — more than double the 38% recorded just three years ago. The average number of AI use cases per physician has also climbed, from 1.1 to 2.3, suggesting that adoption is deepening, not just widening.

81%

of physicians now use AI in practice, up from 38% in 2023

88%

are concerned AI could cause them to lose clinical skills over time

75%

of AI users report reduced admin burden and better job satisfaction

And yet the trust gap persists. A Doximity survey published the same week found that more than 70% of physicians still cite accuracy and reliability as their top barrier to AI adoption. Nearly half flagged legal and regulatory uncertainty as an additional concern. Even among enthusiastic adopters, a remarkable 88% worry that leaning on AI tools could cause them to lose hard-won clinical skills over time — a concern especially pronounced among doctors within their first decade of practice.

The picture that emerges is nuanced. AI is already embedded in medicine at scale — and the benefits are real, with three-quarters of adopters reporting lighter administrative load and nearly half saying they can take on more patients. But the field has not yet solved for the deeper question of how clinicians learn to trust a tool whose reasoning they often cannot see. That challenge is less technical than it is human — and 2026 may be the year the industry is forced to take it seriously.

02 /Healthcare Robotics

At its annual GTC conference in San Jose, Nvidia this week unveiled an ambitious set of healthcare robotics initiatives built around what CEO Jensen Huang is calling the "agentic AI" era — systems capable of acting autonomously toward defined goals, not just responding to prompts. For hospitals, the practical implications are significant and immediate.

The company launched three new tools under a dedicated physical AI platform for healthcare. Open-H is a dataset comprising over 700 hours of surgical procedure video, designed to train robotic systems on the real diversity and complexity of operating room environments. Cosmos-H is an open model family that generates physics-based synthetic training data at scale, allowing developers to simulate surgical scenarios without requiring access to rare or dangerous real-world footage. And GR00T-H is a vision-language-action model — trained on Open-H — capable of receiving plain-language clinical commands and executing complex physical tasks in healthcare settings.

Perhaps the most striking announcement was Rheo, a blueprint for building a complete hospital digital twin — a simulated version of a clinical environment that models workflows, medical device interactions, human movement, and logistics simultaneously. The technology would allow hospitals to test and optimize AI-driven processes in simulation before rolling them out to actual patients and staff, a step that could materially reduce both implementation risk and the governance anxiety that currently slows clinical AI adoption.

03 /Global Health Equity

One of the sharpest criticisms of healthcare AI is that its benefits flow overwhelmingly to wealthy, well-resourced health systems — widening the gap between patients who have access to cutting-edge medicine and those who don't. A major new initiative announced earlier this year is a direct attempt to push against that pattern.

In January 2026, the Bill & Melinda Gates Foundation and OpenAI announced a joint $50 million commitment in funding, technology, and technical support to build AI health capacity across African countries. The initiative begins in Rwanda, with a target of reaching 1,000 primary healthcare clinics by 2028. The practical focus is on the kinds of use cases where AI can have immediate, tangible impact in under-resourced settings: clinical decision support for frontline health workers, diagnostic assistance in areas without specialist access, and tools designed to work within the connectivity and infrastructure constraints common outside major urban centers.

The announcement has been welcomed by global health experts, though it also carries familiar cautions. WHO Regional Director for Europe Hans Kluge has warned broadly that without "clear strategies, data privacy, legal guardrails, and investment in AI literacy, we risk deepening inequities rather than reducing them." The Gates-OpenAI initiative will be closely watched as a test of whether that warning can be proactively answered — or whether it becomes another case where good intentions outpace implementation.

Sources

Week of March 28, 2026