Family Doctor Calls for Clear Rules on What AI Should Automate

A physician's dilemma

A 22-year-old woman came to Dr. Travis Bias complaining of lower back pain and requesting anti-inflammatory medication. During the visit, her responses grew evasive. When Bias stopped typing and asked how everything else was going, the real story emerged: she had been living in her car for three weeks, hadn't showered in a week, and was struggling to find work.

Bias, a family medicine physician and deputy chief medical officer at Solventum, prescribed the medication but knew it barely addressed her actual needs. The case manager at his facility was only available once weekly and already overextended. His patient needed stability and safety — resources the healthcare system wasn't equipped to provide efficiently.

That encounter, Bias writes in a commentary first published by Medscape, illustrates both the promise and peril of artificial intelligence in healthcare. AI could have flagged the patient's housing status before the visit, identified community resources, and freed up time for deeper conversation. But it also raises an urgent question healthcare has yet to answer systematically: which aspects of care should be automated, and which require irreplaceable human presence?

Why it matters

Healthcare organizations are implementing AI tools at scale without industry-wide consensus on where automation helps versus harms patient care. As corporate America experienced historic job losses in October 2025, healthcare faces similar disruption. Getting the boundaries wrong carries billion-dollar consequences in outcomes and workforce burnout — making this a strategic priority for health system leaders, not just a clinical question.

The augmentation argument

Bias advocates for a middle path between uncritical AI adoption and resistance to automation. He points to clear candidates for AI assistance: completing disability paperwork, EHR documentation, safety screening checkboxes, and identifying patient-specific community resources. These administrative tasks drain clinician time without requiring human judgment or empathy.

What must remain human, he argues, is what researcher Allison Pugh calls "connective labor" — the empathy and trust-building that makes patients feel heard. Primary care physicians often treat patients for decades, remembering family milestones and reading unspoken cues that signal deeper problems. That longitudinal relationship, Bias writes, often determines whether patients follow through on difficult treatment plans.

The risk of over-automation became visible when AI systems began refilling medications with no human clinician in the loop — a development Bias notes occurred despite ongoing concerns about chatbots occasionally hallucinating information or omitting critical details.

What healthcare leaders should do now

Bias calls for professional medical societies including the American Academy of Family Physicians, American Medical Association, and American Nurses Association to convene task forces linked to real-world pilots. These groups should map which clinical tasks require human connection and which can be safely augmented by AI, considering desired population health outcomes and available public health workforce.

For his patient living in her car, ideal care would have included time for extended discussion of housing options, health maintenance during transition, and exploration of resources — with space for natural human interaction. AI could augment that encounter by surfacing relevant information and handling documentation, but it cannot replace the moment when a physician stops typing, makes eye contact, and asks the question that reveals what's really wrong.

The details in this analysis were first reported by Medscape in a commentary by Travis Bias, DO, MPH, FAAFP.