Asking 'Is AI good for medicine?' is the wrong question, and it is slowing real progress

A STAT News opinion piece argues that sweeping generalizations about AI in medicine, both positive and negative, obscure the more useful question of which specific tools work for which specific clinical problems. The author contends that this overgeneralization is actively delaying adoption of AI applications that already have solid evidence behind them.

Why it matters

The debate about whether AI is good or bad for medicine has become a substitute for the harder, more specific work of evaluation. When a hospital system delays a radiology AI tool with demonstrated diagnostic accuracy because leadership is still debating AI philosophy, that is not caution, that is a failure of decision-making infrastructure.

The real ask here is for leaders to build the capacity to evaluate tools on their own terms: what problem does it solve, what is the quality of the evidence, where does it fit in the workflow, and what does failure look like. That kind of structured thinking is harder than a policy stance, but it is the only approach that actually serves patients.

What to watch

Whether health systems begin publishing use-case-specific AI evaluation frameworks rather than continuing to issue broad AI governance policies that treat all tools as interchangeable.