Giving AI a seat at the decision table does not automatically speed up professional judgment; it can actually stretch the deliberation cycle. The 2018 JAMA study of IBM Watson for Oncology showed that oncologists who received Watson’s ranked treatment options spent more time discussing each alternative, and in many cases chose a regimen that diverged from both the AI’s top pick and their usual practice.

The extra friction comes from a hidden incentive: clinicians treat the AI’s output as a “second opinion” that must be validated, not a shortcut. This validation step triggers a cascade—first, they retrieve the supporting literature; second, they reconcile the recommendation with patient preferences; third, they document the rationale to satisfy compliance audits.

Each loop adds minutes, but also deepens the clinician’s engagement with the case, sharpening their expertise rather than eroding it. The paradox is that the very presence of a powerful model forces humans to practice their core skill—critical appraisal—more rigorously, turning automation into a catalyst for expertise growth instead of a time‑saving crutch.