Penn Medicine’s Cook Says You Can’t ‘Set It and Forget It’ With Clinical AI

In Penn Medicine’s radiology department, Dr. Tessa Cook emphasizes that clinical AI deployments must be treated as ongoing practice transformation, not one‑off projects. Her team runs roughly 2.5 million imaging exams annually, so any AI tool is evaluated in a controlled, sequential manner. By staging pilots one after another, they capture detailed lessons learned, avoid the chaos of multiple concurrent pilots, and ensure each model is rigorously vetted before scaling. This systematic approach aligns with broader health‑system goals of safety, efficacy, and measurable ROI.

Cook also highlights the power of incremental change. Rather than overhauling infrastructure in a single, risky rollout, her group makes small, observable adjustments, monitors impact, and iterates. This tactic mitigates change fatigue among clinicians and IT staff, respects their day‑to‑day responsibilities, and builds confidence in new technology. The strategy dovetails with a strong emphasis on stakeholder engagement: clinicians, informatics specialists, and IS leaders are involved early, providing feedback that shapes the solution before it reaches production. Early buy‑in reduces wasted effort and prevents the classic failure of implementing tools that don’t fit real‑world workflows.

Beyond process, Cook stresses relationship‑driven communication. Monthly steering committees bring senior leaders together, but meetings are cancelled if there’s no substantive agenda, preserving trust and respecting busy schedules. Direct, timely emails or brief calls replace unnecessary gatherings, reinforcing a culture where time is valued. This blend of disciplined evaluation, incremental rollout, and thoughtful stakeholder interaction offers a replicable blueprint for health systems seeking to harness clinical AI without sacrificing operational stability or clinician satisfaction.