The ROI of AI Agents: Lessons from a Real-World Deployment

Health systems are sitting on ordered care that never gets delivered. Screenings go uncompleted, follow-ups go unmade, and the staff hours to close those gaps simply don't exist. AI agents are changing what's possible, and the results are showing up in ways that go beyond a single metric: patients reached who would have fallen through, revenue from care that was already ordered, and clinical hours returned to the work only humans can do. Northwestern Medicine's lung cancer screening pilot is a concrete example of what this looks like in practice. AI agents contacted 1,705 patients with outstanding CT orders. 177 completed their screenings. Nine came back with positive findings. This session gets into what it took to stand up that pilot, what patient acceptance actually looked like, and how health system leaders should think about measuring impact when evaluating this technology. What You'll Learn How to evaluate AI agents for clinical outreach and pick the right first use case A practical framework for measuring impact across patient outcomes, revenue, and workforce capacity What patient acceptance of AI actually looks like, and why the assumptions are usually wrong Where this technology is ready, and where it isn't yet