Anthropic and the Trust Moat: The Battle for Agentic AI in Health Care
2026
| Case No.
SM418
| Length
18 pgs.
In 2026, Anthropic faced a defining strategic choice in health care: deploy its Claude models directly into health systems as a product, or license its Constitutional AI safety architecture as infrastructure for other vendors to build on. Drawing on Stanford Health Care’s early experience with agentic AI tools, including an agentic tumor board, and Anthropic’s launch of Claude for Healthcare, the case traces a four-tier spectrum of agentic complexity in clinical AI, from simple chat completions to fully autonomous multi-agent systems. It examines the asymmetric stakes of getting AI wrong in a clinical environment, the governance and procurement dynamics that shape how health systems evaluate AI vendors, and the environmental and infrastructure costs of scaling AI deployment. Students are asked to weigh whether safety and governance can function as a durable competitive moat in enterprise health care AI, or whether it is a head start that rivals such as Google, Microsoft, Epic, and OpenAI will close as they take governance more seriously.
Learning Objective
Students learn to evaluate an enterprise AI vendor’s business model and revenue drivers, assess agentic AI use cases and their cost/ROI tradeoffs in health care, analyze whether safety and governance can serve as a durable competitive moat, and consider what governance structures health systems need to responsibly evaluate and adopt AI agents in clinical and administrative workflows.
This material is available for download by current Stanford GSB students, faculty, and staff, as well as Stanford GSB alumni. For inquires, contact the
Case Writing Office.
Download