Health systems have spent 20 years optimizing for the patient who searches, clicks, and reads. They are not optimizing for the agent that queries, evaluates, and routes. Those are two different audiences — and most organizations are only ready for one of them.

The digital front door was built on a human assumption: that discovery begins with a search, passes through a website, and ends in conversion. Agentic AI doesn't use doors. It uses structured pathways, machine-readable attributes, and decision logic that operates entirely outside your owned channel. The routing is already happening. The question is whether health systems are in the decision set - or invisible to it.

The infrastructure making this possible isn't speculative. Model Context Protocol (MCP), now an open standard backed by Anthropic, OpenAI, and Google DeepMind, defines how AI agents connect to external tools and data sources. NLWeb, launched by Microsoft in May 2025, turns websites into machine-queryable endpoints. Together, they create an execution layer on top of your digital ecosystem. And most hospital websites aren't built to be legible to it.

Chris Boyer and Reed Smith work through what this shift actually requires:

Why the patient journey now runs conversation → AI interpretation → machine routing → conversion — and health systems control only the last step
What breaks when machines encounter unstructured provider bios, inconsistent service line naming, and scheduling availability gaps
Why brand strength built on emotional resonance doesn't translate to machine-readable signals — and what does
The gap between "78% of health systems engaged in AI projects" and the 52% that feel operationally ready to implement them
What a practical machine readiness audit looks like, and who inside the organization should own it

The organizational problem is as hard as the technical one. Marketing owns content but rarely owns schema. IT owns infrastructure but rarely thinks in terms of machine-readable patient experience. Someone has to own machine readiness as a cross-functional problem. Right now, almost no one does.

If your digital strategy was designed for the patient who searches, clicks, and reads - it was not designed for the agent that queries, evaluates, and routes.

Mentions From the Show:

Learn more about your ad choices. Visit megaphone.fm/adchoices

TP476: Good Enough for People Is Not Good Enough for Machines