Thanks Luke - in addition to the pace, the very nature of traditional control points is changing. Looks like we have similar conclusions on the impact of AI, particularly on SaaS's.
The cost of leaving vs. value of staying diagnostic holds up. I'd push on the Epic piece though, because the intentionality framing misreads what happened slightly.
Epic didn't architect proprietary formats as a lock-in play. They built APIs that worked for their platform in the environment they were operating in. FHIR was not as available as you might think when that platform developed it's API layer. The Lock-in was a byproduct, not the blueprint. The outcome is identical to what you describe, but the cause matters. It explains why Epic's response to openness pressure looks like compliance, not transformation. They're not unwinding a deliberate strategy. They're adapting a platform that was never designed around portability.
And AI didn't force that open. CMS did. The July 2026 deadline under CMS-0057 removed the switching cost by regulatory fiat before the market economics fully shifted. AI has made leaving more feasible. The mandate removed the permission structure for staying closed. The article gives AI more credit than it's due here.
The move I'd watch is what Epic does with the control point it still holds. The data layer is opening. The intelligence layer is not. Epic is ramping proprietary AI inside the platform — ambient clinical decisioning, workflow tools that deepen the more you stay in the ecosystem. They're migrating the control point from Language toward Governor backed by an Exclusive. Your framework predicts this move. It just doesn't name it for Epic.
On MCP: you're right that it's moving faster than FHIR ever did. The ecosystem density is real. The Linux Foundation governance move was smart. But MCP in healthcare is not a near-term FHIR displacement. The authentication gaps are documented and significant, and regulated data environments don't pilot through that kind of exposure. FHIR is slow and mandate-driven, but it has regulatory standing MCP doesn't yet have in this space. The more interesting scenario is MCP becoming the transport layer underneath FHIR-compliant workflows. Not a replacement. Infrastructure the standard runs on.
Thanks Lisa, that's well said. Agree re Epic - the lock-in was a byproduct. AI's lowering the activation energy to move between Langauge control points - but there are many other factors at play.
Interesting read, the speed of change right now is hard to compare to anything we’ve seen before.
Thanks Luke - in addition to the pace, the very nature of traditional control points is changing. Looks like we have similar conclusions on the impact of AI, particularly on SaaS's.
The Exclusive diagnostic is the sharpest thing in this piece. Every
control point argument eventually collapses into the same question:
what is underneath it that AI cannot route around? Complexity premiums
are gone. First-mover advantage compresses from years to quarters.
What remains is either proprietary data, workflow context, or physical
infrastructure — and most SaaS companies, if they are honest, have
none of those. They have a UI and a switching cost. The window to build
something genuinely structural is narrowing faster than most roadmaps
are moving. Which control point type do you think most mid-market SaaS
companies can still realistically claim before that window closes?
Lisa, you know the healthcare space well. What's your read on my section on Epic / FHIR? Is it accurate?
The cost of leaving vs. value of staying diagnostic holds up. I'd push on the Epic piece though, because the intentionality framing misreads what happened slightly.
Epic didn't architect proprietary formats as a lock-in play. They built APIs that worked for their platform in the environment they were operating in. FHIR was not as available as you might think when that platform developed it's API layer. The Lock-in was a byproduct, not the blueprint. The outcome is identical to what you describe, but the cause matters. It explains why Epic's response to openness pressure looks like compliance, not transformation. They're not unwinding a deliberate strategy. They're adapting a platform that was never designed around portability.
And AI didn't force that open. CMS did. The July 2026 deadline under CMS-0057 removed the switching cost by regulatory fiat before the market economics fully shifted. AI has made leaving more feasible. The mandate removed the permission structure for staying closed. The article gives AI more credit than it's due here.
The move I'd watch is what Epic does with the control point it still holds. The data layer is opening. The intelligence layer is not. Epic is ramping proprietary AI inside the platform — ambient clinical decisioning, workflow tools that deepen the more you stay in the ecosystem. They're migrating the control point from Language toward Governor backed by an Exclusive. Your framework predicts this move. It just doesn't name it for Epic.
On MCP: you're right that it's moving faster than FHIR ever did. The ecosystem density is real. The Linux Foundation governance move was smart. But MCP in healthcare is not a near-term FHIR displacement. The authentication gaps are documented and significant, and regulated data environments don't pilot through that kind of exposure. FHIR is slow and mandate-driven, but it has regulatory standing MCP doesn't yet have in this space. The more interesting scenario is MCP becoming the transport layer underneath FHIR-compliant workflows. Not a replacement. Infrastructure the standard runs on.
Thanks Lisa, that's well said. Agree re Epic - the lock-in was a byproduct. AI's lowering the activation energy to move between Langauge control points - but there are many other factors at play.
Love the thinking and the discussion. !