The opportunity is there to reduce the cost of healthcare, but with the way healthcare is run in the USA, do you think patients will see cost reductions?
My guess is insurance companies are finding every way they can to leverage AI to reduce their costs. Some of these ways will be perfectly legitimate and there will probably be some ways that it happens at the expense of patients.
Either way, I highly doubt that any cost-savings will be passed down in our for-profit healthcare environment.
Spot on about AI amplifying the incentives it’s given.
I’ve been working on a framework that tackles this head-on by making those incentives — and compliance rules — unchangeable at the protocol level. In healthcare, that means AI can’t drift into unsafe, non-compliant territory because the lawful constraints are built into its operation, not just into a policy doc.
If the incentives are aligned with patient safety, the AI’s outputs stay aligned too — no matter the pressure to cut costs or “move fast.”
Governance works best when it’s in the code, not just in the meeting notes.
You've nailed the core truth: AI doesn’t change the incentives, it scales them. If your system rewards cost-cutting over care, AI will industrialize denial. If you use it to mask staffing shortages, you’re scaling silent risk. And if clinicians treat AI outputs as gospel, you’ve automated bias at speed.
The trust gap here isn’t just about the model, it’s about governance. Without transparent objectives, patient-centered metrics, and enforceable accountability, AI will merely amplify existing structural incentives. Trust in AI healthcare isn’t earned by accuracy alone; it’s earned by aligning the system’s goals with patient outcomes and proving it in the data.
Thanks so much for the thoughtful feedback, Maryam! Do you remember what the papers you read about AI in healthcare were about? Would love to check them out.
The opportunity is there to reduce the cost of healthcare, but with the way healthcare is run in the USA, do you think patients will see cost reductions?
My guess is insurance companies are finding every way they can to leverage AI to reduce their costs. Some of these ways will be perfectly legitimate and there will probably be some ways that it happens at the expense of patients.
Either way, I highly doubt that any cost-savings will be passed down in our for-profit healthcare environment.
Spot on about AI amplifying the incentives it’s given.
I’ve been working on a framework that tackles this head-on by making those incentives — and compliance rules — unchangeable at the protocol level. In healthcare, that means AI can’t drift into unsafe, non-compliant territory because the lawful constraints are built into its operation, not just into a policy doc.
If the incentives are aligned with patient safety, the AI’s outputs stay aligned too — no matter the pressure to cut costs or “move fast.”
Governance works best when it’s in the code, not just in the meeting notes.
That sounds incredible, Philip. Is there anywhere I could read more about how your framework operates?
Thanks for the feedback!
You've nailed the core truth: AI doesn’t change the incentives, it scales them. If your system rewards cost-cutting over care, AI will industrialize denial. If you use it to mask staffing shortages, you’re scaling silent risk. And if clinicians treat AI outputs as gospel, you’ve automated bias at speed.
The trust gap here isn’t just about the model, it’s about governance. Without transparent objectives, patient-centered metrics, and enforceable accountability, AI will merely amplify existing structural incentives. Trust in AI healthcare isn’t earned by accuracy alone; it’s earned by aligning the system’s goals with patient outcomes and proving it in the data.
Yes, the alignment and ongoing proof of function are both equally important. Thanks for your insights, Rachel!
Thanks so much for the thoughtful feedback, Maryam! Do you remember what the papers you read about AI in healthcare were about? Would love to check them out.
That’s a great list, I don’t think I’ve read any of them. Thank you!