- Under fee-for-service, AI caps out at administrative efficiency.
- Value-based structures convert the same capability into compounding margin.
- Your reimbursement mix should determine where you deploy first.
The AI return-on-investment argument in healthcare is really an argument about reimbursement. Under fee-for-service, AI makes you more efficient at billing for the same volume, which is a cost saving with a floor. Under value-based arrangements, the same capability changes what you get paid. Same technology, opposite economics.
Why fee-for-service caps AI out
Fee-for-service pays for activity. If AI helps you document faster, code more accurately and chase fewer denials, the benefit is lower cost per encounter. Real money, and it stops.
It stops because the revenue side is fixed by volume and schedule. Making the administrative wrapper cheaper does not increase what an encounter is worth, so the return is a one-time efficiency step with maintenance costs attached.
That is a cost centre with a good business case, not a growth engine. Nothing wrong with it — just do not model it as compounding.
How value-based structures convert it
Value-based arrangements pay for outcomes and shared savings. Here the same capabilities point somewhere else: identifying the patients who will deteriorate, closing care gaps before they become admissions, keeping cohorts in programmes.
Each improvement raises the payment for the same population, and the effect accumulates because a better-managed cohort stays better managed. The margin expands rather than the cost contracting.
The technology is not different. The contract is.
AI under FFS reimbursement caps out at administrative efficiency; only VBC structures convert AI into compounding margin.
What your reimbursement mix implies
If you are mostly fee-for-service, deploy for administrative relief and be honest that you are buying efficiency. Size the programme accordingly and do not promise transformation.
If you have meaningful value-based exposure, put your first agents against the population-health workflows where outcomes move payment. That is where the compounding is.
If you are transitioning, build for the destination. Administrative agents rarely convert into population-health agents, because the data and the governance differ.
What to do next
- 01Split your AI business case by reimbursement model. One case for two economics is how programmes get misjudged.
- 02If you have value-based exposure, aim the first agents at outcome-moving workflows rather than admin.
- 03If you are mid-transition, build for the contract you are moving to.