Pricing is the least discussed, most consequential decision a health AI startup makes, and most founders back into it rather than choosing it deliberately. The model you pick is not just a monetization mechanic. It signals to a buyer what kind of company you are, and it determines whether a CFO, a medical director, or a health plan actuary is the one who has to approve you.
The five models and how they actually behave
Per-seat or per-provider pricing. You charge per licensed clinician or user. This is the easiest model to sell into a single department because it maps to a familiar software budget line, and it is the fastest to close for tools that clearly save an individual clinician time, like documentation or inbox management. Its weakness: it caps your revenue to headcount, and it gives the buyer a very easy lever to cut, seat reduction, the moment budgets tighten.
Per-facility or per-site pricing. Common for infrastructure and workflow tools that serve an entire department regardless of headcount, like an imaging triage tool or a bed-management system. It simplifies procurement because it is one number per site, but it undervalues you in large, high-volume facilities unless you tier by bed count or annual volume.
Per member per month (PMPM). The dominant model for anything selling into a health plan or a value-based care contract, because it matches how the buyer itself is paid. PMPM pricing ties your revenue to population size rather than usage, which is attractive to a payer because it is predictable, but it means your revenue is disconnected from actual utilization, so you need real unit economics discipline if usage turns out higher than modeled.
Per-episode or per-encounter pricing. You charge per patient interaction, per scan read, per triage event. This aligns cost with usage cleanly and is easy for a finance team to model against volume, but it means your revenue is volatile month to month and dependent entirely on the buyer's patient volume, which you do not control.
Outcomes-based or shared-savings pricing. You get paid based on a measured result, reduced readmissions, reduced total cost of care, avoided ER visits. This is the model every founder wants to pitch because it sounds like alignment, and it is the hardest to execute, because it requires a clean attribution methodology, a baseline both sides agree on, and enough balance sheet to survive the lag between delivering the service and collecting payment on a result measured months later.
Why healthcare pricing is not like normal SaaS pricing
For payers and managed care organizations, every vendor pricing model gets evaluated against medical loss ratio (MLR), total cost of care, and quality metrics, because that is the lens through which every vendor pitch gets scrutinized. A pricing model that does not map cleanly to how the buyer's own economics work will get stuck in finance review even if the clinical champion loves the product. This is the single most common reason a well-liked pilot never converts to a contract.
Which model fits which buyer
- Selling to an individual clinical department or practice: per-seat or per-encounter, because the buyer's budget authority is limited and departmental.
- Selling to hospital operations or IT: per-facility, tiered by volume, because it is the cleanest to model against a capital or operating budget.
- Selling to a health plan or at-risk provider group: PMPM, because it matches their own revenue model and actuarial planning cycle.
- Selling into a value-based or risk-bearing contract where you can measure a hard outcome cleanly: outcomes-based, but only once you have enough operating history to model the downside if the outcome does not hit.
The transition trap
Many digital health companies start on per-seat or per-encounter pricing to get early revenue quickly, then try to move upmarket to PMPM once they are selling to payers. The transition is harder than founders expect, because moving from a usage-based model to a population-based model requires the company to now underwrite population-level risk, not just deliver a service. Deciding when and how to transition to PMPM is as much a finance and actuarial decision as a sales decision, and companies that make the switch without building that internal capability first tend to mis-price their first PMPM contracts badly.
A practical checklist before you set your price
- Identify who actually signs the check: a department head, a CFO, or a plan actuary. That person's budget structure should shape your model, not your cost structure.
- Model your unit economics under your highest plausible utilization scenario, not your average one, especially for PMPM and outcomes-based deals.
- Build a simple, defensible attribution methodology before you ever propose outcomes-based pricing. A vague "we reduced readmissions" claim without an agreed baseline will not survive a real negotiation.
- Price for the second and third contract, not just the first. A pricing model that only works because your first customer under-negotiated will not scale.
The takeaway
There is no universally correct pricing model in health AI. There is only the model that matches how your specific buyer's budget and incentives already work. Get that match wrong and you will spend a year in finance review for a product your clinical champion already wanted to buy.
