Picking a pricing model is a values exercise dressed up as a spreadsheet. We knew that going in, and we still spent six weeks on the spreadsheet because the wrong model would have made every customer conversation harder for years.
Models we considered
- Per-message / per-token. Punishes the practices that use the AI most — the ones we want using it most.
- Per-chart. The same chart can be 60 pages or 4. Penalizes specialty.
- Per-patient. Bigger clinics subsidize smaller ones, but the unit (a patient) decouples from the value (a vet hour).
- Per-vet. The unit of value matches the unit of pricing. A 4-vet clinic pays 4×; the 4-vet clinic gets 4× more value.
Why per-vet won
Vetch exists to give vets back hours. The thing we charge for should be the thing we change. A 4-vet clinic that adopts ambient scribe gets back roughly 4× the hours; a 1-vet clinic gets 1×. The price tracks the benefit.
It also makes the business model legible to a practice owner in 30 seconds. "How much per vet?" is a question a CVPM can answer to a board without a calculator.
What we excluded from the price
Front desk seats, vet techs, volunteers — unlimited at every tier. SMS, voicemail, email — included. AI inference — included. The marginal cost of one more user, message, or token is small enough that gating it would create more friction than revenue.
Where it gets weird
Specialty practices employ residents who function as half-vets in production. We bill them as half-vets. Shelters operate on volunteer DVMs. The shelter tier is sliding-scale by intake volume, free under 200 intakes/year. Equine field-call practices have one DVM and a high revenue per visit; we land them on standard per-vet pricing — the math still tracks.
If you want the spreadsheet, email me. It's not pretty, but it's honest.