AI Development ยท 11 min read
Per Seat Is Dying: How Token Pricing Is Rewriting SaaS Budgets
For twenty years software cost the same every month. AI turned it into a utility bill.
Per seat SaaS pricing gave finance teams predictable bills. AI features are priced on tokens, credits and outcomes, and vendors from Salesforce to Cursor to Zendesk have moved to usage models. Here is why it is happening, what it does to budgets, and how buyers should respond.
- 20 years per seat was the SaaS default
- 3 models tokens, credits, outcomes
- Unbounded an AI line item with no cap
Frequently asked questions
Is usage based pricing always worse for buyers?
No. Light users pay less than they would per seat, and metered pricing lets small teams try tools cheaply. It is worse for buyers who value predictability and for heavy users, which describes most companies once a tool succeeds.
Why do vendors not just raise seat prices instead?
Some have. But a seat price high enough to cover heavy AI users overprices light users, and light users churn. Metering lets vendors keep the low entry price and charge heavy users what they cost.
What is outcome pricing?
Charging for a result rather than for usage or seats. Per resolved support ticket, per qualified meeting, per completed task. It shifts the token cost risk to the vendor and ties spend to value.
How should a small company approach this?
Set a monthly cap on every metered tool, review usage monthly, and prefer flat priced tools for anything that does not need AI at its core. Treat metered AI as a utility with an owner, not as software with a licence.