The Per-Seat Pricing Model Is Breaking, and Credits Are What's Replacing It
For two decades, the SaaS pricing conversation was simple: how many seats do you need. That question is losing its meaning fast. In the PricingSaaS 500 Index, 79 companies now run some form of credit-based pricing, up from 35 at the end of 2024, a 126% jump in about eighteen months. More broadly, over 60% of AI SaaS companies now use usage-based pricing in some form, up from roughly 30% three years ago.
Why seats stopped making sense
A seat was always a proxy for value: one person, one login, one unit of work capacity. That proxy breaks down the moment the software itself is doing the work instead of a human clicking through it. A five-person team with an AI agent handling what used to take fifteen people doesn't need fifteen seats, and it also doesn't want to pay for fifteen. At the same time, the vendor's own costs stopped being flat. Every AI feature carries a variable inference cost that scales with usage, not with headcount, and a flat per-seat price with unlimited AI usage baked in is a fast way to destroy gross margin.
Credits solve both problems from opposite directions. The customer pays closer to what they actually consume. The vendor's revenue scales with the cost driving it, instead of being decoupled from it.
The infrastructure that made this easier
Usage-based pricing used to be a genuine engineering project: metering, invoicing edge cases, usage dashboards, none of it simple to build correctly. That got a lot easier in 2026 after Stripe folded Metronome's metering infrastructure into its own stack, letting most vendors handle millions of usage events per second as a managed service rather than something they build in-house. Gartner separately forecasts that 40% of enterprise SaaS will include some outcome-based pricing component by 2026, up from 15% in 2022. The tooling caught up to the incentive at roughly the same time, which is a large part of why the shift accelerated this fast.
What to ask before you sign
If you're evaluating a tool priced in credits, read the credit-burn math the way you'd read a cloud bill, not a subscription line item.
- What specific action consumes a credit, and does that mapping stay fixed, or can the vendor change it later?
- What happens at the limit: a hard stop, a forced upgrade, or silent auto top-up billed to your card?
- Can you see a real usage calculator or historical burn rate before committing, not just a marketing price per credit?
- Does your actual usage pattern, spiky or steady, work in your favor under this model, or would a flat plan be cheaper for how your team really works?
Credit pricing isn't a trick to extract more revenue. It's a more honest mapping between cost and value than per-seat pricing ever was for AI-native tools. But "more honest" still means it needs due diligence before you sign, not less.