Why we ditched per-seat pricing for a credit model
After two years of seat math, our customers were paying for chairs nobody sat in. Here's how we re-priced the platform around what people actually do.
Per-seat pricing made sense when software was something you opened every morning. An AI app builder is not that: some weeks a team ships four products, other weeks nobody logs in at all. Charging for chairs punished exactly the teams we wanted — the ones who invite a designer for one afternoon.
A credit is one unit of agent work. Building a screen, running a fix loop, generating an image — each has a visible price before you confirm it. You can hand a workspace to ten people and still only pay for the work they actually asked for.
The migration taught us that predictability matters more than cheapness. So credits never expire mid-month without warning, every run shows its cost in the chat, and the monthly allowance resets on a fixed date rather than a rolling anniversary.
If you are pricing your own product, the lesson generalises: bill for the outcome your customer talks about internally. Nobody in a standup says 'we used four seats yesterday'.