Seat-based pricing made sense in the SaaS 1.0 era. AI changed the unit economics overnight. Here's how to price for a world where the work is done by tokens, not chairs.
Meriç Karpat
Founder, RevTune
Seat-based pricing was a brilliant invention for its time. In 2010, "more users" was a clean proxy for "more value." Salesforce taught the entire industry to count chairs, and an entire generation of SaaS pricing pages copied the pattern.
Then AI happened.
The moment your product starts using LLM tokens, the seat-based model collapses. A single user can drive 100x more cost than another single user. A "free trial" can rack up triple-digit inference bills overnight. Suddenly the customer who looked profitable yesterday is bleeding you dry today.
You can't price that with seats. You have to price it with usage.
The trap is to think "usage = API calls." That's the cheapest possible interpretation, and it's the one that pisses customers off. Nobody wants to watch a meter while they work.
Better unit metrics:
Pick the one that scales linearly with your customer's value, not your infra cost. If they don't feel ripped off when the bill arrives, you picked the right unit.
The cleanest model I've seen for AI products is base seat + pooled usage. A small monthly fee gives access to the workspace. Heavy usage gets billed against a pooled allowance shared across the team.
It's predictable for finance. It scales with value. And it doesn't punish exploration — small experiments stay free, real work pays for itself.
One pricing essay every other week. No filler, no spam, unsubscribe anytime.
Requests go to a real inbox until self-serve subscriptions ship.