In per-seat pricing, the bill is a fixed amount for each person who has access to the software, usually per month or per year. It is the most familiar SaaS pricing model: easy to understand, easy to budget, and revenue grows as the customer’s team grows.
Slack bills per active user at $7.25 a month on Pro, and credits back seats that go unused. Snyk bills per contributing developer. Harvey sells named legal seats at roughly $1,200 a year.
Potio’s take: Seats work when the seat holder is the person who gets the value. They break when the customer is staffing a team to produce value somewhere else, because then buying the minimum number of seats is the rational choice. Per-seat pricing is not dead for AI, but it has stopped being the default.
Calendly is the case for seats. The person who sets up an account is the same person whose time gets saved, so more seats genuinely means more value and customers add them without being pushed.
Seats go wrong in three ways. Customers share logins to save money when the per-seat price feels high. Seats go unused and turn into shelfware, which surfaces at renewal as a downgrade. And when software replaces work rather than supporting it, the customer’s headcount falls exactly as your value rises.
That third problem is why AI changed the picture. GitHub Copilot’s flat $10 seat was predictable for buyers and exposed for GitHub, because one developer running heavy workloads all day could cost more than the seat earned. Most AI companies now sell seats plus a consumption or output layer rather than choosing between them.
Internal tooling is the category where seats stay right. Jira, Linear and Retool sit several steps from the customer’s revenue, and charging per bug closed or feature shipped would distort how engineers work. Seats track adoption without interfering with it.
Some vendors bill only certain seats. Loom charges creators and not viewers, and ServiceTitan charges field technicians and not office staff. Leaving a population unmetered drives the adoption that sells the metered seats.
No. Seats still work where the software helps a person do their job, as with Copilot for developers or Harvey for lawyers. They break where the software does the job instead, so most AI companies add a usage or output layer on top.
Per-seat charges for every licence bought, used or not. Per-active-user charges only for people who actually used the product in the period, which is fairer to the buyer but needs a clear definition of “active”.
Price so each person gets distinct value from their own account, and use seats only where that is true. If sharing is common, the metric is probably wrong.
More on this: SaaS Pricing Models: 12 Types, Examples and How to Choose
I'm a 3x founder and former CEO of Toggl. I work hands-on with SaaS & AI teams to fix pricing, packaging and monetization.
Book a call