Seat-based pricing
The classic SaaS pricing model where customers pay per active user — common for copilot-style products (Cursor, GitHub Copilot, Notion AI) but eroding for autonomous agents.
Seat-based pricing was the SaaS default for two decades because it mapped neatly to organizational charts. For AI copilots that augment a specific human (the developer, the salesperson, the writer), it still works — value scales with the number of humans using the tool.
For autonomous agents it fits poorly. A single agent might serve thousands of customers; pricing it per internal seat undercharges. A team of one might run twenty agents; pricing them per seat overcharges. Most 2026 agent products use usage- or outcome-based pricing instead.
The hybrid emerging in 2026 is seat-based for the human-facing surface (the dashboard, the chat) and usage- or outcome-based for the agent work happening behind the scenes.
Frequently asked
When does seat-based pricing still make sense for agents?+
When the agent is bound to a specific human user (a coding copilot, a personal assistant). When agents serve many customers from one tenant or run unattended, seat-based misaligns price and value.