Per-seat, per-user, and team-based pricing — and its slow decline in the age of AI agents.
AI makes your product more efficient. Efficiency means fewer users needed. Fewer users means fewer seats. The math on how this compounds into an NRR cliff — and what to do before it hits.
Everyone says switch to usage-based. But seat pricing wins when collaboration is the core value, usage is uniform, and procurement needs predictability. A framework for when it's the right call.
SaaS companies know usage-based pricing is better — but they keep counting seats anyway. Here's why, and what it costs them.
Not dead, but declining. Per-seat pricing still works when collaboration is the core value (Figma, Slack) and usage is roughly uniform across users. But AI agents are compressing seat counts, and many companies are layering usage-based components on top of seat licenses.
When AI makes your product more efficient, customers need fewer human seats — reducing revenue even as value delivered increases. This creates a compounding NRR cliff: your best customers generate the most AI usage but need the fewest seats.
Layer a platform or workspace fee that isn't tied to human headcount, then add usage-based pricing for AI agent actions, compute, or outcomes. Microsoft's approach — maintaining seat pricing as an anchor while expanding through Copilot consumption — is the most common transitional strategy.
Platform fees (flat rate per workspace or organization), usage-based pricing (per API call, per transaction), outcome-based pricing (per resolved ticket), or hybrid models combining a base fee with metered usage. The right choice depends on how uniformly customers use your product.