Per-Seat Pricing Is Dying, and Agents Are Part of Why

Per-seat pricing is running out of room, according to SaaStr’s Jason Lemkin, and not for one reason — for three that are converging at once. Pricing headroom is exhausted after years of price hikes, core SaaS functionality is commoditizing faster than vendors can differentiate on it, and — the newest pressure — AI agents are increasingly the ones evaluating and recommending purchases, which breaks a pricing model built around counting human logins.

The cost data backs up the “headroom exhausted” half of the argument. Vertice’s SaaS Inflation Index has been running 12%-16.4% through 2026, against roughly 2.7% general inflation — SaaS prices rising four to six times faster than the broader economy. Zylo’s 2026 SaaS Management Index, tracking more than 40 million licenses and $75 billion-plus in spend, found average enterprise SaaS spend has climbed to $55.7 million a year, up 8%, while the actual portfolio of apps in use stayed flat at 305. Companies are paying more for the same footprint, not buying more software — which is exactly the condition under which buyers start asking why price scales with headcount instead of usage or outcome.

The agent-driven piece of the argument is already showing up as a measurable failure mode, not just a prediction. A Forrester/Siteline benchmark found roughly 30% of AI agent runs hit search or retrieval errors trying to locate a vendor’s own pricing — and when agents can’t find it, they abandon the vendor’s site for third-party sources instead. Vendors whose pricing exists only behind a “contact sales” form, structured for a human buyer to navigate, are invisible to the agents now doing early-stage evaluation.

For B2B sellers, the practical read isn’t that seats disappear overnight — it’s that pricing pages built for humans-per-seat need a second, machine-legible version, and the sales motion needs to account for a buying committee that now includes software making the first pass.