Information Week has raised a question most SaaS pricing teams have been dodging: what happens to per-seat pricing when the buyer is an AI agent, not a person? The honest answer is that the industry's dominant revenue model was built for human users logging in, and agents do neither.

Per-seat pricing — charging a fixed rate per named user per month — has carried SaaS for two decades because it scales linearly with headcount. An AI agent inflates usage, calls APIs, and triggers workflows without ever occupying a seat. One customer can run hundreds of agent-driven transactions through a product licensed for fifty humans. Revenue and value delivered decouple, and the account either overpays or, more often, the vendor undercharges massively.

The problem is structural, not cosmetic.

What breaks first?

• Seat counts stop mapping to usage. Agents act on behalf of users who never log in, so the core meter of SaaS pricing goes dark.
• Procurement loses its anchor. Buyers negotiate on seats because seats are countable. Agents make consumption unpredictable and hard to budget.
• Packaging gets awkward. Features priced as user add-ons — seats with extra permissions, admin tiers — have no buyer when the "user" is a bot.

Which alternatives are on the table?

Vendors experimenting in this space generally fall into a few camps. Consumption pricing charges per API call, per action, or per task completed — metering the actual work agents do. Outcome-based pricing ties fees to results, such as resolved tickets or closed bookings. Hybrid models keep a base platform fee and layer usage charges on top.

Each carries real tradeoffs, and Information Week's framing implies them without pretending any model has won:

• Consumption pricing is measurable but volatile; customers hate unpredictable bills.
• Outcome pricing aligns value and cost but demands attribution you may not be able to prove.
• Hybrids are easiest to sell and easiest to game.

When does the old model still work?

Seat pricing is not dead. It remains defensible where AI augments a human's workflow inside the same seat — a copilot embedded in an existing licensed product. It fails where agents operate autonomously, at machine speed, on behalf of absent humans. The failure condition is volume without headcount.

What should product managers do Monday morning?

Audit which parts of your product an agent can invoke without a human in the loop. Measure it: what share of API traffic, actions, or workflows no longer traces to a logged-in user? If that number is growing, your pricing is leaking, and patching seat definitions in the contract will only delay the reckoning.

The deeper shift is that pricing is becoming a product decision, not a finance decision. As agents mediate more software usage, the teams that instrument value at the action level — and can price it — will set the terms for the next SaaS generation.