What is the difference between outcome pricing and output pricing?

An output is something the AI produced. An outcome is a business result the customer would have paid for anyway. A resolved ticket is an output. A retained customer is an outcome. Most of what markets itself as outcome-based pricing is output-based pricing with better positioning, and the gap matters operationally, because the definition of resolved is a product decision that has to be measurable, auditable, and hard to game.

What does a growth team own in hybrid pricing?

Four things, all of which live in the product rather than a price book. The price metric is a retention decision, since a metric that grows with customer success produces expansion without a sales call. Meter visibility is an activation decision, because a user who cannot predict their first bill will not run the workload that would activate them. The overage policy is a churn decision. And the upgrade moment becomes an event the product detects rather than a page the user visits.

Pricing After Seats: SaaS Pricing When Agents Do the Work

Why did per-seat SaaS pricing break? The seat survived every previous platform shift because value scaled with people. More users, more work done, more willingness to pay. Agents cut that link. A support agent resolving conversations end to end replaces the person the seat was counting, so per-seat pricing captures a fraction of the value while the customer's headcount and your revenue shrink together.

What replaces it is rarely a clean jump to paying for results. The dominant pattern is hybrid: a platform fee that gives the CFO a floor, plus a variable component tied to the work the AI actually does. That makes the meter a product surface rather than a finance artifact, and it moves 4 decisions onto the growth team: the price metric, meter visibility, the overage policy, and the upgrade moment.

For 2 decades, SaaS pricing had one stable unit: the human with a login. Every model was a variation on counting people. AI products broke the unit. An agent does not log in, does not hold a license, and finishes work while the seat count stays flat. When the unit of work detaches from the unit of pricing, the pricing page stops being a finance artifact and becomes a product surface. That makes it growth's problem.

Why did the seat break?

The seat survived every previous platform shift because value scaled with people. More users meant more work done meant more willingness to pay. Agents cut that link. A support agent that resolves conversations end to end replaces the person the seat was counting. Price it per seat and you capture a fraction of the value while your customer's headcount, and your revenue, shrink together.

The market has already voted. Pure per-seat pricing dropped from roughly 21% to 15% of SaaS companies in a single year, and surveys of SaaS CEOs in 2026 find nearly all of them planning to retire seat-based pricing within 2 years. What replaces it is rarely a clean jump to paying for results. The dominant 2026 pattern is hybrid: a platform fee that gives the CFO a predictable floor, plus a variable component tied to the work the AI actually does. Salesforce runs usage, outcome, and subscription models simultaneously through Agentforce. Anthropic cut enterprise seat prices for Claude while pushing harder into usage. Some vendors keep the seat language and quietly redefine it, selling agent seats at a premium. The label survives. The unit underneath is already something else.

The measurement side breaks in the same place, since MAU and seat counts stop describing an account once 1 supervisor directs 40 agents. That is part of the argument in AI-Led Growth.

Is outcome pricing really outcome pricing?

The flagship examples are real. Intercom's Fin bills $0.99 per conversation the AI fully resolves, with no charge for failed attempts. Zendesk prices automated resolutions in the $1.50 to $2.00 range. Sierra charges only when its agent finishes without a human handoff. The direction is right. The vocabulary is ahead of the reality.

An output is something the AI produced. An outcome is a business result the customer would have paid for anyway. A resolved ticket is an output. A retained customer is an outcome. Most of what markets itself as outcome-based pricing in 2026 is output-based pricing with better positioning, and the gap matters operationally, because the definition of "resolved" is not a legal clause. It is a product decision that has to be measurable, auditable, and hard to game. The moment revenue depends on a definition your model can influence, pricing integrity becomes an engineering problem. Someone has to own the meter, and it will not be the CFO. I took that argument further in When Resolved Becomes a Price, where the eval rubric turns out to be the billing engine.

Pricing is a funnel stage now

In classic PLG, monetization was a wall: a plan page, a credit card form, an upgrade prompt when you hit a limit. You optimized it like a page. With usage and outcome components, monetization is continuous. The customer decides whether the price is fair every single week, because every week produces a bill that reflects what the product did.

This changes what buyers fear. They do not fear paying for value. They fear variance. The products that win with variable pricing treat the meter itself as a product surface: consumption the customer can see in real time, alerts that fire before an overage instead of after, caps the customer controls, and a bill that never surprises anyone. A meter the customer trusts converts. A meter that feels like a taxi running in traffic churns, no matter how fair the rate is.

I have watched a billing flow restructure move activation more than any onboarding change shipped in the same period. That was under seat pricing, where billing touches the user a few times a year. Under usage pricing, billing touches the user constantly. Whatever leverage the billing flow had before, the hybrid era multiplies it.

What does growth own in the hybrid era?

The price metric is a retention decision. Pick a metric that grows with the customer's success and expansion happens without a sales call. Pick one that punishes engagement and your best users become your most price-resentful.

Meter visibility is an activation decision. A new user on a variable plan who cannot predict their first bill will not run the workload that would activate them. Showing the cost of an action before the action is onboarding now, and it is the same budgeting problem the free tier faces, which I covered in The Free Tier Meets the Inference Bill.

The overage policy is a churn decision. Hard limits interrupt work at the exact moment the product is most valuable. Silent overages produce the invoice that triggers the cancellation review. The version that keeps accounts is the boring one: warn early, degrade gracefully, and let the customer set the ceiling.

The upgrade moment is a lifecycle asset. Under hybrid pricing, upgrades stop being a page a user visits and become an event the product detects: usage approaching the floor, an agent workload the current plan was not shaped for. The prompt that fires on that event, with the account's own numbers in it, is the highest-converting message the company sends.

None of this lives in a price book. All of it lives in the product. That is the argument for growth owning it.

The uncomfortable part

Pricing experiments used to be a spreadsheet and a landing page test. In the hybrid era they require metering infrastructure, which means R&D time, which means the growth PM has to spec a meter with the same rigor they spec a funnel event: what is counted, when it is counted, what the customer sees, and how it reconciles. Most growth teams cannot write that spec today. The ones that can are deciding their company's revenue model while everyone else is still A/B testing the pricing page header.

The seat is not coming back. Pricing has become something the product does every week rather than something finance revisits every year, and a company still treating it as a quarterly slide will learn its new rate from a competitor's page. The work has already moved. The ownership should follow it.