For twenty years, software buyers and sellers agreed on a convenient fiction. You pay for a seat because a person sits in it, and the person does the work, so the seat is roughly the work.
Nobody believed the seat was the value. It was the easiest thing to count that moved in the same direction as the value. More reps, more seats, more pipeline. It was a proxy, and it held because the software never did the job itself.
That assumption is now broken, and the pricing pages are showing it.
Three prices for one product
Look at what the biggest name in CRM charges for its agents. Salesforce launched Agentforce at a flat $2 per conversation in October 2024, added Flex Credits at roughly ten cents per action in May 2025, and then added per user licenses starting around $125 a month. All three models run at the same time.
That is not a pricing strategy. That is a company with enormous pricing talent openly admitting it does not yet know what unit of value it is selling. Per conversation assumes the value is the interaction. Per action assumes it is the work performed. Per user assumes it is still the seat. Each one is a different theory of what the product is, and the market has not picked one.
Customer service moved first, because the outcome there is easy to see. Zendesk announced outcome based pricing for its AI agents in August 2024, charging for issues the agent resolves without a human. Intercom prices its Fin agent at $0.99 per outcome, with a monthly minimum, and charges nothing when a conversation reaches the team without a defined outcome.
These are the honest versions. The vendor is saying: we will be paid when the work is done, not when someone logs in.
Outcome pricing moves the fight, it does not end it
Here is what the enthusiasts skip. The moment you price on outcomes, the definition of the outcome becomes the contract.
Read the fine print on any resolution based model and you find a working definition of success, usually something like the customer did not ask for more help after the agent's last answer. That is a reasonable rule and also a contestable one. A customer who gave up is indistinguishable from a customer who was helped. A customer who called the phone line instead never shows up as a failure at all.
Under seat pricing, the buyer and seller argued about headcount. Under outcome pricing, they argue about what counts as done. That second argument is harder, more technical, and it is decided by whoever controls the measurement. Right now that is almost always the vendor, because the vendor's system is the one recording the outcome.
Buyers who celebrate the death of the seat should notice who now holds the meter.
The budget problem nobody priced in
The other reason seats survived is that finance teams love them. A seat is a line item you can forecast a year out. Usage and outcomes are not. An agent that resolves more tickets than expected is a success that blows the budget, which is a strange incentive to hand a CFO.
This is why the retreat back toward seats is already visible. When Salesforce layered a per user license on top of its consumption options, it read like a response to buyers who wanted predictability more than precision. Expect more of this: hybrid contracts that look like seats on the invoice and like consumption in the fine print, with true ups at renewal.
So the seat will not die cleanly. It will be repurposed as a budgeting container that no longer describes anything happening inside it.
What actually changes
The interesting consequence is not on the vendor side. It is on the buyer side.
For two decades, a software purchase was a staffing decision in disguise. You bought tools for the people you had. Now a purchase can be a decision about which work no longer needs a person at all, and the price is a direct statement of what the vendor thinks that work is worth. Ninety nine cents to resolve a ticket is a number a support leader can compare to a loaded hourly cost. That comparison never existed when the software just made people faster.
That makes pricing a strategic signal, not a procurement detail. When a vendor charges per outcome, it is betting it can do the job. When it insists on seats for an agent product, ask why. Sometimes the answer is that buyers wanted predictability. Sometimes it is that the agent cannot reliably finish the work, and the seat hides that.
The seat was always a proxy for labor. The industry is now pricing the labor directly, and finding out how much of it the software can actually do. Watch which unit a vendor is willing to be paid in. It tells you more about the product than the demo does.