outcome-based pricing

advancedspreading

A pricing model in which you pay not for software use or time spent, but only for a measurable result actually achieved, for example a resolved customer complaint.

In plain terms

Imagine that instead of paying an hourly rate to the plumber, you only pay them if the tap really stops dripping. Outcome-based pricing applies this same principle to the world of software and AI agents: the provider earns only when the work is truly completed.

For traditional software, companies pay per user per month, whether they use them or not. An AI agent, however, performs tasks independently and does not provide tools for humans, so the question of 'how many colleagues use it' loses its meaning. Instead, what matters is how many cases were closed, how many appointments were booked, and how many invoices were processed.

This is attractive to the buyer because part of the risk shifts to the seller: if the agent does not perform, you do not have to pay for it. The provider hopes to earn more with good performance than with a fixed subscription.

How it works

Both parties agree in advance, in a contract, on what counts as a result, for example, that the agent resolved the client's question without human intervention.
They agree on a unit price: the cost per resolved case, booked appointment, or collected debt.
The system logs every case and marks whether it was successful; a failure is counted, for example, if the case ultimately had to be taken over by a human.
At the end of the month, the provider bills only for successful cases. There is often a base fee or a cap alongside to keep costs predictable.
Parties occasionally review measurements and disputed cases because defining the 'outcome' is rarely clear-cut in practice.

An everyday example

A web store sets up an AI agent for customer service chat and pays the provider a small fee per resolved inquiry. If you ask where your package is and close the window satisfied after the agent's reply, that counts as a billable case for the store. If you request a human because you received no substantive answer, the store does not pay for the agent's work.

Who is affected

entrepreneur, executiveWhen purchasing an AI service, you must decide between a subscription, a usage-based fee, or a performance-based fee, which fundamentally determines your costs and risks.

employeeIf an agent's work is priced per unit, it becomes directly comparable to the cost of your work, which can also affect your role. → Work & HR

developerYou must build the agent so that every outcome is reliably measurable and auditable afterward; otherwise, billing is impossible. → Code

everyoneAs a customer, you increasingly talk to an agent whose operator is interested in marking your case as resolved.

A common misconception

MythOutcome-based pricing is always cheaper for the buyer, as they pay only for success.

In realityIf the agent performs well and resolves many cases, the total amount may be higher than a fixed subscription. Risk decreases, but costs do not necessarily.

Limits and risks

Defining an "outcome" is delicate: if a case counts as resolved when the customer does not reply, it may be billable even if the customer simply gave up.
The measurement is mostly performed by the provider's own system, so it is worth stipulating independent verification or spot checks.
The monthly cost is harder to plan because it fluctuates along with traffic and the agent's success.
Poorly chosen metrics create distorted incentives: the agent may be tuned to close easy cases quickly instead of providing real help.

Similar, but not the same

usage-based pricing – There, you pay for the resources used (such as a token or query), even if the final result is unusable.
per-user subscription – There, you pay per person and per month for access, regardless of how much work is completed.
success fee – A success fee is typically a one-time commission tied to large deals, while outcome-based pricing is built on many small, recurring, automatically measured cases.

Related concepts

Broader

AI business models, software pricing, software as a service

Narrower

per-solution pricing, task-based pricing

On the edge

usage-based pricing, AI-agent, digital workforce, return on investment, service as software

A brief history

The principle is old: the "Power by the Hour" arrangement was introduced in 1962 by the Bristol Siddeley engine manufacturer, which later became part of Rolls-Royce, and the name is now a Rolls-Royce trademark. In this model, operators of business aircraft paid a fixed fee per flight hour for engine maintenance and replacement as needed, rather than paying for repairs on an ad-hoc basis. In the world of AI agents, it became known in 2023 when Intercom began pricing its Fin customer service agent at $0.99 per resolved inquiry.

Where things stand

As of late 2026, this is an established solution mainly for customer service AI agents, because results are easiest to measure there. In other areas, providers mostly offer a mixed model: a base fee, usage-based charges and an outcome-based fee combined.

Status: 2026 November

In the news

Appears in 1 Agent Age stories; the curve shows weekly occurrence in the news magazine.

Performance-based pay rises spread, with AI gaining a role in pay decisionsAug 7, 2026 · Work & HR

Also known as

outcome-based pricing · performance-based pay · output-based pricing · outcome-based pricing · outcome pricing · results-based pricing · pay-per-resolution · pay-per-outcome

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.