Buyer's guide
Outcome-Based Pricing for AI Agents
Seats and tokens measure how much AI you use. Outcome-based pricing measures what it gets done. Here are the pricing models for AI agents, what has to be defined before any of them work, and how Tekscape prices AI Outcomes as a Service.
Outcome-based pricing for AI agents ties what you pay to an agreed result, such as a resolved request or a document processed correctly, rather than to seats, tokens or hours. It works only when the outcome, the baseline and the acceptance test are defined before the agent goes live. Many providers blend it with a fixed fee, because some work an agent does, such as routing or triage, is not an outcome on its own.
The models
Five ways AI agents are priced
| Model | You pay for | Fits when | Who carries the risk |
|---|---|---|---|
| Per seat | Each user with access | People use the AI as a tool in their own work | You: unused seats still cost |
| Usage or tokens | Model calls, tokens or minutes | Volume is predictable and the tool is self-serve | You: costs rise with volume, results or not |
| Per outcome | Each completed result | The result is easy to count and verify | The provider, if the outcome is defined well |
| Managed fee with acceptance gates | Each agent in service, after it passes its test | Work is varied and needs ongoing operation | Shared: the agent earns its place before it is billed as a service |
| Blended | A fixed base plus outcome or usage components | Some of the work is countable and some is not | Shared, by design |
How it works
What outcome-based pricing means in practice
Sierra, which prices its customer-service agents this way, puts it simply: "With outcome-based pricing, Sierra gets paid only when we complete a task for you." Sierra also notes the limits: criteria for each outcome are agreed up front, escalations are usually not charged, and some interactions suit consumption pricing better, so it sometimes offers a blended model.
The same lessons apply to any AI agent. The hard part is not the price per outcome. It is agreeing what counts as one, what does not, and how a disputed case is settled.
Before you sign
What to define before any outcome-based price works
The outcome unit
Exactly what is counted, in words both sides would apply the same way to a borderline case.
The baseline
How the work is done today, so the price can be judged against the cost of doing nothing.
The acceptance test
The bar an agent must clear on real work before it is billed as being in service.
Human hand-offs
What happens when a person has to step in, and whether that case counts.
Exclusions
Bad source data, out-of-scope requests and upstream system failures that should not count against the agent.
Reporting and disputes
Who reports the numbers, how often, with what evidence, and how a disagreement is resolved.
How Tekscape prices it
How Tekscape prices AI Outcomes as a Service
- Step one
Free outcomes session
We pick one workflow with you, write down the outcome and today's baseline, and agree the acceptance test.
- Step two
Fixed-scope pilot, one-time fee
A six-week path from design to the agent running on real work under human review, for a fee agreed up front.
- Step three
Monthly fee, only after the agent passes
An agent that passes its acceptance test moves into managed service for a monthly fee, reported against the outcome every month. One that does not pass never goes into service, and you keep the findings.
We give a fixed scope after the outcomes session rather than a generic price list.
Running costs
The model bill is part of the outcome, not a surprise
Model usage is the cost that grows quietly. In a managed service, the provider routes each task to the cheapest model that passes the test, so the running cost is part of what is managed.
Frequently asked questions
Sources
- Sierra: Outcome-based pricing for AI agentsSource of the quoted definition and the caveats on defining outcomes, escalations and blended pricing.
Get a fixed scope and a test
Bring one workflow to a free AI outcomes session and leave with the outcome defined, the baseline measured and a scoped pilot.
Book a free AI outcomes session