OpenAI is reportedly experimenting with a different way to charge some enterprise customers: payment after the AI completes the job. The change points to a broader rethink in software pricing, as vendors look beyond traditional subscriptions and customers press for clearer value from costly AI tools.
A New Test For Enterprise AI Pricing
According to The Information, OpenAI has started giving some large customers the option to pay only after its AI successfully finishes a task. One example named in the report is handling customer support requests.
The arrangement has taken shape over the past few months, The Information reported, citing a person with direct knowledge of the deals. The move had not previously been public, and an OpenAI spokesperson declined to comment.
The model is known as outcome-based pricing. Instead of charging a flat subscription fee, or billing purely by usage, the vendor is paid when the software produces a defined result.
For buyers, the appeal is straightforward. If an AI system is meant to complete work, the bill is easier to justify when it is connected to completed work. For vendors, the model can make AI easier to sell to customers who are watching budgets closely and comparing several tools at once.
Why Subscriptions Are Under Pressure
The shift comes as AI products remain expensive to run. The Information reported that these tools have not yet accelerated revenue growth at software companies, creating pressure to find pricing models that better match what customers believe they are receiving.
At the same time, corporate IT budgets are being pulled in different directions. The source article points to competing tools such as Anthropic's Claude, which recently pushed harder toward billing by usage.
That leaves vendors with a practical challenge. A fixed subscription can feel disconnected from performance, while usage billing can still charge customers even when the output does not create business value. Outcome-based pricing tries to close that gap by linking payment to a result the customer recognizes.
Several pricing approaches are now visible across the market:
- Subscription pricing, where customers pay a fixed recurring fee.
- Usage billing, where payment is tied to how much the tool is used.
- Outcome-based pricing, where payment depends on a completed task or measured business result.
Startups Pushed The Model First
OpenAI is not the only company moving in this direction. The source article describes outcome-based pricing as a model that startups have pushed especially hard.
Sierra and Fin charge only for tasks completed by AI without human involvement. Fin is now being acquired by Salesforce for $3.6 billion.
Cognition, a coding assistant, has made a different kind of promise to corporate customers. It offers credits of up to $10 million if the software does not produce results worth at least the price paid.
More established software vendors are also shifting. Adobe, HubSpot and Zendesk are named in the source article as older vendors heading toward similar pricing ideas.
Salesforce now gives companies multiple ways to pay for its AI product, Agentforce. Those options include individually negotiated contracts connected to how much the AI increases revenue or reduces costs.
Salesforce CEO Marc Benioff said at an investor conference that customers want different ways to buy and price software. The idea, as described in the source, is not only to count completed calls and attach a price to them. It is to connect the charge to the value the AI helped create, such as revenue that rose because of the software.
The Attribution Problem
Outcome-based pricing sounds simple until the parties need to decide who deserves credit for the outcome. If a sale closes, a support cost falls or a campaign is completed, the result may not come from the AI alone.
Stripe has already published guidelines on the problem. The source article notes that results can also come from product changes, marketing campaigns or seasonality.
That matters because unclear attribution can lead to disputes. A customer may argue that its own work, timing or market conditions created the result. A vendor may argue that the AI produced the measurable gain that triggered payment.
For outcome-based AI pricing to work, both sides need rules that define success before the bill arrives. Those rules must explain what counts as a completed task, what evidence is required and how outside factors are handled.
Adobe Shows A Hybrid Path
Adobe has also said part of its newly bundled AI suite, CX Enterprise, will be billed by the value it creates. The source article gives completed ad campaigns as an example of that value.
Adobe did not name specific prices. The company is also keeping subscription and usage models for other tools, and existing AI features for photo and video editing are excluded.
That approach shows where the market may be heading: not one single pricing model, but a mix. Some AI functions may remain subscription products. Others may be billed by usage. Workflows with clearer outputs may become candidates for outcome-based contracts.
OpenAI's reported test fits into that larger pattern. As AI vendors try to prove value and customers demand pricing that reflects performance, the invoice is becoming part of the product strategy.