OpenAI is testing AI payments tied to successful task outcomes
According to The Information, OpenAI is testing payments for AI services with large customers that are made only after a task has been successfully completed. Salesforce, Adobe, HubSpot and Zendesk are also moving in the same direction; attribution of success to AI remains contentious.
According to The Information, OpenAI has in recent months begun offering some large customers the option to pay for an AI service only after a task has been successfully completed, such as handling a customer support request. The information came from a source familiar with these arrangements, a spokesperson for OpenAI declined to comment, and the move had not previously been publicly known.
This outcome-based pricing model was championed primarily by startups Sierra and Fin (Salesforce is now buying Fin for 3.6 billion dollars) — both charge only for tasks that AI completes without human intervention. Cognition, the creator of a coding assistant, offers business customers credits of up to 10 million dollars if the software fails to deliver results worth at least the price paid. According to the article, established suppliers Adobe, HubSpot and Zendesk are also moving in the same direction. For the newly unified AI suite CX Enterprise, Adobe plans to tie part of its billing to the value created, such as the number of completed advertising campaigns, but has not disclosed specific prices; subscriptions and usage-based billing remain in place for other tools, and existing AI features for photo and video editing are excluded from this.
According to The Information, the shift away from flat-rate subscriptions is being driven by the high operating costs of AI products and the fact that they have not yet significantly accelerated revenue growth at software companies. Customers also have tight IT budgets because of competing tools; for example, the Claude model from Anthropic recently shifted further toward usage-based billing. Salesforce now offers customers individually negotiated contracts for its Agentforce product tied to how much AI increases revenue or reduces costs; CEO Marc Benioff said at an investor conference that the aim is to charge a payment based on how much AI actually delivered, rather than a fixed amount for the number of calls handled.
According to the article, the main problem with this model is assigning credit for the outcome — as this approach becomes more popular, it is increasingly difficult to determine whether cost savings or a closed deal resulted from the software or from the customer. Payments company Stripe has already published guidance on this, stating that an outcome may also stem from product changes, marketing campaigns or seasonality; without clear attribution rules, customers may question whether the AI supplier actually caused the success.
Why it matters
For companies buying AI services, this opens up the possibility of paying only for value actually delivered instead of a flat fee, reducing the risk of paying for an unused tool. At the same time, it leaves both sides with a dispute over who deserves credit for the outcome — without clear attribution rules (as proposed by Stripe), there is a risk of billing disputes between the supplier and the customer.
Relevant practical impact
What this means
For a business
Companies buying AI services (e.g. for customer support) can negotiate payments tied to outcomes instead of a flat fee with major suppliers, but must bear in mind that attributing success to AI versus other factors (marketing, seasonality, product changes) is contentious and, without clear rules, can lead to billing disputes with the supplier.
Sales and marketingCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.