OpenAI is now billing select enterprise clients only after its artificial intelligence successfully completes a task, a move that shifts the risk from the buyer to the vendor.
According to a report by The Information, the tech giant has begun offering large accounts an arrangement where payment is triggered solely by successful outcomes, such as resolved customer support tickets. This approach mirrors a strategy pushed by startups like Sierra and Fin, which charges only when the AI executes a job without human intervention. Salesforce recently acquired Fin for $3.6 billion and now ties pricing for its Agentforce platform directly to revenue gains or cost reductions. The coding assistant Cognition offers corporate clients credits worth up to $10 million if the software fails to deliver results matching the price paid. Established vendors including Adobe, HubSpot, and Zendesk are also moving toward this model.
Why the subscription fee is under pressure
High operating costs are forcing the industry to rethink its billing structures. So far, AI products have not accelerated revenue growth at software companies, putting pressure on firms to adopt outcome-based pricing. Meanwhile, strained IT budgets and competition from tools like Anthropic‘s Claude, which pushes harder toward usage-based billing, mean customers want to price in different ways. Salesforce CEO Marc Benioff noted at an investor conference that the goal is not simply to charge for a number of closed calls, but to charge based on the actual revenue the AI brings in.
Who gets credit becomes the sticking point
As this model becomes more common, attributing success becomes harder. A result could stem from product changes, marketing campaigns, or seasonality rather than the software itself. Payment provider Stripe has already published guidelines on this issue, warning that without clear rules, customers might dispute whether the vendor drove the success. Adobe has announced it will bill part of its CX Enterprise suite by the value created, such as the number of ad campaigns completed. Specific prices were not named. Subscription and usage models remain for other tools, and existing features for photo and video editing are excluded.
What it means
For companies relying on these tools, the financial risk shifts. Instead of paying a flat fee regardless of performance, businesses pay only when the AI delivers a verified result. This reduces the cost of experimentation but places the burden of proving success squarely on the vendor.
