OpenAI starts charging some customers only when AI delivers results

OpenAI has quietly begun allowing some of its largest enterprise customers to pay only when its AI successfully completes assigned tasks, according to a report Monday from The Information. The move marks a departure from the company's standard token-based billing and places OpenAI alongside Salesforce ↘ 3.94% and other AI providers experimenting with pricing models that tie costs to results rather than usage.theinformation

Pay for What Works

The arrangement, which has not been formally announced, is limited to select major accounts rather than offered as a general option. The terms, participating customers, and specific pricing remain undisclosed. Under the new model, enterprises are billed when OpenAI's AI completes a task end-to-end — such as handling a customer support interaction — rather than paying per token regardless of whether the work succeeded.thenextweb

The shift addresses a growing pain point for enterprise buyers. Under conventional token billing, customers absorb the cost of every failed attempt, making expenses difficult to predict. One developer running a hundred agents in parallel reportedly accumulated $1.3 million in OpenAI tokens over 30 days, an extreme but illustrative example of how costs can spiral without guaranteed results. Outcome-based pricing transfers that risk from the buyer to the vendor, giving OpenAI a direct financial incentive to ensure its models deliver.thenextweb

An Industry in Transition

OpenAI is not the first to move in this direction. Intercom charges $0.99 per resolved conversation through its Fin agent, billing nothing for unresolved ones. Zendesk went further in May, restricting charges to what it calls "Verified Resolutions," confirmed by an automated evaluation within 72 hours. Salesforce's Agentforce initially charged $2 per conversation regardless of outcome before pivoting to a consumption-based credit system at roughly 10 cents per action.thenextweb

A May survey by Futurum Group found that 43% of enterprise buyers prefer consumption-based pricing and 27% prefer outcome-based models, with fewer than one in five still favoring traditional per-seat arrangements. "Outcome-based pricing is becoming a market standard," wrote Keith Kirkpatrick, the firm's research director for enterprise software, noting that vendors offering only seat-based pricing are now being disqualified before evaluations begin.thenextweb

What Comes Next

The pricing experiment arrives as OpenAI prepares to launch Astra, its next-generation agent-focused model. CEO Sam Altman previewed the model to select customers in early August, describing it as potentially the first where AI "invents new things in a way that matters". OpenAI's quarterly revenue reached $6.7 billion in the second quarter, up 18% from the prior quarter, though rival Anthropic's sales doubled over the same period, according to The Wall Street Journal News Corp ↘ 0.03%. Defining what counts as a successful "outcome" beyond straightforward customer support resolutions remains the central challenge as agentic AI expands into multi-step tasks where completion is a matter of judgment.stocktwits+1