OpenAI Quietly Shifts to Pay-Per-Success Pricing Model for Select Enterprise Customers

OpenAI is quietly rolling out outcome-based pricing for select enterprise clients, charging only when AI successfully completes a task — a bold industry shift that transfers financial risk from buyers to vendors, though defining 'success' for complex AI tasks remains a critical unresolved challenge.

• OpenAI is quietly allowing select major enterprise customers to pay only when its AI successfully completes a task, a model known as outcome-based pricing, rather than the traditional per-token billing structure.
• Outcome-based pricing is gaining traction across the AI industry, with companies like Intercom and Zendesk already charging per resolved customer support interaction, while Salesforce has moved toward consumption-based pricing, which still bills for failed actions.
• The shift transfers financial risk from buyers to AI vendors, but defining 'success' remains a key challenge, especially for complex multi-step agentic tasks, and enterprises unable to forecast costs tend to delay full deployment until pricing is predictable.