OpenAI Quietly Rolls Out Outcome-Based Pricing, Charging Some Customers Only When Tasks Succeed

OpenAI has started offering some large enterprise customers a new billing option: pay only when its AI actually completes a task, according to a report by The Information citing a person with direct knowledge of the arrangements. The change, which has not been previously disclosed, applies to use cases such as automated customer support requests. An OpenAI spokesperson declined to comment on the report.

The move marks a departure from OpenAI's standard subscription and per-token API pricing, and signals that the company is adopting a billing model pioneered by AI-native startups rather than legacy software vendors.

Outcome-based pricing spreads across the industry

OpenAI is following a pricing strategy that startups have pushed hardest. Sierra and Fin — the latter currently being acquired by Salesforce for $3.6 billion — already charge customers only for tasks their AI agents complete without human intervention. Coding assistant startup Cognition goes further, promising corporate customers credits of up to $10 million if its software fails to deliver results worth at least as much as what they paid.

Established software vendors are moving in the same direction. Salesforce now lets Agentforce customers negotiate individual contracts tied to revenue lift or cost savings rather than flat fees. CEO Marc Benioff described the shift at an investor conference: rather than charging a fixed amount per call closed, Salesforce wants to charge based on how much revenue the AI actually generated for the customer — for example, billing $2 if the AI drove $20 or $40 in new revenue.

Adobe has said it will bill part of its newly bundled AI suite, CX Enterprise, according to value created, such as the number of ad campaigns completed, though it has not disclosed specific prices. Existing subscription and usage-based pricing remains in place for other Adobe tools, including its established photo and video editing AI features. HubSpot and Zendesk are reportedly moving toward similar structures.

Why the industry is rethinking subscriptions

The shift is driven by economics on both sides. AI products remain expensive to run, and according to The Information, they haven't meaningfully accelerated revenue growth at software companies despite the added cost. At the same time, customer IT budgets are under pressure from competing tools — Anthropic's Claude has also pushed harder toward usage-based billing in recent months, adding competitive pressure on pricing structures industry-wide.

The attribution problem

The central challenge with outcome-based pricing is proving the AI actually caused the result being billed for. Payment provider Stripe has published guidelines on this problem, noting that a cost saving or closed sale could stem from unrelated factors like product changes, marketing campaigns, or seasonal demand rather than the AI itself. Without clear attribution rules, customers could dispute that a vendor's software drove a given outcome, creating friction in contract negotiations and billing disputes.

What this means

Outcome-based pricing shifts commercial risk from buyer to seller — a signal that vendors are betting their models now perform reliably enough to guarantee results, or that competitive pressure is forcing them to make that bet regardless. For enterprise buyers, it removes the risk of paying for AI that doesn't work, but it also requires agreeing on measurable, disputable success criteria upfront, which is harder than it sounds for anything beyond narrow tasks like resolved support tickets. Expect attribution methodology — not the AI's raw capability — to become the next competitive and legal battleground as this pricing model spreads from AI-native startups to incumbents like OpenAI, Adobe, and Salesforce.