OpenAI and The Information announce a partnership

Decision brief

What changed

OpenAI is testing enterprise pricing that charges for successful AI outcomes rather than token usage, and The Information reported the model has already begun in pilot form. The coverage says the change shifts billing from compute consumed to success-based pricing, but no public criteria have been disclosed for how a task is judged successful.

Why it matters

If this pilot expands, enterprise buyers would need to evaluate AI vendors less on unit-cost efficiency and more on whether they can define, measure, and audit business outcomes reliably. That makes evaluation frameworks, workflow instrumentation, and contract terms more strategically important, especially for multi-step agent use cases where success is subjective. Assumption: decision-makers should treat this as an early commercial signal rather than a broadly available pricing change, because the coverage describes a pilot rather than a general rollout.

Affected roles

CEO CFO COO CTO

Evidence

The New Stack reported that OpenAI is testing outcome-based enterprise pricing and attributed the existence of pilots to reporting by The Information. The single provided article is directionally consistent on the core fact of pilot testing, but it is based on secondary reporting and does not include independently verified public pricing terms from OpenAI.

What remains uncertain

It is still unknown how OpenAI defines a 'successful' outcome, how disputes would be handled, which enterprise customers are in the pilot, and whether this model applies broadly or only to selected agent workflows. The supplied coverage does not confirm commercial terms, governance requirements, or whether customers will accept OpenAI-controlled evaluation mechanisms.

Monitor next

Watch for OpenAI to publish or customers to disclose concrete pilot terms, especially the success criteria, evaluation process, and dispute/audit provisions for outcome-based billing.

Analytical support, not advice — assumptions and open questions stated above.

Source coverage