OpenAI has reportedly begun offering some major customers a different way to pay for artificial intelligence: charging only when its AI completes tasks. Techmeme, citing The Information, says the option has been introduced in recent months and is being offered to some major customers. The report frames the move as part of a broader pricing experiment across the AI sector. According to the same item, Salesforce and other AI providers are also testing outcome-based pricing, where payment is tied to task completion rather than simply access, usage, or seat count. The available summary does not specify which OpenAI customers are using the option, what tasks qualify as completed, how completion is measured, or how prices compare with existing usage-based plans. It also does not name the other AI providers beyond Salesforce. That makes this a developing pricing story rather than a settled shift in OpenAI’s commercial model. The material claim is narrower: some major customers are reportedly being given the option, not that OpenAI has broadly replaced its current pricing structure. For buyers, the appeal is straightforward if the mechanics hold: paying for completed work may be easier to justify than paying for model calls or tokens. For vendors, the risk is that outcome-based pricing requires clearer definitions of success, stronger measurement, and more confidence that AI systems can reliably finish the work customers value. Who benefits: Large customers may gain more leverage if they can tie AI spending to completed tasks. Vendors with reliable task automation may also be able to charge against outcomes rather than raw usage. Who's exposed: AI providers could take on more performance risk if they charge only when work is completed. Customers also need clear definitions of what counts as a completed task before treating the model as lower-risk.