OpenAI CFO Sarah Friar introduced "useful intelligence per dollar" as a metric for measuring AI return on investment on July 17, 2026. Rather than adoption metrics such as sheet counts or active user numbers, this method divides total costs by the number of tasks that meet quality standards. The cost per successful task includes model pricing and compute usage, as well as time spent on human verification, retries, and rework. Friar stated that the market has long measured software success by adoption rates and argued that a more robust metric is needed to gauge the value of AI.
Meanwhile, examples of shifting billing units to outcomes are leading on the business application side. On April 14, HubSpot changed the pricing for its Breeze Customer Agent from $1.00 per conversation to $0.50 per resolution. On May 18, Zendesk restructured its pricing into a three-tier model that charges only for resolutions that pass verification by its AI evaluation model. Intercom offers its service at $0.99 per outcome. The usage-based portion of OpenAI's ChatGPT Enterprise continues to measure usage in tokens or credits.
OpenAI Chairman Bret Taylor discussed token economics and how to measure AI investment during a CNBC appearance on July 20.
Reporting and Reactions (Consolidated from 2 sources):