OpenAI proposes a new framework for measuring AI investment returns
OpenAI released a new framework on Friday, arguing that the value of AI should be measured by the number of successfully completed tasks and total cost, rather than traditional software metrics such as user count or activity levels. This comes at a time when global AI spending is projected to reach $2.59 trillion by 2026, yet only 12% of CEOs report gains in both cost savings and revenue.
At a Glance
- OpenAI is introducing a new way to measure the value of artificial intelligence investments, as companies shift from technology trials to scaled deployment and need to demonstrate business impact.
- In a blog post published Friday, OpenAI Chief Financial Officer Sarah Friar outlined a framework for measuring AI based on outcomes, with metrics including the number of tasks successfully executed and the total cost of completing those tasks.
- "For years, software was measured by adoption: seats, active users, renewal rates," Friar said in a separate LinkedIn post. "AI is different—it needs to be measured by the work it completes."
Deep Dive
The move comes as companies face increasing pressure to prove that AI investments are generating measurable business value.
According to Gartner's forecasts, total global AI spending is expected to reach $2.59 trillion in 2026, a 47% increase year over year.
A study released by PwC in January showed that only 12% of CEOs said AI has already delivered both cost reductions and revenue growth. Overall, 33% of respondents reported gains in either cost or revenue, while 56% said they have not yet seen significant financial benefits.
This issue has become more prominent as external attention grows on AI token consumption—the amount of AI processing companies pay for when using large language models.
Palantir CEO Alex Karp recently said that many business leaders are increasingly frustrated with the economics of LLMs, questioning whether rising token costs translate into return on investment.
"Enterprises are tired of this," Karp said in an interview with CNBC.
Karp made these remarks while promoting Palantir's own technology, acknowledging that his company has a commercial interest in this debate.
In the blog post, Friar introduced what she calls an "effective intelligence per dollar" scorecard, a framework built around four factors: whether the technology performs meaningful work, the cost per successful task, the reliability of output results, and whether each dollar of AI investment generates more value as usage grows.
"Tokens only create value when they translate into work people can use," Friar wrote. "As models become more capable, they can take on longer, more complex tasks: maintaining context, performing multi-step reasoning, collaborating across tools, and adjusting along the way."