OpenAI's chief financial officer Sarah Friar is urging companies to abandon the metrics they have used for decades to evaluate software and adopt a new yardstick for artificial intelligence — one built around work completed rather than the number of people using a tool.

In a blog post published Friday July 17, 2026, and elaborated in a LinkedIn post reported by Axios, Fortune and CFO Dive, Friar introduced what she calls a "useful-intelligence-per-dollar" approach. The core argument: the metrics that defined the SaaS era — seats, active users, renewals — fail to capture the value AI creates. For ongoing analysis of how enterprises are grappling with AI economics, follow our AI industry coverage.

The four questions

Friar proposes that organizations judge AI value by answering four questions:

1. Is AI completing work that matters?
2. What does each successful task cost?
3. Can people depend on the result?
4. Does each AI dollar produce more value as usage grows?

"For years, software was measured through adoption: seats, active users, renewals," Friar wrote on LinkedIn. "AI is different — it needs to be measured by work accomplished."

The framing recasts token consumption — the amount of AI processing a company pays for — not as a cost to be minimized but as raw material that only becomes valuable when it converts into usable output. "Tokens create value when they transform into work people can use," she wrote, arguing that as models grow more capable they can take on longer, multi-step tasks that maintain context, reason across tools and adapt as they go.

Why OpenAI is pushing this now

The pitch lands at an uncomfortable moment for the AI industry. Worldwide spending on AI is forecast to reach $2.59 trillion in 2026, a 47% increase year-over-year, according to Gartner figures cited by CFO Dive. Yet evidence that the spending is paying off remains thin. A PwC study released in January found that only 12% of CEOs said AI had delivered both cost and revenue benefits. Another 33% reported gains in either cost or revenue, while 56% said they had seen no significant financial benefit so far.

That gap is producing visible frustration at the top of the corporate ladder. Palantir CEO Alex Karp recently told CNBC that many business leaders are "tired" of the economics of large language models, questioning whether rising token costs are translating into returns — comments Karp made while promoting Palantir's own technology and acknowledged as commercially motivated.

For OpenAI, which sells the tokens in question, reframing the conversation around outcomes rather than consumption is strategically vital. If customers measure success by work accomplished, then paying more for a frontier model that completes a complex task reliably looks like efficiency, not waste.

A shift away from the seats-and-renewals mindset

The "useful-intelligence-per-dollar" idea aligns with a broader rethinking of how AI should be priced and measured. Traditional enterprise software was sold per seat; usage-based AI pricing breaks that model, because a single seat can now consume vastly different amounts of compute depending on the task. Friar's four questions effectively ask buyers to build internal accounting around tasks completed and cost per successful task — a much harder number to track than headcount, but one that reflects what the technology actually does.

Not everyone is convinced the new yardstick favors buyers. Critics note that "work accomplished" is slippery to define and audit, and that vendors have every incentive to classify AI output as successful. Dependability — Friar's third question — is particularly hard to measure when models produce confident but incorrect answers, the phenomenon known as hallucination. Without independent verification, the metric risks becoming a marketing tool rather than an accountability mechanism.

Still, the proposal reflects a real shift in the conversation. After two years of breathless adoption metrics, executives are demanding evidence that AI spend is earning its keep. The companies that can answer Friar's four questions with data — whether they use OpenAI's models or a competitor's — will be the ones whose budgets survive the inevitable scrutiny.

Sources

  • CFO Dive, "OpenAI pushes new yardstick for measuring AI investments," July 17, 2026
  • Axios, "Exclusive: OpenAI's CFO pitches a new way to measure AI's value," July 17, 2026
  • Fortune, "OpenAI's CFO: 4 questions that reveal if your AI spend is paying off," July 17, 2026

Stay ahead on enterprise AI

The economics of AI are being rewritten in real time. Bookmark our homepage for the latest AI business coverage.

Read more AI news →