Breaking

OpenAI CFO Proposes 'Useful Intelligence Per Dollar' Framework to Measure Enterprise AI Returns

July 17, 2026 at 09:43 EDT

Sales & Business
Foundation Models

OpenAI Chief Financial Officer Sarah Friar has introduced a new framework aimed at helping companies gauge the value of their AI spending, urging finance leaders to measure "useful intelligence per dollar" rather than raw token prices or total outlays.

OpenAI · CFO Framework

Stop counting tokens. Measure "useful intelligence per dollar."

OpenAI CFO Sarah Friar wants finance leaders to judge AI spend by the value of the work completed — not sticker price or total outlay. The core test: is the value of what AI does growing faster than the cost of producing it?

questions to score AI return

benchmark figures — the framework is fully conceptual

~30h

/ month saved from a 30% workflow reduction, per OpenAI's own ROI guidance

The four questions

Is real work getting done? Customer inquiries handled, code shipped, contracts reviewed.
What is the cost per successful task? Model usage plus retries and human review.
How accurate is the output? Fewer corrections and escalations mean higher return.
Does value per dollar rise as usage scales? Value should climb — not costs exponentially.

Old lens vs new lens

Price per token

Cheapest sticker wins — routine tasks routed to the least-costly model.

→

Value per dollar

Higher-performing models argued to be cheapest by outcome.

The pitch

Judged by results — accuracy, tasks completed, time saved — premium models can beat cheap ones on total value, even at a higher price.

The skeptics

"Value over price" can double as a case for pricier models. Soft gains are hard to quantify — and only a small share of firms report substantial AI returns.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.