OpenAI Pushes 'Useful Work Per Dollar' Framework for Enterprise AI Measurement

OpenAI published two advisory pieces in mid-July 2026 arguing that traditional business metrics are inadequate for evaluating AI investments. [1][2] In the more substantive piece, CFO Sarah Friar introduced a four-dimension scorecard — useful work, cost per successful task, dependability, and return on compute — targeted at enterprise finance and executive audiences. [2] Both pieces are openly promotional, positioning OpenAI as a standard-setter for how enterprises measure AI ROI as agentic deployments grow.

If the framework gains traction, it shifts enterprise buyers toward measuring AI by output value rather than model price — a framing that favors agentic, higher-cost deployments where OpenAI competes. A vendor-defined measurement standard also gives OpenAI influence over how its own products are benchmarked against alternatives.

OpenAI published two pieces within three days in mid-July 2026, making the case that standard business metrics — cost per seat, token prices, headcount saved — do not capture what enterprises actually receive from AI deployments. The first, a general advisory on managing AI spend in the agentic era, frames the right lens as 'useful work per dollar': measuring the value of outputs rather than the price of inputs. [1]

The follow-up formalizes this into a named scorecard with four dimensions: useful work (volume and quality of tasks completed), cost per successful task (spend normalized to actual completed work), dependability (consistency and error rate), and return on compute (value extracted per unit of compute invested). [2] Sarah Friar, OpenAI's CFO, is the named author — a deliberate choice that signals the content is aimed at finance and executive audiences rather than technical buyers.

Both pieces are openly promotional. Their goal is to give enterprise customers a vocabulary that justifies higher AI spend, specifically on agentic workflows where task-completion value is easier to argue than per-token cost. Neither piece references independent measurement research, academic benchmarks, or third-party audits. The reactive searches generated by this thread returned mostly off-topic results — financial asset management content, general AI adoption statistics — and no substantive counter-voices from analysts or rival vendors.