OpenAI's CFO says the AI scorecard is useful intelligence per dollar - completed work that matters, full cost per success, dependability, and whether each dollar buys more as you scale.
Via OpenAI: A scorecard for the AI age
On July 17, 2026, OpenAI CFO Sarah Friar published a plain question every owner already feels in the gut: how do we get more value from AI spend? For years, software success meant adoption - seats bought, users active, licenses renewed. AI breaks that scoreboard. A cheaper token can still lose money if it needs three retries and a partner's afternoon to clean up.
Friar's proposed scorecard is useful intelligence per dollar. It asks four things: Is AI finishing work that matters? What does each successful task actually cost? Can people depend on the result? Does each AI dollar buy more output as usage grows?
That is not a vendor slogan for SMEs. It is the language a managing partner or practice owner needs before the next budget cycle - especially if staff already pasted client work into personal ChatGPT accounts and nobody can show what it produced.
The direct answer: stop defending AI by license count. Start defending it by cost per completed job in the system where the work already lives.
Token price is a trap metric. Friar's point is blunt: a lower-cost model may burn more attempts, more latency, and more human review. A pricier model may finish in one pass. The business cares about the full cost of a successful outcome - model spend plus staff time, rework, and escalation - measured against the value of that outcome.
She walks through concrete workflows: customer issues resolved, code changes shipped, contracts reviewed, time returned to people. For a finance team, 'done' might mean a forecast pack that reconciles before the meeting - not a chat window that felt productive.
OpenAI's own framing of the GPT-5.6 family (Sol / Terra / Luna) is a tiered starting point, not a permanent routing rule. The economics of the full task should choose the model. On the Artificial Analysis Coding Agent Index, Friar cites GPT-5.6 Sol reaching a new high on DeepSWE v1.1 while using fewer output tokens than another leading model - an efficiency claim worth verifying against your workloads, not copying wholesale.
Dependability has a price tag too. Track ready-to-use vs needs-correction vs needs-escalation. If AI drafts everything and humans rewrite everything, you bought a very expensive autocomplete.
For regulated firms - law, accounting, dental, advisory - add one more column: what data left the building on the way to 'done.' A cheap successful task that leaked client content is not a win on the scorecard; it is a liability with a productivity costume.
None of that requires an enterprise AI office. It requires an owner who stops confusing 'we bought Copilot' with 'we finished more billable work.'
'The ultimate scorecard for the age of AI could be looked at as Useful Intelligence per Dollar.' - Sarah Friar, OpenAI CFO
This is exactly what a Workflow ROI Audit is for: find where AI saves real time and money, and where it only accelerates the wrong work. You get a prioritized, costed roadmap - not a tool pitch.
If usage is already informal, pair that with a Shadow-AI Risk Assessment so the scorecard is not built on data walking out the side door. The destination is governed operations - someone accountable for whether each AI dollar still buys useful work as the firm scales.
Vendor-neutral on purpose. The scorecard is the point, not the logo on the invoice.
Friar's four questions are a useful filter for any firm still reporting AI progress as 'seats deployed.' Useful work. Cost per success. Dependability. Value at scale. If you cannot answer those for one core workflow, you are not ready for the next subscription - you are ready for a baseline.
Start with a free AI assessment when you want a clear read on where AI is paying and where it is theater: Request your free AI assessment.
This article summarizes publicly reported information and is for general informational purposes only. It does not constitute legal, tax, financial, investment, security, or compliance advice. AgentsROI.ai is not a law firm, accounting firm, or registered investment adviser. Facts, pricing, statistics, and product capabilities cited here reflect the sources listed at the time of writing and may change. Readers should verify current information independently and consult qualified professionals regarding obligations specific to their industry, jurisdiction, and circumstances-including applicable New York State and New York City requirements. AgentsROI.ai may have commercial relationships with vendors mentioned; where material, such relationships are disclosed. Nothing in this article is an endorsement of any specific AI product, model, or provider.
