OpenAI's CFO sets a zero-day close goal and a scorecard for AI
Sarah Friar's five lessons read like a finance operating manual. The scorecard section is the part to copy.
What happened
OpenAI CFO Sarah Friar published five lessons on August 10, 2026 from building an AI-native finance function. Her team set two goals, a zero-day close and continuous forecasting. She writes that the team is still building toward both. Everyone on her team builds AI dashboards and tools with ChatGPT Work and Codex.
Her controls are explicit. Define what data AI can access, what actions it can take, when approval applies, and when issues escalate. She also proposes a four-question scorecard for each AI workflow. One example in her post is IR-GPT, a tool her team built for investor relations work.
What it means
Read past the vendor angle. Friar sells AI, and she still writes, “AI accelerates the work. People own the result.” She also writes, “The cheapest model isn’t always the most economical.”
Review and rework count as cost, not only the model bill. OpenAI research she cites shows finance professionals already reach beyond finance work with AI. About 40% of their specialized use falls outside traditional finance, and 22% is engineering tasks.
Why it matters to you
The scorecard works for any finance team. For the close, she suggests cycle time, the share of transactions reconciled automatically, exceptions needing review, and time to explain a variance.
What to watch
Friar’s team is still building toward a zero-day close and continuous forecasting. Watch for OpenAI to say it reached either goal, and with what results. Watch whether other CFOs publish scorecards built on the same four questions. Cycle time and time to explain a variance are the numbers that will show progress.