What building an AI-native finance function taught me

TL;DR - OpenAI CFO Sarah Friar's first-person account of running finance as an "AI-native" function, distilled into five lessons spanning automated forecasting, controls, and measuring AI ROI. It matters as a practitioner's view of what changes operationally when a back-office function is rebuilt around LLM tooling rather than bolting AI onto existing workflows.

• Framed as five lessons from lived deployment inside OpenAI's own finance org — an executive viewpoint/op-ed, not a product launch or research result.
• Named application areas: automated forecasting (planning/close cycles) and financial controls, implying AI is used in workflows where auditability and accuracy are hard requirements.
• Explicitly raises AI ROI measurement, signaling that justifying spend and quantifying productivity gains remain open problems even at an AI-first company.
• Content available here is only the summary blurb, so the specific five lessons, tooling stack, and any metrics are not verifiable from the provided text.