OpenAI’s CFO ditched the monthly close scramble for a ‘zero-day’ model

When OpenAI’s finance chief arrived two years ago, the team was tiny and the company was exploding. The tools were world-class, but the work was stuck in a familiar loop: hunting for data, reconciling spreadsheets, and building slide decks to explain what happened last month. The ambition now isn’t just a faster close — it’s a zero-day close and a forecast that updates itself continuously. That means a real-time, traceable view of the company’s actuals against approved plans, where AI drafts the variance explanations and flags the exceptions that need a human’s judgment. Finance still owns the sign-off, but the period-end scramble to reconstruct the business starts to vanish.

The shift required more than plugging in new software. The team ran a hackathon that paired sales engineers with finance staff, which yielded IR-GPT, a custom tool grounded in approved investor relations materials to handle diligence questions. Similar GPTs are now being built for procurement and tax. The lesson for any CFO: give people secure, capable AI and let them identify the work worth transforming — bottom-up experimentation has to meet a top-down strategy focused on the decisions that actually move the business.

Redesigning the work itself is the real unlock. The CFO describes changing the fundamental unit of work from assembling inputs to making decisions. Instead of pulling actuals from one system, purchase orders from another, and accruals from a spreadsheet, the team is building a continuously reconciled foundation where every variance can be traced to the underlying activity. That same foundation powers a rolling forecast that shows not just what might happen, but which choices could alter the outcome. It’s a stark departure from the static spreadsheet model that still dominates most finance functions.

This is a concrete playbook for an AI-native function, not a vague sales pitch. The five lessons are pragmatic: broad access to tools, structured experimentation, redesigning workflows around decisions, building clear accountability into automated steps, and measuring the dependable work AI completes. The quiet implication is that the finance talent who thrive will be the ones who learn to build the tools their work requires, carrying their expertise further instead of being buried by process. Whether a zero-day close is genuinely achievable at scale remains an open question, but the direction of travel is hard to ignore — finance is becoming a real-time function, and the teams that don’t redesign their work will simply be left explaining what they missed.