Over the past two years OpenAI rebuilt a small finance team supporting rapid company growth into an AI‑native finance function focused on real‑time reconciliation and continuously updated forecasting. The organization set two bold goals: a zero‑day close and automated, continuously refreshed forecasts.
Why they changed and what they aimed to achieve
Finance has become a real‑time function. Beyond faster book closings or more frequent forecasts, OpenAI aimed to let leaders see the business as it changes, help them act sooner, and give finance teams more time to influence outcomes. The ambition was to make AI a core way of working rather than a bolt‑on tool.
Starting point and initial steps
When the author joined two years ago, the finance team was small and many close and forecasting tasks were manual and repetitive: finding records, explaining variances, and assembling decision inputs. Although OpenAI had access to advanced AI tools, the team needed to learn how to redesign finance around those tools.
Two central ambitions: zero‑day close and continuous forecasting
The zero‑day close aims to provide leaders with a reconciled, traceable, real‑time view of the company’s financial position. Continuous forecasting builds on that reconciled foundation to show how the business is changing, what could happen next, and which decisions might change the outcome. The operating model connects approved spending plans, general‑ledger actuals, purchase orders, accruals, and transaction details into a continuously reconciled view where each variance traces back to underlying activity. AI prepares initial explanations and highlights exceptions; finance validates numbers, applies judgment, and signs off.
Five practical lessons for CFOs
- Broad access plus structured experimentation
People need the freedom to explore AI in the context of their own work. OpenAI ran a finance hackathon that included sales engineers and asked participants to bring tasks they wanted to transform. One output was IR‑GPT, a custom GPT grounded in investor relations’ approved materials to answer diligence questions quickly and consistently. Teams also began building custom GPTs for procurement and tax. The hackathon turned AI from an abstract capability into a working tool in one day: identify a recurring task, build a solution, test it with colleagues, and iterate. The lesson for CFOs: combine bottom‑up experimentation with clear top‑down priorities.
- Redesign work from source data to decision
Finance teams spend enormous effort assembling decision inputs—reconciling data across systems, explaining variances, producing slides. AI changes the unit of work and enables redesigning the entire path from source data to decision. For the close, the goal is a continuously reconciled model where approved inputs, source checks, and finance‑owned review replace the post‑period scramble. That reconciled foundation supports interactive forecasting and scenario planning so leaders can see what changed, why, and which actions could alter the quarter or year.
- Enable finance professionals to build the tools they need
OpenAI research cited in the article reports that 40% of finance professionals’ specialized AI use involves work outside traditional finance and 22% involves engineering‑related tasks. Team members use ChatGPT Work and Codex to build custom dashboards and tools, moving away from static Excel and PowerPoint to live dashboards that sit on the full context and data of the business. One teammate with no prior coding experience used Codex to build a tool that converts monthly advertising forecasts into weekly and daily plans, accounting for weekdays and holidays while keeping every number tied to the approved model. The people closest to the problems can now shape solutions and rethink how the work is done.
- Governance and human‑centered control
IR‑GPT illustrated the importance of control: building a custom GPT on vetted sources can produce a strong first draft in seconds for investor diligence, but humans remain central. Investor relations reads and edits the draft, adding judgment and ensuring consistency. CFOs should work with IT and governance teams to define which data an AI can access, what actions it may take, when approvals are required, and when issues must be escalated. Every output should link to a reliable source, every forecast should include a clear explanation, and every change to an approved baseline should require finance authorization. Practical controls—usage limits, budgets, role‑based access, model‑routing rules, and approval thresholds—are straightforward to implement.
- Measure outcomes, not consumption
CFOs need an AI scorecard grounded in operating performance. Buying more seats or consuming more tokens is not in itself meaningful. Useful metrics for finance include: cycle time to close, share of transactions reconciled automatically, number of exceptions requiring review, time to explain a variance; and for forecasting: forecast accuracy, refresh frequency, time to produce a new scenario, and the quality of decisions supported by the forecast. A cheaper model is not always more economical if a better model delivers reliable answers with fewer attempts and less review.
Why this matters for CFOs and the company
Finance sits at the center of strategy, capital, data, risk, and performance, giving CFOs a unique view and a mandate to lead AI transformation. An AI‑native finance function is defined by faster cycles, stronger controls, better decisions, and more time for judgment. The practical path is: start with a consequential workflow, give people the tools, help rebuild the work, keep accountability clear, and measure outcomes. The ultimate aim—zero‑day close and continuous forecasting—expresses a broader ambition: a finance team that understands what is happening as it happens, shows leaders what could happen next, and helps the company make better decisions sooner.
The result is greater capacity in finance, clearer visibility for the CFO, and a working model for how the entire company can convert intelligence into durable value.
