OpenAI showed how finance teams use Codex for reporting and scenario planning

OpenAI on May 12, 2026 demonstrated how Codex can be embedded in the daily work of financial teams, not just in engineering processes. In a new OpenAI Academy material, the company gathered five practical scenarios: from monthly business reviews and board packs to model verification and scenario planning.

Where Codex Saves Time

The main idea of the material is simple: financial professionals don't need to start with a blank page or manually assemble the first version of documents before a meeting with the CFO or board of directors. Codex is proposed to be used on top of already familiar work artifacts — closing tables, revenue and expense dashboards, forecast updates, comments from metric owners and past presentations. As output, the team receives not an abstract chatbot answer, but a draft of a specific file that can be reviewed, corrected and passed further along the process.

• Preparing narrative for monthly business review based on period close, forecast and owner comments
• Cleaning and QA of financial model before presentation to management or board
• Updating a regular CFO or board reporting pack with fresh data
• Building variance bridge by revenue, margin, expenses, EBITDA and cash flow
• Recalculating base, downside and upside scenarios for forecast and operating plan

In all five cases, OpenAI emphasizes the same thing: Codex takes real input data from the current work cycle and accelerates the first assembly. This removes routine work, but does not eliminate financial expertise. The team spends less time on mechanical consolidation of materials and more time on verification of figures, formulation of conclusions and preparation of questions for the business. Especially where deadlines are tight and sources are scattered across spreadsheets and chats.

How the Process Works

In OpenAI's examples, Codex is asked to work not based on a general task description, but on a specific set of files and channels. For MBR this could be close workbook, revenue dashboard, forecast update, previous slide deck and owner notes. For scenario planning — operating model, headcount plan, cash forecast, latest actuals and leadership notes. It is important that for each material number, the system is given the requirement to specify the source: workbook tab, dashboard, tracker or owner note. This makes the result suitable for review, not just for reading.

There is a separate emphasis on integration with the company's work tools. OpenAI lists Google Drive, SharePoint, Box, spreadsheets, documents, presentations, Slack, Teams, Gmail and Outlook Email as typical points from which Codex can take context and where it can return results. That is, this is not a demonstration on synthetic dataset, but a scenario where the financial team asks to update the May CFO pack, verify the FY27 model or explain the forecast-to-actual movement for April in terms of existing processes.

Where Control Is Needed

Despite automation, OpenAI separately emphasizes the boundaries. If Codex cleans a model, it can safely fix structure, references, formulas, period labels or sign conventions, but should not silently change business assumptions. If a number without confirmation appears in a variance bridge or reporting pack, it should be marked as unconfirmed. For high-stakes review, the system should return not only the updated file, but also a QA memo prioritizing risks: what corrections were made, what assumptions remain disputed and what cells or tabs the finance owner should verify.

In essence, OpenAI proposes not a replacement for the FP&A team, but an accelerator of its cycle. Codex handles the first pass through the data, finding gaps between sources and preparing the document structure well. But responsibility for interpretation, final wording, assumption approval and sending materials to management remains with people. For finance, this is critical: the cost of error in a model or board pack is too high to leave the decision entirely to the machine.

What This Means

Codex is increasingly positioned not just as a tool for developers, but as a work environment for functions with heavy operations and many files. For financial teams, this is a particularly telling case: if AI can assemble MBR, verify models and quickly recalculate scenarios based on real data, then the next wave of AI adoption in companies will not only go through IT, but also through the office of the Chief Financial Officer.