OpenAI Finance Pro Uses ChatGPT to Create Interactive Forecasts

Kyle Kober, a member of the strategic finance team at OpenAI, has demonstrated a novel application of ChatGPT Work that promises to transform how financial forecasts are reviewed and manipulated by leadership. Previously, Kober spent considerable time manually updating spreadsheets and rerunning numbers to respond to executive queries. Now, he utilizes ChatGPT Work to automate much of this process, turning dense spreadsheets into interactive 'sites' that allow for real-time scenario testing and comparative analysis.

Transforming Financial Data with AI

Kober outlines a five-step process for converting a standard forecast spreadsheet into an interactive review tool. The first step involves opening the ChatGPT Work desktop app or web interface and linking the forecast spreadsheet. Crucially, this is accompanied by providing product context for the decision being evaluated. This ensures the AI understands the underlying business drivers and the purpose of the forecast.

The full discussion can be found on OpenAI Youtube's YouTube channel.

The second step is to connect the spreadsheet to the tool. Kober shares the link to his team's forecast and provides context behind the decisions being evaluated, such as a revenue forecast for FY26 and Week 4 retention results. This data-rich input allows ChatGPT Work to accurately interpret and represent the financial model.

Next, Kober defines the audience, the decision to be made, and the deadline. For instance, he specifies a prompt for 'tomorrow's CFO review' and asks the AI to help decide whether to update the Week 4 retention assumption based on its impact on forecasted revenue and gross margin. This structured input guides ChatGPT Work in tailoring the output for specific stakeholders and decision-making needs.

Interactive Sites for Dynamic Analysis

The core of Kober's demonstration lies in the creation of an 'interactive site.' He instructs ChatGPT to build a site where leadership can compare forecasts and test specific assumptions. These sites are described as being 'between a spreadsheet and a presentation,' offering a balance of simplicity for executive review and interactivity for dynamic analysis.

Unlike static slide decks, these AI-generated sites can run different scenarios in real time. Kober showcases how leaders can interact with the site to compare the annual plan with Q1 and Q2 outlooks, adjust key drivers like 'Week 4 retention' or 'New user acquisition,' and immediately see the impact on revenue and gross margin. This capability is vital for informed decision-making, as it allows for immediate visualization of how changes to assumptions affect the overall financial outlook without any risk of altering the original, authoritative spreadsheet.

The interactive site also includes features like an audio briefing. This offers a concise summary of changes, the rationale behind them, and the decision status, all generated on demand. Leaders can listen to this briefing to quickly grasp the key insights before diving into the interactive elements.

Ensuring Accuracy and Approval

The final step in Kober's process is a crucial validation phase. He emphasizes the importance of checking all results generated by the AI against the original model. This verification step ensures the integrity of the data and the accuracy of the AI's calculations and interpretations. Before officially approving any proposed changes, Kober reviews the exact proposed modifications, maintaining a human oversight in the decision-making loop.

Kober concludes by highlighting the efficiency gained. This method allows financial teams to transform complex forecast spreadsheets into user-friendly, interactive tools that empower executives to explore different scenarios and make data-driven decisions without the risk of breaking or corrupting the underlying financial models.