Key Points
- 1.OpenAI uses ChatGPT and its Sites feature to streamline the monthly financial forecasting process.
- 2.Sites allows users to generate and host interactive, no-code web applications without managing underlying infrastructure.
- 3.The generated forecasting tool displays data science baselines, prior forecasts, and aggregate scenario projections alongside visual trendlines.
- 4.Finance teams can adjust forecast curves at the individual account level directly within the user interface.
- 5.Modifications made to specific account forecasts automatically recalculate and update the aggregate model in real time.
Summary
Jackson from the OpenAI data science team explains how his team collaborates with strategic finance to streamline the monthly forecasting cycle. Finance teams frequently face a time-consuming scramble at the start of each month while waiting for updated actuals, modifying complex spreadsheets, and consulting product, marketing, and go-to-market teams to assemble a reliable forecast for the CFO. To reduce this friction, OpenAI created a workflow that produces an interactive user interface for cross-functional collaboration.
Using ChatGPT Work, Jackson initiates the process with a prompt requesting the model to pull the latest actuals, update the data science baseline, and build an interactive forecast application using Sites. Sites is an internal OpenAI capability enabling staff to create and host no-code web tools without needing dedicated server infrastructure. The prompt also requests account-level scenario testing and the integration of contextual data from internal documents and communications.
The generated application presents a quarterly overview displaying actual performance, the data science baseline, the prior forecast, and an adjustable finance scenario. It includes an aggregate revenue trajectory graph for the remainder of the fiscal year, allowing analysts to visually evaluate upcoming trends against historical benchmarks.
Below the aggregate view, the interface provides an account adjustment matrix for granular analysis. Demonstrating the workflow on a major account named Apex Inc., Jackson shows how analysts can drag the scenario trendline directly within the interface. As individual account adjustments are made in the browser, the entire financial model updates in the background to reflect changes in aggregate company revenue immediately.
Monthly forecasting challenges
Finance teams typically experience a monthly scramble waiting for actuals data, updating spreadsheets, and aligning with product, marketing, and go-to-market teams before presenting forecasts to leadership. OpenAI encountered the same friction and aimed to improve the workflow through collaboration tools.
Generating interactive forecasting apps with Sites
Using ChatGPT, a prompt is submitted to ingest the latest actuals, refresh the data science baseline, and build an interactive forecast site. Sites functions as a no-code platform used internally at OpenAI to create and host web applications without infrastructure management.
Aggregate and account-level forecast adjustments
The resulting application provides an aggregate view comparing baselines, past forecasts, and customizable scenarios with visual trend graphs. Users can also inspect individual client accounts, such as Apex Inc., and drag scenario lines in the interface to dynamically recalculate overall company projections.
Worth watching for
Finance professionals, data scientists, and operations teams interested in building automated, interactive forecasting tools with AI.
- chatgpt
- financial-forecasting
- data-science
- no-code
- openai