We’ve been thinking about what modeling should look like in an AI-native research stack.
Most AI modeling tools take the same approach: an agent writes code to populate or edit an Excel workbook. As the model grows, the agent has more numbers to track, more cell locations and formulas to get right, and more sources to preserve, all while following an ever-growing set of instructions for each company’s reporting and modeling nuances. The result becomes slow, brittle, and difficult to audit.
Implied takes a different approach. We give the agent an AI-native spreadsheet where AI lives in the formulas.
Instead of remembering and reproducing every number, the agent builds the model using AI formulas that populate values and link each one back to its source. Analysts can inspect sources, edit the model directly, and export it to Excel with formulas intact.
Modeling is one use case of a much bigger foundation.
The sheet is integrated into Implied’s chat, agents, live events, and workflows. Those connections make it possible to take the work further:
- Across models: Build or update 20 models in parallel, in the background.
- Across data: Have your agent incorporate alternative data daily without you opening a sheet.
- Across live events: Incorporate management’s latest guidance into forecasts as a call unfolds.
- Across your coverage: Compare commentary with consensus and flag potential gaps, company by company.
The video above is a first look at building a model from scratch and updating it after earnings. Over the next few weeks, we’ll show more of the workflows this foundation enables.