Learn with AI
Check an AI forecast in a live workbook before sharing it
Turn a short sales example into a formula-based forecast, label its assumptions and test a changed month. Keep the workbook as the result, not only the chat answer.
Separate observations from assumptions
January, February and March sales are 120, 180 and 240 in B2:B4. These are illustrative observations. A forecast based on three points does not establish seasonality, demand or a reliable trend. Label forecast rows explicitly so a teammate cannot mistake them for recorded sales.
Ask for an inspectable calculation
Ask the in-app Pro assistant to add a forecast using the average absolute increase, preserve the original values and explain the assumption. The changes are 60 and 60; the average increase is 60, so April is 300. A chart makes the pattern visible but does not validate the forecast.
Inspect the formula in the workbook
For this small exercise, =B4+((B3-B2)+(B4-B3))/2 gives 300. Inspect the formula bar rather than accepting a number quoted in chat. Check that forecast cells refer to the observed data and are not merely typed constants.
Challenge the assumption
Change March to 210. The increases become 60 and 30; their average is 45, and April becomes 255. If the displayed forecast remains 300, check whether the formula or referenced sheet is wrong. Restore 240 after the test. A different forecasting method may be valid, but it needs its own explanation and expected result.
Keep the document useful after the conversation
Save the workbook to your account and invite a reviewer to discuss the assumptions. Use a view-only public link only if the sample data is suitable for public access. The Arte Sheets ChatGPT plugin has been submitted for review and is not yet publicly approved; access to a configured test connector is separate. When available to your account, that connector can edit the shared workbook, but the rendered results still need checking in the editor. You can practise this workflow now with the in-app assistant.
Try the example
Use the original values or change them to test your understanding.