An agent created through Co-Pilot Studio is not able to read the excel properly

This issue is likely due to how the Customer Insight template agent in Copilot Studio is interpreting your Excel data. The generic chat uses a broader, more flexible model for file understanding, while the template agent may have stricter or different parsing logic.

Troubleshooting steps:

1. Check Excel formatting: Ensure your Excel file has clear headers and consistent data types. Avoid merged cells, hidden columns, or complex formulas.
2. Agent configuration: Review the Customer Insight agent’s settings—some templates expect specific column names or data structures. Try uploading a simplified version of your Excel file (just a few rows and columns).
3. File upload method: Make sure you’re using the “Upload file” option within the agent’s chat, not outside or in a different context.
4. Test with another template: Try creating a new agent with a different template or a blank agent and upload the same file to compare results.

Hello,

Thank you for the input, have removed every format in the excel. There are no merges as well. Also, it doesnt have any complex formula just a simple excel pivot with 60 months of data. Still the agent is throwing wrong results. Tried the same with csv as well. Please suggest if there is anything else that can be done

Hi

• Do you have one sheet and one table only, or are there multiple? Are you using plain numbers in your Excel?
• Have you tried with another very simple test Excel (just a few rows/columns) to confirm if the issue is format-related?
• When you added the file in Add knowledge, did you give it a detailed description (columns, units, time range) so the agent can interpret it correctly?
• Also, is the switch “Allow the AI to use its own general knowledge” enabled? If it is, try disabling it so the agent is forced to ground its answers only on your Excel.

Hi Nikhil Anant Dambal

It seems the issue isn't with your Excel file itself, since the generic chat reads it correctly, but with how the Co-Pilot Studio agent is interpreting it. Agents like the Customer Insight template expect a specific data format and column names, so if your headers or data types don't match, it can produce wrong results.

Pivot tables may also cause issues because some agents can't read derived data properly—they often require raw values. To fix this:

Copy your pivot table results and paste them as values in a new sheet.

Ensure the column headers exactly match what the template expects.

Remove any extra spaces or special characters.

Test with a smaller subset of data to confirm the agent reads it correctly.

For CSVs, make sure decimal points, commas, and date formats are consistent.

Check if the agent has any preprocessing or aggregation settings enabled, as these can sometimes conflict with your data.