Ask a question of a spreadsheet instead of building another pivot table

Upload your spreadsheet and ask questions in plain English instead of wrestling with pivot tables and formulas.

For anyone whose data lives in files · 7 steps · 7 min

How it works today.

You have a spreadsheet with months of sales data, customer records, or inventory counts. Someone asks a question that should take thirty seconds to answer, but you know it means opening Excel, creating a pivot table, dragging fields around until the layout makes sense, then building a formula to calculate the percentage or trend they actually want.

By the time you finish, you have a worksheet full of intermediate calculations and a chart that answers this specific question. Next week, when they ask a slightly different version, you start over. The data is right there, but extracting insights feels like archaeology.

Before you start.

All of it has to be true, or step one fails in a way that is annoying to debug.

Your data is in a CSV, Excel, or Google Sheets file you can download
The file has column headers and consistent data types in each column
You have access to a generalist AI agent like Claude, ChatGPT, or Gemini
The spreadsheet contains no sensitive personal information you cannot share with the AI service

The steps.

Upload your spreadsheet to the AI agent
Drag and drop your CSV or Excel file directly into the chat interface. Most agents can read common spreadsheet formats and will confirm they can see your data structure. If the file is too large, export just the columns you need for your question.
02
Ask your question in plain business language
Type the question exactly as someone would ask it in a meeting. The agent will examine your data structure and identify which columns and calculations it needs.
Paste this
```
Looking at this data, [your specific question]. Show me the answer and explain how you calculated it.
```
03
Request the specific format you need
If you need a chart, table, or summary for a presentation, ask for it directly. The agent can create visualizations or format results as tables you can copy into reports.
Paste this
```
Present this as a table I can paste into a slide deck, with the top 5 results and percentages rounded to one decimal place.
```
Verify the logic with a spot check
Pick one result you can manually verify and check it against your original data. This catches cases where the agent misunderstood a column name or made an incorrect assumption about your data structure.
05
Ask follow-up questions without re-uploading
The agent retains context about your data within the same conversation. Ask related questions, request different time periods, or drill down into specific segments without starting over.
Paste this
```
Now show me the same analysis but only for [specific time period/category/region] and compare it to the overall average.
```
Export results you want to keep
Copy tables directly from the chat into your documents, or ask the agent to format results as CSV data you can paste into a new spreadsheet. Save the conversation link if your platform supports it.
Document the question and method for next time
Write down the exact question phrasing that worked well and note any data preparation steps you took. This creates a template for similar analyses and helps colleagues ask effective questions of the same dataset.