Turn your spreadsheets into a database you can ask questions

Somewhere there is a folder of spreadsheets that grew up on their own: members on one sheet, payments on another, a third someone started for volunteers. The same person is spelled two ways, a column means something different on each tab, and every question starts with twenty minutes of filtering. In this recipe the agent turns them into one database with proper tables that link to each other. Then you, and the people you share it with, ask it questions in plain words.

What you need

• XataWorks, open, with the agent answering. If it is not yet, Getting started covers the one decision to make first.
• The spreadsheets, as Excel workbooks (.xlsx or .xlsm) or .csv files.
• To share the database with other people: XataWorks Pro.

1. Attach the spreadsheets and ask

Attach the files to a chat, and ask the agent to look before it builds:

Turn these spreadsheets into one database. First tell me what one row is on each sheet, which columns link the sheets together, and where the same thing is written different ways. Then propose the tables, how they link, and a value list wherever a column only ever holds a few answers.

The agent reads the files and answers in the chat, so you can correct the design while it is still only words. A sheet with “Paid”, “paid” and “PD” in the same column is exactly what to settle now.

2. Let it build and fill the database

That’s right. Build it as Club_Records, tidy the values as you go, and bring every row in.

A database name has no spaces, so “Club Records” becomes Club_Records. The agent creates the tables, fields, relationships and value lists, adds the rows (the rows go straight from the file into the database, not through the conversation) and opens the database beside the chat. Creating it asks you first; allowing that also lets the agent fill and change this database for the rest of the chat.

3. Check what arrived

How many rows went into each table? Are there payments that don’t match a member?

Look through a table or two in the database’s Table view. If you would rather load a very large file yourself, go to a layout for the table and choose More ▸ Import Records…, which takes CSV, tab-separated and Excel files and lets you match each column to a field.

4. Ask it questions

Now the questions that used to take an afternoon:

Who has not paid this year? What did we take in each month of 2026? Which members joined in the last ninety days and have never volunteered?

The agent answers from the records without opening anything, and reading your own database never asks. Ask it to show the matching records and it finds them in the open database. To keep or send an answer, ask for a report instead.

5. Share it

Choose the database’s name in the list to open its own page. Under Sharing, press Sharing… and add each person by email address, at Can view, Can edit, Can design or Admin. They can then open it from their own phone or computer. In their own XataWorks, their agent can look up its records too, and change them if they have Can edit.

Where this recipe stops

• The database is a copy. Once the records are in, changes to the old spreadsheets don’t reach it. Retire the spreadsheets, or import again with the option to update matching records.
• Calculated fields are worked out, not imported. A total or a count in the old sheet becomes a calculation or summary field in the database, so check it agrees with the sheet once.
• Sharing runs from your computer. People reach the database while the computer holding it is on, with XataWorks running and web access on. To keep it reachable, use Move to another of my computers… on its page to move it to a computer that stays on.
• Other people’s agents look records up. On a shared database an agent finds records by their values and can’t run the database’s own queries. A question over thousands of records is best asked on the computer that holds it.
• Nothing is lost by accident. Removing records always asks, even after you have allowed changes, and a snapshot is taken first.

Read more

• Databases: tables, relationships, importing, the database’s own page, and what the agent asks.