Is Your Data Ready for Copilot? A Checklist for Small Businesses

Many small businesses are being told that AI will save them hours every week. Some of that is true. But there is a step most sales pitches skip: an AI assistant is only as good as the files, folders and systems it reads from.

If your sales figures live in three versions of the same spreadsheet, Copilot does not know which one is right. If a customer appears as "ACME Ltd", "Acme Limited" and "acme", it may count them as three customers. AI does not fix those problems. It repeats them, quickly and confidently.

This checklist covers the six areas we look at first when a business asks whether it is ready for AI.

1. Do you have one version of each key number?

Pick your five most important numbers: revenue, gross margin, cash in the bank, active customers, and whatever else drives your business. For each one, ask where the official figure lives.

If the answer is "it depends who you ask", fix that first. An AI assistant that searches across your files will find every version and may pick the wrong one, or blend them together.

2. Is your data in proper tables?

AI tools read structured data far more reliably than a spreadsheet laid out for printing. Watch for these signs of trouble:

• Titles, logos or notes above the column headers
• Merged cells and blank rows used for spacing
• Subtotal and total rows mixed in with the data
• More than one table on the same sheet
• Column headers that are blank, repeated or renamed from month to month

In Excel, the quick win is to convert each data range into a formatted table (Home, Format as Table) with one header row and one row per record. Copilot in Excel works best with data laid out this way too.

3. Are your categories consistent?

AI answers questions like "which product line grew fastest?" by grouping your data. If the same thing is spelled several ways, the groups are wrong and so is the answer.

Look for label variants in your key columns: "Edinburgh" and "Edinburgh " with a trailing space, "N/A", "n/a" and blank, or supplier names typed freehand. Add a dropdown list (Data, Data Validation) to any column people type into.

4. Are numbers and dates stored as numbers and dates?

A number stored as text looks normal on screen but is ignored by SUM, and can be misread by an AI tool. Dates typed in mixed formats, such as 03/04/2026 and 2026-04-03 in the same column, can be read as different months.

5. Who can see what?

Microsoft 365 Copilot works within each person's existing permissions. That is good for security, but it also means Copilot can surface any file a person can already open, including files that were shared too widely years ago.

Before rolling it out, review who can open your finance, HR and payroll folders, and remove "anyone with the link" sharing from sensitive files. This is a data protection job as much as an AI one.

6. Can you check the answer?

Treat AI answers like the work of a new member of staff: helpful, fast, and in need of checking. For any number that feeds a decision, someone should be able to trace it back to a source file and reproduce it.

If nobody on the team can do that today, a simple dashboard built on the same clean data gives you a trusted reference point to compare AI answers against.

Where to start

You do not need to fix everything at once. Start with the one spreadsheet your team uses most, run it through a free check, and fix what it finds. Then move on to the next one.