How to use AI to find insights in your sales data
Export a report from your POS or accounting software, remove names and card details, upload it to an AI assistant and ask specific questions. Check its math.
To get insights from data you already have, export a report from your POS, booking or accounting software as a spreadsheet, remove customer names and payment details, and upload it to an AI assistant like ChatGPT, Claude or Microsoft Copilot. Then ask it specific questions: which products make the most margin, which days are slowest, which customers have stopped coming. Check its totals against your own report before you act on anything.
What data do you already have?
More than most owners think:
- POS or online store: every sale with date, time, items and amount
- Booking software: appointments, no-shows, cancellations and repeat visits
- Accounting software like QuickBooks or Xero: income, expenses and overdue invoices
- Google Business Profile and review sites: what customers say about you
- Spreadsheets: job logs, inventory counts, quotes
Most of these can export a CSV or Excel file from a reports screen.
How do you do it, step by step?
- Pick one question. "Which services bring in the most profit per hour?" works better than "analyze my business."
- Export the report that answers it, covering at least a few months.
- Clean it up. Delete columns with customer names, emails, phone numbers and card details. Replace names with a customer number if you need to tell people apart.
- Upload it using the attach button in ChatGPT, Claude or Gemini. OpenAI's help pages describe this data analysis feature for spreadsheets and CSV files.
- Ask it to check the file first: "List the columns, date range, total sales and any blank or duplicate rows." If its total doesn't match your report, stop and fix that.
- Ask your question, and ask for a table or a chart.
- Ask follow-ups like "Show that by month" or "What would change if I dropped the bottom five items?"
What questions are worth asking?
- Which items or services have the best and worst margin?
- What are my slowest days and hours, and how slow?
- Which customers spent the most last year, and which haven't been back in 90 days?
- How does this quarter compare with the same quarter last year?
- From these 200 reviews, what do customers praise and complain about most often?
The customer questions feed straight into targeted marketing with your customer data.
What can go wrong?
AI tools can misread a column or make a math error, and they say it with confidence. Check every number you plan to act on against your own report. Ask it to show the formula it used.
Privacy is the other risk. Keep names, contact details, health information and card numbers out of consumer AI tools, and use a business plan that doesn't train on your data. Is it safe to put customer information into ChatGPT? covers what's safe to upload.
How do you make it a habit?
Pick three or four questions that matter every month and ask them the same way each time. Better still, put the answers in a weekly numbers report that builds itself, so you're not exporting files by hand.
We set up reports and safe AI analysis on the data your business already has through our reports and customer data service.
Common questions
- Can ChatGPT analyze an Excel or CSV file?
- Yes. ChatGPT lets you upload spreadsheets and CSV files, with tighter limits on the free plan, and it runs calculations and draws charts from them. Claude, Gemini and Microsoft Copilot in Excel can do similar work.
- Is it safe to upload business data to AI?
- Remove customer names, contact details and anything covered by privacy rules before uploading, and use a business plan that doesn't train on your data. Totals by product, day or location are usually fine to share.
- What questions should I ask AI about my sales data?
- Specific ones you'd act on: which products have the best margin, which days and hours are slowest, which customers haven't been back in 90 days, and how this quarter compares with the same one last year.
- Can AI tell me why sales went down?
- It can show where they went down, like a product, a day of the week or a customer group, and suggest possible reasons. It doesn't know about the road construction out front, so treat its reasons as ideas to check.