Part 3: Using AI to see trouble coming · MODULE 13
You can build a bad payer report in about 20 minutes. Export your aged receivables and invoice history from Xero, replace customer names with codes, and paste them into ChatGPT, Claude or Gemini with a structured prompt. Then check the AI's figures against Xero before you act on them. The report tells you who to contact first, who's getting slower and how much of your cash is at risk.
7 MIN READ · UPDATED 10 OCTOBER 2026
IN THIS MODULE
What a bad payer report shows
Step 1: Export the data from Xero
Step 2: Anonymise it
Step 3: Run the prompt
Step 4: Check the answer
Step 5: Act on it
Common questions
What should a bad payer report show?
A useful report answers the questions you'd otherwise have to work out by hand from the aged receivables report.
Who to contact first, ranked by how much is overdue and for how long.
Which customers are paying more slowly than they used to.
How much of what you're owed is more than 90 days overdue.
Whether one customer owes a large share of your receivables.
A suggested next step for each customer: reminder, call, payment plan or letter of demand.
Step 1: Export the data from Xero
Open the Aged Receivables Summary report. Set it to age by due date, so the columns show how overdue each balance is, rather than how old the invoice is.
Export it to Excel or Google Sheets, then save it as a CSV file.
For trends, also export your sales invoices from the last 12 months with invoice date, due date, paid date, amount and contact.
The aged receivables summary tells you where things stand today. The invoice history tells you which way each customer is heading, which is usually the more useful of the two.
Step 2: Anonymise it before you paste
Customer names, emails and phone numbers are personal information when the customer is an individual, including sole traders whose business name is their own name. Australia's privacy regulator recommends not entering personal information into publicly available AI tools. The AI doesn't need names to do the analysis.
Sort the list by balance and replace each name with a code, such as C01, C02 and so on.
Keep the key that matches codes to names in a separate file on your computer.
Delete columns you don't need, such as email addresses, phone numbers and street addresses.
Module 17 covers privacy and AI settings in more detail.
AI PROMPT
You're helping me, the owner of an Australian small business, understand who owes me money and what to do next. My payment terms are [number] days. Customer names are replaced with codes.
I've pasted two things below:
A. My aged receivables summary from Xero, aged by due date.
B. My invoices from the last 12 months with invoice date, due date, paid date and amount.
Please give me:
A short summary of my receivables in plain English: total owed, total overdue, and the amount more than 90 days overdue, each as a dollar figure and a percentage.
A ranked list of customers to contact first, with the amount overdue, the oldest overdue amount, and a suggested next step: reminder email, phone call, payment plan or letter of demand.
For each customer in B, the average days from invoice to payment for each of the last four quarters, and a flag on anyone whose average has risen by more than 7 days.
Any customer who owes more than 25% of my total receivables.
A table showing your calculations so I can check them against Xero.
Use only the figures provided. Don't invent details. Some customers may be individuals, so keep any contact plan within the ACCC and ASIC guideline of no more than three contacts a week.
A. [Paste aged receivables summary]
B. [Paste invoice list]
Step 4: Check the answer before you act
AI tools are good at spotting patterns and writing summaries. They also make arithmetic mistakes, especially with long tables, and they can sound confident about a wrong number. A few quick checks catch most problems.
Compare the total owed and total overdue with the totals on your Xero report. They should match to the dollar.
Spot-check two or three customers' figures against Xero.
Look at the calculation table. If a customer's average days to pay looks odd, check whether a credit note or a part payment confused it.
Treat the suggested next steps as suggestions. You know which customers are going through a rough patch and which ones always pay at the end of the month.
EXAMPLE
Say the AI reports $58,050 owed in total, but Xero shows $59,050. The calculation table shows it skipped a row where the CSV wrapped "1,000.00" in quotation marks. Pasting that row again fixes the total. Without the table, you'd never have seen why the numbers didn't match.
Step 5: Act on the report
Contact the top five customers on the list this week, using the next step the report suggests or a better one if you know the customer.
Move anyone with a rising days-to-pay trend to shorter terms on new work.
If one customer owes a large share, keep their invoices on a tighter schedule and consider a deposit on their next job.
Run the report again in a month and compare. The trend over time is the most useful part.
Common questions
Can AI find bad payers in my Xero data?
Yes. If you export your aged receivables and invoice history from Xero, an AI tool such as ChatGPT, Claude or Gemini can rank customers by risk, spot customers who are paying more slowly and suggest next steps. Check its figures against Xero before you act, because AI tools can make arithmetic mistakes.
Is it safe to paste Xero reports into ChatGPT or Claude?
Replace customer names, emails and phone numbers with codes first, because names of individuals and sole traders are personal information. Check whether the tool uses your chats for training and switch that off. Australia's privacy regulator recommends not entering personal information into publicly available AI tools.
How often should I run a bad payer report?
Monthly is enough for most small businesses, with a quick look at the aged receivables report every week. Comparing each month's report with the last one shows which customers are slowing down before they become a problem.
