22% of revenue, paying 19 days late
The owner exported 14 months of invoices as a simple file. The file held client, dates, amount, and date paid. She removed bank details and cut client names to an initial plus an industry. She uploaded the file to Claude and asked direct questions. She also asked Claude to flag any odd pattern, even one it could not explain. She ran the same check in ChatGPT.
Start with one job only. Fill in the three lines. Keep the bot on that job.
The one job this bot does: ______________________________
What it may do on its own: ______________________________
What it must ask me first: ______________________________
Two lines from this case, side by side. No new figures.
The owner exported 14 months of invoices as a simple file. The file held client, dates, amount, and date paid. She removed bank details and cut client names to an initial plus an industry. She uploaded the file to Claude and asked direct questions. She also asked Claude to flag any odd pattern, even one it could not explain. She ran the same check in ChatGPT.
The work took under an hour with follow-ups. One client was 22% of invoiced revenue and paid 19 days late against 30-day terms, every time. Her day rate had not changed in 3 years, and overheads had risen about 19%. She sent a price review letter to five legacy clients and a payment-terms email to the late payer. One flat-fee add-on paid under minimum wage per hour, so she retired it. Claude also called a dip seasonal, but the dip was her 3-week holiday.
The after line is the source's claim, not our measurement.
smalloffice.space · Get a case like this in your inbox each morning. Subscribe free
Read this case on your phone now, and keep it for later. Type your email. We send this case to that address, and you get the next cases each morning too.