42 receipts matched, 85% perfect
SquidCircle is a small consultancy. It tested ChatGPT Operator on its own books. The agent got a folder of digital receipts for one month of business costs. It signed in to the accounting software. It matched each receipt to a bank transaction. It flagged any line that did not match. The test used the consultancy's own records, not a client account.
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.
SquidCircle is a small consultancy. It tested ChatGPT Operator on its own books. The agent got a folder of digital receipts for one month of business costs. It signed in to the accounting software. It matched each receipt to a bank transaction. It flagged any line that did not match. The test used the consultancy's own records, not a client account.
The agent matched 42 digital receipts. It handled 85% of entries perfectly. Vendor names, amounts, and tax categories came out right. The test saved about four hours of manual data entry in one afternoon. It failed on one blurry handwritten scan. The source is the consultancy's own test, with no client results. The source reports no limit beyond the scan failure and its own call for a human final check before money moves.
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.