Series: AI × Supply Chain build breakdown
Stack: Claude Code · Shopify export · Xero export · Python · Google Sheets
The problem
B2B invoices live in Xero, the orders live in Shopify, and they drift. Missing invoices, wrong prices, subscription orders that never crossed over. Catching it by eye every month is slow, and the misses are real money walking out the door.
What I built
Drop the two month-end exports in a folder, run one script. It finds the files itself, lines invoices up against orders, spits out a workbook, then Claude reads it and writes me a findings note sorted by what matters. I only look at the real problems.
Where it failed
The first runs over-flagged like crazy. Half the “mismatches” were nothing: €0 sample orders, units of measure logged differently in each system, a few invoices raised straight in Xero on purpose. It only got useful once I taught it those patterns so they stop being noise. A subscription order that doesn’t sync to the accounts isn’t a fire, it’s a known gap. Flag it and move on.
The actual job isn’t “find the differences”. It’s “find the ones that mean something”. Most are noise.
Leverage rating: 5 / 10
It kills a recurring part-time job and catches money that used to leak every month. Recurring plus financial, that’s a 5.
Related: clean, reconciled commerce data is exactly what the Shopify Ops Starter installs. The script is forkable, email me.