Turn order emails into spreadsheet rows

Every hour this workflow reads the order emails sitting in your inbox, has an AI read the order out of each one, and adds a row per ordered item to your spreadsheet - so nobody has to re-key them.

Each row is Customer, Order reference, Item, Quantity, Requested delivery date and the Gmail message ID the order came from, so any row can be traced back to the email behind it. Rows are only ever appended; nothing already in the sheet is changed. Values are written exactly as they were read, so a value that starts with '=' cannot turn into a spreadsheet formula and nothing is silently reinterpreted.

Accuracy is the point, so the workflow refuses to guess. If the model is not certain, or any of the five values for an item is missing, the whole email is left out of the sheet and labelled 'Order needs review' for a person to pick up - a wrong quantity costs far more than a missing row. Orders are written all or nothing: one unusable line means none of that email's lines are added. Emails that were handled are labelled 'Order logged', and both labels are excluded from the search, so a re-run can never record the same order twice. One email that cannot be read is logged and skipped; the rest of the batch still goes through.

To set it up, paste your spreadsheet link, name the tab to append to, and say how to find your order emails using Gmail search terms - for example subject:(order OR "purchase order"). Give the tab a header row matching the six columns above before the first run.