Reviewing cards after parsing
What to look at in the catalog preview so the AI never sells the wrong product at the wrong price.
The AI shows a preview of the cards before publishing. This is the cheapest moment to fix a data error: after publishing, these are the numbers customers are quoted.
Price and units
The most common problem is a price in the wrong unit. Check a few expensive and a few cheap items: if the file listed dollars, or a price per pack instead of per piece, the order of magnitude gives it away.
Name
The name has to be recognisable to a shopper. Technical strings like “SKU-40021 / carton” are worth replacing: the AI will not find the product when someone asks in plain language.
Stock
Items with zero or empty stock should not go on sale as available. If you keep stock in a POS rather than in the file, syncing is more reliable than maintaining it by hand.
Categories
Check that products landed in sensible groups. How accurately the AI offers an alternative when an item is missing depends on it.
Duplicates
The same product uploaded twice at different prices produces different answers to the same question. Remove duplicates before publishing.
Once corrected, publish the catalog — the bot starts answering from the updated data.