Post by Invisible Technologies

A merchant's first order can't happen until their menu is live. Grubhub processes thousands of menus every month from restaurants, convenience stores, and local shops in a variety of different formats: handwritten notes, PDFs, Website URLs, photos and more. Manually extracting and reformatting every item, price, and item modifier to get it consumer-ready was arduous and manual. Every minute of delay was a day a merchant wasn't generating revenue. Grubhub partnered with Invisible to optimize the process. Together, we built a pipeline that captures menus the moment they're submitted, automatically classifies each by format, and routes it to custom AI agents that transform it on Grubhub's exact menu structure. Human specialists review the outputs to ensure accuracy. The results: 👉 ~1,000 menus processed per month 👉  75-80% reduction in extraction handling time 👉  98% quality rate We’re making AI work for marketplace operations. Link to the full case study in the comments.

Video Content