A logistics company is building a Power Apps canvas app for warehouse clerks to process incoming supplier delivery notes. Clerks photograph each paper delivery note with a tablet, and the app must automatically extract the supplier name, delivery date, and line-item quantities into editable form fields before the clerk confirms and submits. The delivery notes come in dozens of different supplier layouts. The architect wants the fastest path to add this capability directly within the canvas app without training a fully custom vision pipeline. Which AI capability should the architect design into the app?