{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/778688c45f5b4b38b1a9da38a912eb69\" frameborder=\"0\" width=\"1242\" height=\"931\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":931,"width":1242,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":931,"thumbnail_width":1242,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/778688c45f5b4b38b1a9da38a912eb69-224b6cd8924bb90e.gif","duration":132.629333,"title":"Andy: Local AI Invoice Reconciliation Review","description":"This Loom explains how Andy uses local AI image recognition to validate invoice and order reconciliation and avoid hallucinated errors. The speaker describes a common back office task of comparing invoices to what was ordered, noting that if AI is unsure it must flag the result for human review instead of guessing. Andy runs 100 percent locally for the QVEC track, taking an invoice image and a CSV of order details to determine whether it is reconciled automatically with high confidence or needs review due to OCR limitations. The system aims to reduce manual effort by having humans only check the uncertain cases."}