video1.0<iframe src="https://www.loom.com/embed/cf20dc7d8246425d94f285071e723828" frameborder="0" width="1728" height="1296" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12961728Loomhttps://www.loom.com12961728https://cdn.loom.com/sessions/thumbnails/cf20dc7d8246425d94f285071e723828-1deb2b8abd692457.gif121.03This Loom demonstrates a Malaysia-specific AI intent receipt extraction app using Cloud Sonnet 4.6 and the Anthropix tool use API. It shows a live e-voice receipt where fields populate from streamed JSON, correctly separating ssd and service charge rather than placing them in one tax bucket, noting ssd replaced gst in 2018 and service charge has its own line. The system returns a confidence score for every field, flagging anything under 85 percent for human review, following an audit pattern. In testing with 8 Malaysian and international receipts, it achieved 8 out of 8 on merchant and total currency and 7 out of 7 on date and SST where present.