Real cases

Industrial artificial intelligence in rawhide chew treat factories

A rawhide chew treat factory resembles no other food plant. At the head there is process chemistry inherited from tanning — soaking, liming, dehairing, fleshing, bleaching. In the middle there are hundreds of people hand-forming knotted bones, sticks, donuts, braids and rolls, an operation where value comes from human dexterity and not the machine. At the center there is a drying of twenty-four to seventy-two hours that rules the whole plant's capacity. And at the tail there is the discipline of an exporting food industry: food safety, chain security and the destination country's labeling rules.

This sub-sector's pains live exactly at the joints between those four realities. The production order and the drying load sheet are still paper, and take a shift to exist for the system. The drying chamber's panel is twenty years old and its readings die on the screen, so nobody correlates the curve with the final moisture or with the complaint arriving weeks later from the distributor's warehouse. The laboratory works in a spreadsheet the ERP does not see. And in a plant manufacturing for several brands, the line clearance between one brand and another is signed without objective evidence. On top of all that weighs the capital pain: a container rejected or recalled at destination, dragging several client brands down at once.

These pages hold twelve cases landed on this sub-sector, organized by the IRIS architecture: Connect's capture modes, the ERP integrations with their license savings angle, the coder's stitching, the in-line Edge vision cases, the auditable evidence pack and the flagship case coordinating everything against the recall at destination. None replaces anyone: in all of them, the person stays in command and keeps signing.