[Knowfa Case Study] From AI-OCR to AI Agent: Improving Order Automation Accuracy by 'Teaching' AI Like a New Employee | Salt Kansai

Usac System Co., Ltd. announced that Salt Kansai Co., Ltd., a food wholesale company, has implemented the 'Knowfa Order AI Agent.' Salt Kansai deals with a wide range of food products, such as salt and sugar, and had long struggled with handling diverse order formats from various customers.

Previously, AI-OCR had limitations in recognition accuracy; issues like tilted FAX sheets or handwritten characters necessitated double-checking by humans. Furthermore, rules for unit conversions (e.g., kg to pieces) and customer-specific business practices relied on the experience of veteran staff, making standardization and knowledge transfer difficult.

With the introduction of Knowfa, the company succeeded in automating these 'tasks requiring judgment' by instructing the AI on rules via prompts. During test operations, it was confirmed that the AI could automatically split orders with complex delivery dates and enable junior staff to input orders with the same quality as veterans. Moving forward, the company plans to integrate Knowfa with the RPA 'Auto Job Meijin' to create a system that almost entirely automates the process from order receipt to entry into the core system, strengthening its response to future labor shortages.