Operational problem
Rising order volume does not establish that robots are already autonomously packing orders. Duhands said on November 3 that Poomgo’s cumulative domestic and overseas volume for January through September 2026 rose 60.3% year on year. According to ZDNet Korea, the company began training a physical AI system earlier this year and is currently focusing that work on packaging tasks. It plans to combine initial operation with remote human intervention, expand autonomy in stages, and pursue a proof of concept to test cost per transaction. Those are development plans, not verified commercial robot performance or demonstrated cost savings.
Evidence to request
The company separately reported that existing packaging automation equipment at its XFC fulfillment center, which opened in May, lifted hourly packing output per worker by about 90% versus manual work. It also reported a third-quarter daily peak of about 22,000 orders. These company-provided figures must not be attributed to the robot now being trained. A buyer evaluating a trial should request the definition of completed orders, matched product mixes, human intervention time, exception rates, damage and rework rates, and fully loaded cost per order under comparable conditions.
Practical deliverable
Maintain separate columns for the current automated line and the robot training or PoC. Record the product mix, carton dimensions, hourly completed orders, remote interventions, recovery procedure, safety owner, and quality results in each column. This is an editorial procurement checklist, not a Duhands performance target.
Hold conditions
Do not claim autonomous deployment, labor replacement, or a robot-driven cost reduction without operating logs and like-for-like cost and quality data. This update distinguishes the operating XFC equipment from the newer robot learning program.
About the Author
PackingMaster: PaperPackLog editor covering the paper packaging industry through public evidence and practical procurement, quality, and operations questions.