Fujitsu, SoftBank Robotics and two Japanese logistics organizations have started an artificial intelligence pilot aimed at standardizing how damaged home-appliance packaging is assessed during delivery.
The project, launched in Japan on November 2, will run through January 15, 2027. It is designed to determine whether image-based AI can help manufacturers, retailers and logistics providers make more consistent decisions when cardboard packaging is scratched, dented or crushed but the product inside may be undamaged.
Fujitsu said home appliances can be returned or redelivered because of minor packaging damage, with decisions varying among companies and delivery locations. The pilot will develop common assessment criteria from rules currently used by members of the Home Appliance Supply Chain Council, then test an AI model in logistics operations.
The initiative is being conducted by the Council for Sustainable Home Appliance Logistics, which comprises the Home Appliance Supply Chain Council and the Distribution Economics Institute of Japan. Japan’s Ministry of Land, Infrastructure, Transport and Tourism selected it for a subsidy program promoting the use of digital technology by shippers and logistics operators.
Combining image recognition and operational data
Fujitsu will contribute its Kozuchi AI platform and Amalgamation AI for Vision image-recognition technology. The company says the system can train with relatively small data sets, apply specified assessment conditions and provide reasons for its determinations.
SoftBank Robotics will provide a system that analyzes smartphone images of cardboard packaging, compares the visible damage with predefined criteria and records the result for later review.
The two logistics organizations will manage the project, develop industry-wide criteria and supply packaging images and operational expertise. Fujitsu and SoftBank Robotics will provide the AI technology and analyze the data.
The partners plan to measure the model’s assessment accuracy, test methods for standardizing decisions and evaluate whether the system can be used in real logistics settings. They will also analyze whether the approach reduces returns and redeliveries.
Pilot has not yet reported results
The announcement describes a validation project rather than a completed deployment. It does not provide a baseline return rate, expected cost savings or an estimated reduction in carbon emissions.
After the pilot, the partners intend to publish the packaging-damage criteria and consider whether the method could apply beyond home appliances. Any broader rollout will depend on the accuracy and operational feasibility established during the trial.
Fujitsu and Japanese builders test AI construction oversight