More Warehouse Operations Adopt AI-Enabled Vision Systems

Gartner recently predicted that by 2027, half of companies with warehouse operations will replace traditional scanning-based counting with artificial intelligence-enabled vision systems.

These new systems will use 3D cameras, computer vision software, AI-powered pattern recognition, and machine learning. However, integrating such systems will take more work and time than it might seem. The software will need training data and companies might have to develop their own that is specific to their products and needs.

The upside for companies is that, with the proper work, the combination of pattern recognition and machine learning might allow companies to reduce the amount of manual work necessary for processing and warehouse operations. Visual processing can extend beyond visual spectra into various levels of infrared, for example, to perform quality control and advanced levels of sorting unavailable to human vision.

“AI-enabled vision systems will propagate quickly in warehouse operations as the value proposition is so evident; not only for inventory management, but also monitoring that can identify safety issues and ergonomic problems for workers in real-time,” Carly West, senior director analyst in the Gartner supply chain practice, said in prepared remarks.

The projection was the result of a Gartner survey of 506 supply chain professionals “with input into management and operations processes from December 2023.” A fifth of them had already adopted AI-enabled vision systems. What will drive more adoption, according to Gartner, is improving cost and performance of the camera hardware and improving the pattern recognition.

Gartner has three suggestions for those looking to adopt such systems for their warehouses. First, there is no single vendor that has the technology and experience to address all the possible needs of different companies. As prices drop, companies might be able to experiment with various systems to find combinations that work for them.

Next, they suggest starting with the easy use cases — cycle counting or worker safety monitoring — that are “tailor-made” for AI vision systems. Then learn lessons early on to apply to future use cases.

Third, mature companies with good manual processes should consider more advanced applications like process ergonomic monitoring.