Report: Berkshire Grey enhances AI capabilities for warehouse robots

Automated Warehouse Online reported on October 29, 2026, that Berkshire Grey had enhanced the AI capabilities of BG Pilot, its software for warehouse robots. The company said it would draw on insights gained from a decade of warehouse operations and billions of robotic picks.

Automated Warehouse Online’s report points to Berkshire Grey using experience from warehouse operations and robotic picking to enhance BG Pilot’s capabilities. The company says the software is intended for warehouse tasks involving variations in packaging and SKUs, but the excerpt does not specify the scale of improvement or the conditions covered. Concrete evaluation results will be needed to assess the impact of deployments.

The report focuses not on an update to the robots themselves, but on incorporating operational knowledge accumulated in warehouses into software capabilities. The figures of a decade of operations and billions of picks are based on the company’s account, but the article excerpt does not specify the scope of data collection or the methods used for training and evaluation. The company says the software addresses differences in packaging and multiple SKUs, but provides no figures showing the extent of improvement under specific working conditions. In previous coverage, we reported that Berkshire Grey had opened a customer innovation center in Haarlem, the Netherlands article on the opening of the European center. The opening of the facility and the software enhancement reported here are separate developments, but they represent different layers: a site for deploying and operating warehouse automation, and capabilities that use operational data. Turning data into actual operational improvements will depend not only on software updates, but also on validation under the working conditions of each deployment site. In terms of data infrastructure, the approach differs from the integration of search, curation, annotation and evaluation covered in the article on Voxel51 3.0. This report concerns applying warehouse robot operating experience to software; it is not an announcement of a platform for managing and evaluating data. The two should not be treated as the same system. Data acquisition, quality management, training and evaluation, and application in the field should be considered separately. The broader industry implication is that warehouse automation may be differentiated not only by equipment performance, but also by how effectively real-world operational data—including variations in packaging and SKUs—is incorporated into software capabilities. However, the report’s figure that new business opportunities increased by more than 65% over the past six months is not defined or explained in the excerpt, and should not be equated with an increase in deployments or revenue. The next points to verify are the specific features enhanced, evaluation metrics by task, target warehouse conditions, data handling and results in real-world operations. See the original report for details.