CASE STUDY 1 E-commerce fulfilment (continued) Innovation in action (continued)Amazon’s application of physical AI enables robots to perceive, reason and act within the real-world environment of a fulfilment centre, while remaining anchored to the deterministic controls required for safe, at-scale operations. Together, these systems unlock tasks that rule-based robots alone could not tackle, easing repetitive work for employees while expanding automation into previously unstructured areas of the operation. Impact Early deployments, including in Amazon’s next-generation fulfilment centre in Shreveport, Louisiana, have yielded significant results – including improved workplace safety, creation of 30% more skilled jobs onsite, 25% faster delivery to customers and 25% efficiency improvement that will drive lower costs to customers. A new generative AI foundation model, coordinating the movement of the entire mobile robot fleet across the fulfilment network, further improves fleet travel efficiency by 10%. Foundation and learningsThree foundational elements have enabled Amazon to scale intelligent robotics across its global fulfilment network: 1 Decades of computer-vision monitoring data produced an internet-scale catalogue of physical-world images and trajectories, comprising the critical training fuel for reliable physical AI models. 2 A people-first feedback mechanism, centred on front-line employees who work alongside robots, allows engineering teams to directly iterate with users to ensure that automation solutions reduce friction rather than introduce new islands of automation that do not seamlessly integrate into broader workflows. 3 Amazon designs, manufactures and deploys its robotics solutions in-house. Co-location of hardware, AI and operations teams enables rapid build–measure–learn cycles and accelerates rollout. Together, these enablers – massive data, employee co-design and an integrated build pipeline – helped Amazon to scale physical AI safely and consistently across the world’s largest fulfilment network. With this shift, the need for higher-skilled workers (such as roles in reliability maintenance engineering) increased by 30%. Amazon has implemented the Career Choice11 programme to provide tuition support and reimbursement for books and fees for degrees and certifications designed to upskill employees into in-demand and high-paying career paths, and through its mechatronics and robotics apprenticeship programme employees can advance their careers and increase their hourly wages by up to 40%.12 CASE STUDY 2 Electronics manufacturing Adaptive robotics automating highly versatile assembly operations In response to rising labour costs and the global trend towards localized production, Foxconn is promoting the vision of a “scalable AI-powered robotic workforce”. It outlined a three-stage AI-enabled factory development process: digital twin simulation design, human-robot collaboration and, finally, embodied intelligence robotic factories. This phased transformation reflects Foxconn’s broader ambition: not only to automate tasks, but to architect an intelligent production ecosystem to unlock higher industrial value and operational scalability across its global manufacturing network. Physical AI: Powering the New Age of Industrial Operations 14

Physical AI Powering the New Age of Industrial Operations 2025