Physical AI Will Reshape the Economics of Automation

The rapid advancement of frontier technologies is both remarkable and inspiring. What once seemed like science fiction is now within reach.

Physical AI is creating excitement because it bridges digital intelligence and the physical world. As intelligence is increasingly embedded into machines, it is reshaping industrial automation. Unlike traditional goal-based automation, physical AI systems can adapt dynamically to changing scenarios using the same hardware but retrained “brains,” enabling huge flexibility and efficiency gains.

This capability fundamentally changes automation economics and has important implications for capital expenditure. Instead of replacing entire production lines, companies can retrain physical AI systems to handle new requirements with minimal hardware changes, making it easier to support next-generation products and get them to market quicker. In traditional manufacturing environments, human workers often require weeks of training before they can operate effectively on the shop floor. With physical AI systems, training can be conducted overnight in simulated or digital twin environments, leveraging reinforcement learning where robots practice performing tasks thousands of times under varying conditions.

This is game-changing for supply chain agility, as the shift toward nearshoring and onshoring introduces new operational challenges. Moving production out of established hubs, such as China, often results in efficiency losses—typically in the range of 4% to 15%. These losses frequently occur even when production is relocated to alternative lower-cost locations, such as Mexico, India, or Vietnam. Yet the demand for low-cost production remains unchanged, creating a structural tension between resilience and productivity. AI systems, whether physical or agentic, can help bridge the gap by offsetting productivity losses in new manufacturing locations and enabling more consistent performance across regions.

Aging populations are an additional challenge, making it increasingly difficult to sustain historical labor-supply levels. Companies can no longer rely on indefinitely relocating production to lower-cost countries because they are experiencing the same demographic pressure as countries in the West. There is a need to accelerate the adoption of new solutions, and robotics and physical AI systems present one.

That’s not to say adoption is easy.

Since 2023, I’ve been leading an AI transformation initiative at the global manufacturing company Foxconn. The work includes workflow identification, AI platform design, organizational structuring, talent strategy creation, and even incentive design to make the change sustainable.

Physical AI has been part of the transformation strategy from the beginning, and part of our work was a pilot on a brownfield printed circuit board assembly line, where space constraints made automation infeasible. One operator had to manage up to nine irregular components. By training physical AI systems in a simulated environment to recognize and place components accurately, we overcame this challenge. While it may sound simple, it required advanced kinematic, spatial AI, and vision language models that were trained repeatedly in a virtual environment before being deployed on the line.

When implementing AI—either physical or agentic—companies shouldn’t start with minor workflows or launch pilots that don’t impact the core business. AI should be applied where it can deliver real, scalable impact—from individual agents to entire agentic systems—capturing all the multimodal data and feeding it back into the system to retrain the models for improving precision and performance over time.

Talent transformation is also key. No longer can business experts get by without understanding AI or can data scientists be disconnected from operations. Companies need hybrid talent—people who bridge both worlds and focus on areas with the biggest business impact. That alignment is critical to organizational buy-in.

Companies must rethink their legacy IT infrastructures. AI isn’t just another tool—it’s a fundamental shift in how people and systems interact. Moving beyond humans prompting agents, companies are deploying autonomous agents capable of initiating and orchestrating workflows. That calls for new architectures, tech stacks, and change management strategies to support widespread adoption.

CEOs need vision, resilience, and the willingness to personally lead their company’s AI transformation. This isn’t just another tech initiative—it’s likely the most significant change an organization will undergo. That’s why it’s critical not to view AI as a cost but an asset. When we institutionalize organizational knowledge and operational know-how into proprietary AI models, we build sustainable and defensible enterprise value. And physical AI is not just a tool for efficiency—it is becoming a core enabler of resilience, adaptability, and long-term competitiveness in an increasingly complex and dynamic global environment.