Manufacturers are betting big on physical AI, with three-quarters expecting it to reshape warehouse, assembly and production operations. But a new report from Tata Consultancy Services (TCS) reveals a critical gap: nearly half of organisations still have no clear accountability structure for when these systems fail.
Investment Is Accelerating, Not Slowing Down
The findings come from TCS's Future-Ready Manufacturing: TCS Physical AI Readiness Report 2026, which surveyed 300 manufacturing C-suite executives and vice presidents across North America and Europe. The report was published alongside the Farnborough International Airshow 2026.
Notably, no surveyed organisation plans to reduce its physical AI investment, and 26% intend to increase spending. Rather than replacing the workforce, 71% of manufacturers see physical AI as a way to help employees work more safely, efficiently and productively, supporting redeployment rather than displacement.
"As intelligent machines take on repetitive, physically demanding or hazardous tasks, employees can be redeployed and retrained for roles involving supervision, exception management, maintenance, safety oversight and process optimisation."— Anupam Singhal, President of Manufacturing, TCS
Proving Physical AI in the Real World
TCS is working to bridge the gap between digital ambition and physical deployment through initiatives like the TCS Physical AI Gemini Experience Center in Troy, Michigan. The centre shows manufacturers how Google's multimodal Gemini models integrate with robotics, edge computing and digital twins to automate complex industrial tasks.
Early results are measurable: an agricultural technology manufacturer using quadruped-based hazard patrolling has achieved a 90% reduction in safety incidents and a 30% cut in operational downtime. An automotive manufacturer improved first-time-right weld rates by 10–20% using AI-powered weld-quality prediction, while an AI-driven digital twin helped a construction and mining equipment manufacturer boost throughput by 25%.
Who Owns the Risk of Physical AI?
With 44% of manufacturers reporting no clear accountability for physical AI failures, TCS is calling for governance to become the industry's next critical priority. Singhal argues that ultimate accountability for how a system is deployed and operated should sit with the manufacturer, under a clearly identified business or operational leader.
This does not remove the responsibilities of technology providers and other partners, each of whom should have clearly defined accountability for the design, implementation, security and performance elements within their remit. A workable framework, Singhal says, should assign a named owner to every deployment and define which decisions the system can make independently, which require human approval, and when control must be handed from the AI agent back to a person.
Clear escalation thresholds, adequate testing and the authority to pause or override the system need to be designed in from the outset, alongside any regulatory reviews and approvals required for predictable performance and outcomes.