AI has already reshaped digital work. It can generate code, automate IT workflows, support service teams and accelerate decision-making across the enterprise. The next frontier of Physical AI brings intelligence into the environments where products are made, assets are maintained, goods are moved and infrastructure is operated.

HCLTech’s The AI Impact Imperatives, 2026 report defines Physical AI as “the application of AI, including GenAI, Agentic AI, machine learning and autonomous decision-making systems, to perceive, understand and act upon physical environments and operational processes in the real world.” It encompasses intelligent systems across manufacturing, industrial operations, logistics, infrastructure and field operations that can observe conditions, make autonomous decisions and execute actions to achieve business objectives.

In a recent discussion on the report’s findings, Mark Beccue, Principal Analyst - AI at Omdia, spoke with Tamas Foldi, SVP, Kinetic AI at HCLTech, about why Physical AI is gaining momentum and how enterprises can move from experimentation to production.

Physical AI is intelligence in action

“Physical AI is AI which directly interacts with the physical world,” said Foldi. “There are two dimensions to it. One is physical operations, and you can think about all these verticals which are operating in the physical world traditionally, like manufacturing, aerospace, medical devices, warehousing and logistics.”

The first dimension is operational intelligence. This includes improving productivity, throughput, asset health and uptime in physical environments. The second is the movement of intelligence into machines, including robotics, autonomous systems and autonomous vehicles.

“That’s where robotics, autonomous systems and autonomous vehicles come into the equation,” said Foldi. These systems can “sense, decide on the perceived reality of the physical world and act.”

This is what makes Physical AI distinct. It brings intelligence closer to real-world action, where systems can support decisions in the environments where work happens.

“With Physical AI, you are not just improving digital workflows,” said Foldi. “You are empowering your system with more intelligence.”

The opportunity is significant, but maturity is uneven

The report states that 90% of respondents agreed that Physical AI will be critical or important to organizational success over the next three years. However, it also found that respondents at organizations not currently exploring Physical AI, researching or evaluating it, piloting solutions and just deploying initial systems to production outnumbered those with many Physical AI systems in production by more than four to one (81% vs. 19%).

That gap between ambition and maturity is important. Enterprises see the opportunity, but many are still working out how to scale Physical AI safely and effectively.

“Physical AI is one of the biggest untapped opportunities for enterprises,” said Foldi. “Everyone is laser focused on digital AI.”

Momentum is building because several factors are converging at once. Models are becoming more capable. Agentic AI and world models can better understand surroundings, support spatial reasoning and orchestrate actions. Simulation and synthetic data are also lowering barriers to entry.

“In the past, collecting data was super expensive, sometimes unsafe,” said Foldi. “Now, with all these simulation capabilities, you can generate or augment your data, which reduces the time to market and also makes it cheaper.”

This matters because Physical AI depends on learning from physical systems. Simulation, real-world deployment and operational feedback can create a continuous improvement loop, helping organizations test, refine and scale AI-enabled systems with greater confidence.

Physical AI can transform operational performance

Physical AI has relevance across many operational domains. The report found that most respondents selected logistics and supply chain management (63%), field operations (59%), facilities management (56%) and manufacturing (55%) as areas that could benefit from Physical AI deployment.

The benefits also extend beyond the efficiency gains typically associated with digital AI. The report states that increased operational efficiency is the most widely seen benefit (56%). It also highlights benefits unique to Physical AI, such as upgrading and extending the life of physical assets (47%), meeting sustainability goals (46%) and improving physical safety (46%).

Foldi sees strong use cases in quality control, defect detection, predictive maintenance, process optimization, intelligent logistics, warehouse management and robotics.

“Predictive maintenance is huge,” he said. “You can also use robotic automation, like using drones or mobile platforms, to go into industrial manufacturing plants or mines, look for issues and do corrective actions.”

The bigger opportunity comes when Physical AI moves beyond detection and recommendation. A system may identify an asset performance issue, determine the right response and execute the corrective action itself.

“Traditionally, you close a workflow by creating a ticket and a human takes the action. With Physical AI, the system can take that corrective action itself, end-to-end,” said Foldi.

For enterprises, that closed loop is where Physical AI starts to change operational performance. It can help systems adapt and act in physical environments, provided those systems are integrated safely with the processes, controls and governance around them.

The hard part is everything around the model

The promise of Physical AI is significant, but deployment is complex. These systems often operate in real-time, mission-critical environments where safety, resilience, cybersecurity and regulatory compliance matter.

“We talked a lot about AI, but what is really hard in AI programs is not AI itself. It is actually everything around it,” said Foldi.

Physical AI needs to be integrated with legacy systems, operational processes, edge devices, data pipelines, simulation environments and enterprise platforms. Without that integration, organizations risk being left with isolated use cases that never reach scale.

Security is another major consideration. Physical AI can affect assets, workers and live operations, which means systems need to be reliable, governed and secure from the start.

“We are working in real-time environments,” said Foldi. “There are regulatory and safety requirements, human-robot collaboration and heterogeneous environments with very exotic protocols you need to integrate.”

That is why enterprises need to think beyond models. They need the infrastructure, governance and integration capabilities required to make Physical AI operational.

“The hard part is how you are going to integrate it with your enterprise systems in the real world,” said Foldi.

Partners can help accelerate production

The report found that 54% of organizations working with AI partners have deployed Physical AI systems to production, more than double the 26% recorded among organizations not working with partners. It also found that these organizations are more likely to have seen positive outcomes attributed to Physical AI, including improvements in safety and risk posture (75% vs. 48%), improved resource utilization rates (71% vs. 43%) and increased production line uptime (64% vs. 52%).

For Foldi, this reflects the complexity of the environment. Physical AI requires expertise across engineering, AI, robotics, integration, compute, edge deployment and operations.

“Physical AI is hard,” he said. “You need a lot of expertise which is not necessarily core to your business. Having an integrator which can help you through the process is super helpful.”

HCLTech supports this through capabilities including Kinetic AI for robotics, Vision X for edge perception and SmarTwin for digital simulation. These platforms are designed to help organizations accelerate development and focus on outcomes, rather than building every component from scratch.

“We can help customers jumpstart and make sure that we can be laser focused on the business outcomes and not on the technology,” said Foldi. “That’s what really matters.”

Design for scale from the start

One of the biggest lessons from early Physical AI programs is that small, isolated use cases rarely create transformational value. Enterprises need to identify high-impact areas where Physical AI can improve safety, productivity, asset performance, resilience or cost, then design for scale from the beginning.

“Start on the hardest business problems, not necessarily on the easiest one. Start with the most critical one, which can drive the most measurable ROI for your enterprise,” said Foldi.

Physical AI is still maturing, but its direction is clear. As intelligent systems move from digital workflows into physical environments, enterprises have an opportunity to rethink how operations run. The organizations that succeed will be those that connect ambition to execution, integrate AI into real-world systems and build the foundations needed to scale safely.

The next wave of AI impact will not only be seen on screens. It will be felt across factories, warehouses, field operations, infrastructure and machines that can learn, adapt and act.