Intelligence in motion: The dawn of physical AI

For years, when we spoke about AI, we were really talking about software. It helped banks detect fraud, retailers forecast demand and enterprises make sense of tonnes of data. The intelligence stayed inside computers—and humans still did the physical work.

But now we are entering a phase where AI doesn’t just analyse the world, but it interacts with it. Machines can now observe what is happening around them, make decisions in real time and act without waiting for constant human instructions.

Whether it is a robot on a factory floor, an autonomous warehouse vehicle or a laboratory running experiments with minimal intervention—intelligence is now moving into the physical world.

And, this is what many are calling ‘Physical AI’.

Although the term sounds futuristic, most of the underlying technologies already exist.

AI models, robotics, computer vision, sensors, edge computing and digital twins have all matured over the last few years. What is new is how they are coming together. Instead of operating as separate technologies, they are becoming part of a single intelligent system that can continuously sense, learn and respond.

The pace is picking up much faster than many expected.

Several industry estimates suggest the Physical AI market could grow at more than 45% annually over the next decade. That kind of growth shows that this is no longer a research project. Companies are beginning to spend serious money because they see measurable business value.

Digital twins at the centre of transformation

Manufacturing, perhaps, offers the clearest picture of where things are headed.

Traditionally, factories relied on fixed automation where machines performed repetitive tasks extremely well – but at the same time struggled when something unexpected happened. So, whenever there was a defect in a product, or a change in material quality or a disruption on the production line – people were often required to step in.

But, Physical AI changes that equation.

For instance: Computer vision systems can identify defects that might escape the human eye; robots are becoming better at handling variation instead of repeating identical movements all day; digital twins allow manufacturers to test changes virtually before making them on the shop floor.

In simple terms, the factory is gradually becoming a system that learns from every production cycle instead of simply repeating one. The result is no more just higher productivity—but it is a manufacturing operation that becomes more resilient with time.

Digital twins deserve particular attention because they sit at the centre of this transformation. A few years ago they were largely simulation tools—but, today they are evolving into living digital replicas that continuously absorb data from real operations. A MarketsandMarkets study says that the global

market for AI-powered digital twins could expand from around $31 billion in 2025 to more than $225 billion by 2032. This growth reflects how central they are becoming to industrial decision-making.

Building living industrial systems

Now, the same pattern is emerging in life sciences.

Drug discovery has always been dependent upon highly skilled scientists, but much of the work around experimentation is repetitive and time-consuming. AI-powered laboratory automation is helping researchers run more experiments, reduce manual variability and shorten development cycles. Scientists spend less time managing processes and more time interpreting results.

Outside the lab, the impact may be even bigger. Life sciences companies manage some of the world's most complex supply chains. Medicines and biological products move through tightly regulated manufacturing and distribution networks where delays can have real consequences.

Physical AI makes these networks more adaptive by continuously tracking conditions, anticipating disruptions and recommending operational changes before small problems become major ones.

The impact of Physical AI is far beyond manufacturing or pharmaceuticals. Logistics companies are using AI to optimise routes as conditions change throughout the day; energy providers are monitoring infrastructure more intelligently, while agriculture is using AI to improve irrigation, monitor crop health and reduce waste.

These examples somewhere proves that intelligence is no longer confined to enterprise software – but is gradually becoming part of the physical systems that power businesses every day.

This shift also changes what competitive advantage looks like.

For decades, competitive advantage for organisations rested on scale, capital and operational efficiency. While these factors still matter, but increasingly, the winners will be those that can combine digital intelligence with physical operations. Having data alone will no longer be enough—with the real differentiator being how quickly that data turns into action.

The human edge

Technology, however, is only one piece of the puzzle.

Deploying Physical AI is not simply about buying robots or installing new software. It requires organizations to rethink how work gets done, data needs to flow across systems, and infrastructure has to support real-time decision-making.

On top of that, engineers, operations teams and business leaders need to work far more closely than they have traditionally.

Because, ultimately, people remain central to the story.

There is a tendency to frame every AI conversation around replacement. In practice, the more interesting question is augmentation. The greatest value comes when human judgement combines with machines that can process information and respond at a speed that people simply cannot match. AI handles scale and repetition—whereas humans continue to provide context, creativity and accountability.

How ‘intelligence’ will shape our future

Every industrial revolution has changed the relationship between people and machines.

Let’s think of how steam power transformed manual labour, electricity reshaped manufacturing, and computing digitised business processes. Physical AI feels like the next chapter in that journey.

It is the beginning of a new operating model for industry. The opportunity is not just to automate more work—but to build organisations that continuously learn from the world around them.

We have spent years building AI that could think—the next decade will be about building AI that can act. And that may prove to be an even bigger shift.

(The author is CEO & MD, Birlasoft. Views are personal.)