Physical AI Emerges From A Technology Avalanche

Pascal Brier, Group Chief Innovation Officer for Capgemini, leads an internal group called Technology Innovation and Ventures (TIV) that is tasked with imagining the technology agenda for tomorrow. It’s an essential role in an organization that is trusted to help some of the world’s biggest and most complex organizations decide which technologies to invest in.

Brier’s team monitors and assesses a portfolio of around 1000 technologies at any given time that range from different flavors of emerging AI to automation and quantum computing. With these deep insights, they can help the firm’s consultants provide advice to clients and also shape their own future services.

The pace and volume of innovation right now is something Brier calls a technology avalanche. Anyone paying attention knows he’s right.

Not all technologies will amount to something. The trick, according to Brier, is to identify as quickly as possible those with value from those that are not going anywhere or are too early for primetime. That’s no easy feat, but it’s a function that his team specializes in.

Brier is confident that physical AI is ready and leaders must act now.

Why Physical AI Matters Now

What makes this moment different is the convergence and maturity of several technologies that include robotics, spatial intelligence, and compute. It’s a powerful mix of capabilities that is now enabling machines to autonomously perceive, reason, and act in the physical world.

Leading systems can increasingly walk, climb, pick, lift, and navigate autonomously in constrained environments, though capability still varies significantly by task and setting. In many deployments, these systems operate with growing autonomy rather than constant human control, though supervision and intervention mechanisms remain common.

Unlike previous generations of robotics that typically served a fixed function, these physical AI devices learn new tasks and evolve over time. It’s a fundamental difference that shifts machines from being limited tools to adaptive collaborators. Some robots can increasingly be repurposed through retraining and software updates rather than complete redesign.

Some physical AI will take the form of humanoids--robots that look like a person--but in the shorter-term machines will have a wide variety of configurations including specialized industrial robots, collaborative robots (cobots) that work alongside humans, and even drones and autonomous vehicles.

It’s a paradigm shift that will provide industries with capabilities that were previously impossible or uneconomical. 60% of executives interviewed for Capgemini’s report say physical AI will make previously impractical use cases viable in areas such as productivity, resilience, safety, and growth.

Industries and applications for physical AI will differ considerably, but in the near-term some uses are more obvious.

Physical AI Begins To Roll Out

Brier says the early adopters of physical AI include logistics, manufacturing, warehousing, and particularly hazardous environments. The latter has always made sense for robots which can protect humans from dangerous environments such as those high in toxic chemicals, intense heat, or high-radiation areas within nuclear facilities. The addition of intelligence makes these robots vastly more valuable in each instance.

A more liberal definition of physical AI includes both drones and autonomous vehicles. The application of these is already obvious in many contexts including transportation, deliveries, and military uses.

Science fiction has conditioned many people to imagine humanoid robots folding laundry, washing dishes, and becoming household companions. Brier sees progress in this area but says the time horizon for practical humanoids is longer than what most anticipate. The Capgemini report suggests that broad adoption is still about seven years out as a result of limitations in dexterity, reliability, cost, safety, and ROI issues.

Brier is also less concerned that physical AI could displace significant portions of the human workforce, at least in the medium term. Ironically, according to the Capgemini report, 74 percent of executives provide human labor shortages as their reason for interest in adopting robotics.

Brier argues that much of the fear of robots taking most human jobs is created by the lack of understanding of what they can currently do and how they will work alongside people.

Specifically, in his view work will be logically split between what humans and robots each do well. The emphasis will be on functional strengths. Tasks that must be completed quickly and require complex dexterity will be better suited to humans, whereas repetitive actions that include, for example, lifting heavy objects and low safety environments will be best for robots.

Why Waiting May Be Riskier Than Experimenting

According to Capgemini, around 79 percent of business leaders across industries are already embracing physical AI whether in full deployments or just to experiment and 65 percent expect to scale within five years. Evidence to date suggests entry barriers are falling, though organizations still face integration, safety, governance, and workflow challenges.

From his research and experience, Brier notes the technology is now mature enough to move forward and the costs for many organizations won’t be a restrictive consideration. At a minimum, he suggests leaders explore what physical AI can do and where it can play a role. They’ll also discover quickly where it isn’t yet a good fit or not mature enough. For example, even the best physical AI still struggles to mimic the remarkable dexterity of the human hand.

Brier also notes that physical AI is not yet plug and play. There will be a notable time lag between receiving the machines and becoming productive. For example, a new robot delivered at a warehouse won’t jump into action once powered on. He says it needs to be configured and trained so that it understands the environment and can be effective in delivering tasks. Brier adds that introducing more robots to the mix increases complexity too and should be factored into any planning.

The age of physical AI may arrive unevenly, but leaders waiting for certainty could discover they waited too long.