Physical AI has moved from a talking point to a real-world application, forcing enterprises to adapt or risk becoming obsolete.
Armel Roméo Kouassi, Senior Vice President at Northern Trust Corporation, says the key for legacy institutions is augmenting human performance to answer shareholder demands for productivity gains.
Kouassi stresses that despite the move toward autonomy, accountability and human-in-the-loop processes remain critical.
Physical AI is no longer just a talking point for the global elite at Davos. Multi-agent, embodied systems capable of autonomous action are now a real-world application, prompting leaders like NVIDIA's CEO to put price tag of trillions of dollars on the opportunity. And while many companies are still figuring out how to rewrite the rules of enterprise innovation with generative and agentic AI, fully autonomous physical AI pilot programs are rolling out around the world.
Watching the evolution closely is Armel Roméo Kouassi, a Senior Vice President at Northern Trust Corporation. As a results-driven global finance executive, he brings more than 20 years of insight on risk management and accountability to the banking, wealth management, and strategic planning fields. From his vantage point, autonomous systems are an imminent force reshaping how institutions operate and compete.
"Physical AI is the next industrial revolution. It's all about autonomous systems that don’t just generate insights, but take action in the real world." Such systems, he emphasizes, aren't a distant concept. They're already being embedded in real-world systems and operating in live environments.
Banking without bankers: As an example, Kouassi highlights a fully autonomous bank launched in India that targets the country's vast unbanked population with a near-zero fee structure. "There aren't bankers. They just hire engineers to make sure the tool works," Kouassi explains. "In that area, many people don't have formal ZIP codes like in the U.S., so your address is identified through GPS."
Leapfrogging legacy: The model sidesteps legacy infrastructure entirely, showcasing AI as a transformative opportunity rather than a threat, especially in emerging markets. "For the unbanked population, it's a revolutionary tool because the alternative is nothing. It's an entirely different value proposition compared to a sophisticated client who already has access to a full suite of banking products."
The AI mandate: This kind of disruption is putting pressure on traditional financial institutions, with the banking industry outlook showing a sector focused on efficiency gains and seizing the agentic AI advantage. "Financial institutions are driven by their earnings calls," Kouassi notes. "When analysts consistently ask about productivity and AI, banks must respond to return value to shareholders. Those questions drive the trend."
The answer, Kouassi says, isn’t a wholesale rip and replace of old systems. Instead, the key lies in augmenting human performance. "We often talk about the physical transformation legacy systems, but part of AI is removing or augmenting human thinking, execution, and bias." Even with a legacy system, he says, AI frameworks can help boost the performance of its human operators. "Now that legacy system becomes one, two, or three times more effective. Right there, you increase the productivity of a firm."
Policing for machines: The move toward AI autonomy is prompting a conversation about new forms of supervision. As the industry moves toward multi-agent architectures, governance comes to the forefront. "We have to think about the new police, the new control, the new requirements to supervise autonomous robots," Kouassi explains. This is where human-in-the-loop processes become a critical aspect of control.
The human envelope: To illustrate, he uses the example of an AI-prompted report. "The report can be generated autonomously. It can be triggered at a particular time. But who has the ownership?" Ownership, he stresses, cannot be assigned to a machine. "A person has to have the ownership, so I believe the human envelope will continue."
In Kouassi's view, the AI race is escalating from a corporate contest to a national one. He describes a growing trend where data is treated as the strategic fuel for national AI ambitions. This push for sovereign architecture is reshaping the geopolitical map.
Data as a fortress: "We are seeing a trend of geopolitical decoupling because data has become the new gold. It's the fuel that gives a country its competitive advantage in AI," Kouassi says, pointing out that nations like France have begun creating their own closed AI platforms and digital ecosystems. "They understand that if the US government forces a company like Microsoft, Google, or Facebook to provide data, they have to comply."
Whether the motivations are geopolitical or industrial, Kouassi cautions that this isn’t a wait-and-see moment. He advises enterprises to adapt to autonomous, multi-agent systems now or risk being left behind in a productivity arms race. "You adapt or you die. The larger risk is doing nothing and allowing your company to lag behind the massive productivity gains that are clearly here."
