Autonomy, AI, and the future of operations

PRESENTING HIS keynote at the recent Automation Fair event, Rockwell Automation CTO Cyril Perducat described an industry at the threshold of a new era, where AI, robotics, and software-defined architectures are reshaping how factories operate and how decisions are made. Yet amid the unprecedented technological developments, he highlighted that manufacturers still face the same pressures: delivering “the right product at the right time with the right quality,” while meeting expectations from employees, stakeholders, and increasingly complex global environments. 

Disruption, he suggested, is not just a threat, it is also a moment to re-architect industrial systems. “In every crisis moment or disruptive moment lies opportunities to transform things, to reinvent yourself.” This reinvention, however, must be thoughtful, manufacturers have to make architectural, technological, and partnership choices today that will shape their capabilities for years to come. And they must do so with the awareness that technology – in particular AI – may look very different even five years from now. “Some of the things that we use today will be obsolete,” he noted. “And some new, better better will emerge.”

Perducat envisions a shift that goes beyond simply layering AI onto existing systems. The traditional model of manufacturing, with factories programmed to execute predetermined tasks, is being replaced by operations capable of learning, adapting, and making autonomous decisions in real time.

“We want to really embrace this transformation... changing factories from factories that you program to factories that make independent decisions, that learn and adapt,” he said. This does not only alter control architectures; it changes the physical and human workflows inside facilities.

He cited Rockwell’s own Singapore factory, which has been redesigned to accommodate human-robot collaboration at scale. Autonomous mobile robots (AMRs), once segregated from people, now operate side-by-side with them, leading to new thinking around layout, safety, and workforce roles. 

In this context, the last decade has brought unprecedented data access, but Perducat cautioned that availability does not equal value. “This is also noise, so transforming those data into a relevant context decision, and ultimately closing the loop, this is what really matters – transforming an insight into an action.”

This end-to-end loop, he argued, defines true autonomy. It also underscores why AI must be embedded within the architectural fabric of operations, rather than treated as a bolt-on capability or standalone product. 

“This is a reinvention," he said. "This is not just patching AI to automation, or considering AI as another category of product. It's really, for us at Rockwell Automation, the idea of reinventing what we have been doing over decades around the development of automation into the business of autonomy – we believe that all business is becoming the autonomy business."

Three technologies powering the shift

Perducat outlined three mutually reinforcing technology domains enabling this new era of autonomous operations, advancements that are “available today… not technology that needs to be invented in a decade".

Software-defined automation (SDA)

SDA decouples application logic from hardware decisions, giving engineers freedom to design systems that are flexible, scalable, and easier to adapt. “It’s about maximising the flexibility and agility of the system,” he said. Rockwell’s own FactoryTalk Design Studio illustrates this model, enabling application design independent of PLC or hardware selection.

Importantly, SDA does not diminish the role of hardware. Perducat emphasised that Rockwell continues to engineer advanced devices, edge platforms, and intelligent I/O. But with SDA, hardware becomes an adaptable execution layer rather than a limiting constraint.

SDA also opens the door to new forms of self-organisation and self-optimisation. With architectures that can adapt dynamically, systems can be designed around human roles and skill levels, something Perducat said Rockwell is prioritising. “We can design them in a way that they fit the roles of the different people... with an experience that reduces friction and complexity.”

Artificial intelligence

In Perducat’s words, AI is “one more technology” – not a product category, but a capability that should permeate every layer of industrial software and hardware. Rockwell’s strategy is to fuse AI into functional domains such as vision, edge condition monitoring, and control.

He described how LogixAI embeds autonomous control parameter adjustment directly into PLC architectures, while Vision AI provides a unified software layer for diverse inspection and detection tasks. This dual strategy – retrofitting existing products while creating entirely new AI-native systems – helps allow customers to evolve at their own pace.

Robotics and the physical embodiment of AI

Perducat described robotics as AI made physical; machines capable of autonomy, adaptability, and continuous learning. In this sense, he suggested, many industrial machines will evolve into “some form of robots", even if not humanoid.

With Rockwell’s acquisition of Clearpath and OTTO Motors, the company now sits at the intersection of automation, autonomy, and robotics fleet orchestration. AMRs not only transport materials but also act as mobile sensors, enabling new applications in safety monitoring, environmental scanning, and real-time risk detection. “They become mobile sensors,” he said, highlighting the importance of unified fleet management platforms.

The human-machine interface of the future

One of the most transformative outcomes of this technological convergence will be how people interact with industrial systems. Perducat described a shift toward natural interaction models: “Instead of searching for data to make a decision, you can get directly a recommendation.” Whether through chat, voice interfaces, or role-specific digital assistants, the goal is not to replace workers but to empower them, reducing cognitive load and elevating decision quality.

This leads directly into Rockwell’s next frontier: agentic AI. “Instead of providing information, it drives an action,” he explained. Agent architectures will appear throughout Rockwell’s software ecosystem, including early demonstrations within Plex. But they will also manifest physically, through robots and autonomous systems capable not only of analysing but acting.

The journey ahead

Perducat closed with both a challenge and an invitation. The future of industrial operations will be shaped by decisions companies make now—choices about architecture, partners, and strategy. “The decisions you make today will set a lot of your path in the future,” he said. Rockwell intends to be a long-term partner on that journey, providing the systems, expertise, and innovation required to build factories that learn, adapt, and continuously improve.

“We are at the edge of intelligence and physical AI,” Perducat concluded. “This is just the beginning.”

http://www.rockwellautomation.com