At Rockwell Automation Fair 2025 in Chicago, Cyril Perducat set out a vision for industrial operations built on adaptive intelligence rather than fixed instruction. The convergence of software defined automation, embedded AI and robotics is becoming the foundation of modern production, reshaping architectures, workflows and expectations across global industry.
The setting helped underscore the moment. Automation Fair has become a yearly barometer for the pace of change across industrial technology, but this year’s event carried a different weight. Cyril Perducat, Chief Technology Officer, Rockwell Automation, stepped on stage to address a room filled with engineers, strategists and executives wrestling with new pressures. The speed of transformation, the uncertainty of emerging AI techniques and the complexities of workforce expectations formed the backdrop to his message.
Perducat grounded his perspective in the straightforward reality that leaders cannot escape. The world is moving quickly, yet the demands placed on operations remain uncompromising. Plants still need to deliver the right product, at the right time and with the right quality. Teams still need tools that help them make better decisions under pressure. Shareholders and stakeholders still expect reliability on top of innovation. His emphasis remained fixed on navigating this landscape without being distracted by the volume of technological noise that now characterises the industrial sector.
He described the difficulty of separating hype from genuine value, and he urged leaders to resist imitation-based decision making. His view was that organisations must set their own direction rather than simply copying what they saw at events or in competitor case studies. This was connected to a broader warning that technology will shift again within the next five years, particularly as new approaches to intelligence emerge. The architectures chosen today must therefore support evolution instead of locking plants into designs that lose relevance just as new capabilities reach maturity.
Perducat returned repeatedly to the idea of factories that learn and adapt. He positioned this as the defining transformation of the next industrial era. “Technology evolves, but the fundamental challenge does not,” he said. “You still must deliver, and you still have to navigate increasing complexity. The question is how you design an operation that can learn, adapt and improve with every cycle. That is where the future of industrial operations truly begins.”
Building the technical foundations of adaptive operations
Software defined automation sat at the heart of Perducat’s argument. He treated it not as an abstract concept but as a tangible architecture already operating across Rockwell’s ecosystem. The principle removes the traditional binding between application logic and hardware selection, allowing engineers to develop functions independently from the platforms that execute them. This shift creates a different kind of industrial environment.
“Those three advancements are not separate pillars,” he said. “They work together. You need a software defined architecture to give AI the right environment, and you need AI to unlock the potential of robotics. The result is a system that can evolve without friction and without forcing you to rebuild everything each time technology shifts.”
He expanded on this by referencing Factory Talk Design Studio, which allows teams to design applications before committing to specific execution hardware. It is a change that mirrors the evolution seen in the IT world through software defined networking and software defined vehicles. In each case, abstraction created freedom, and freedom accelerated innovation. Rockwell is applying the same principle to industrial control by enabling system behaviour to evolve across a heterogeneous mix of PLCs, edge devices, motion controllers, safety systems and network components.
This is not an argument for diminishing the role of hardware. Perducat was explicit that hardware remains essential. Specialised form factors, deterministic processing, environmental resilience and domain-specific silicon still matter, particularly in high-speed, high-precision environments. The point was that software defined automation allows hardware to do what it does best while giving software the agility needed to manage complexity. It creates a shared stack in which each layer performs the function best suited to its characteristics.
He pointed to the evolution of multidisciplinary control architectures to illustrate how the foundations have already shifted. Control, safety, motion and cybersecurity were once stitched together through separate systems. Rockwell integrated them into unified platforms, reducing friction between domains. Software defined automation extends this idea further by allowing entirely new disciplines to be added without rearchitecting the underlying system.
“With SDA we are no longer limited by the structure of the hardware,” he said. “We can add new disciplines, embed AI directly into the silicon and create architectures that scale across locations and use cases. This is the next evolution of automation, not a departure from it.”
This expansion of the control layer is not theoretical. By embedding AI models within the execution environment, the PLC becomes capable of continuous adjustment based on patterns, deviations and optimisation targets. The classic PLC cycle no longer stops at rule execution. It incorporates learning, evaluation and parameter modification, and these capabilities operate within the deterministic boundaries required for industrial reliability.
AI as a distributed capability within the control fabric
Perducat positioned artificial intelligence as a capability distributed throughout the operational stack. AI is not an isolated layer sitting above control. It is a set of functions integrated into sensors, controllers, edge platforms and execution systems. He was direct about the danger of treating AI as a separate product category.
“Industrial AI has to be real,” he said. “It must fit into existing architectures and deliver measurable value. When you take a proven system and add an intelligence layer that closes the loop between insight and action, you create something that transforms operations without requiring technologies that do not exist yet.”
He detailed how Rockwell has created AI functions aligned with specific industrial needs. FactoryTalk Analytics VisonAI replaces rule-based pattern matching with adaptive interpretation, allowing systems to detect subtle deviations that traditional approaches overlook. AI modernises condition monitoring by analysing vibration, temperature and noise signatures with models that improve as more data becomes available. FactoryTalk Analytics LogixAI embeds closed-loop optimisation within controllers, enabling parameters to adjust continuously in response to shifts in process conditions.
These functions share an architectural quality. They operate close to the physical process, which reduces latency, improves determinism and protects continuity during network interruptions. AI therefore becomes part of the control architecture rather than an adjunct. It provides continuous insight without overwhelming operators with information. It also takes on the task of identifying optimisation pathways by evaluating relationships that are not visible through static thresholds or predefined rules.
Perducat stressed the importance of natural interaction as these systems mature. “The interaction model on the shop floor is going to look very different,” he said. “Natural interaction, whether it is chat or voice, will help teams work with systems that provide recommendations instead of forcing them to search through data. That is how we make AI enhance work rather than complicate it.”
This level of interaction shifts the role of operators from oversight to decision validation. Systems that provide recommended actions, predicted outcomes and contextual explanations reduce cognitive burden and accelerate response. It also supports workforce transition by helping less experienced employees perform at levels that once required years of plant familiarity.
Robotic systems as physical manifestations of intelligence
The robotics landscape formed the third component in Perducat’s technical stack. He described robotics as the physical embodiment of AI, a statement that reframes robots not as mechanical tools but as intelligent operational agents. “A machine that performs a task and adapts as it works is a robot,” he said. “Many machines will evolve into robotic systems, and they will no longer be isolated. They will share space with people, and that has deep implications for safety, management and the way information flows through a plant.”
This redefinition expands the domain of robotics beyond articulated arms and fixed installations. Conveying systems, packaging machines, inspection units and mobile platforms can all evolve into robotic forms as AI augments their ability to interpret, adjust and navigate. It also paves the way for mixed environments where autonomous mobile robots operate alongside humans, moving across dynamic spaces and coordinating with fixed assets.
Perducat highlighted the role of mobile robots as roaming sensors. Their ability to traverse a plant gives them a unique position in operational intelligence. They can monitor safety compliance, collect equipment condition data, verify environmental parameters and identify spatial anomalies. These insights become more valuable when combined with fleet management platforms capable of coordinating hundreds of devices, balancing tasks and optimising routes.
This integration is strengthened by Rockwell’s autonomy platform, developed through the acquisition of Clearpath and OTTO Motors. It allows mobile and fixed robots to be managed within a unified system, which supports coordinated behaviour, shared decision making and cross-platform learning. Fleet management becomes more than path optimisation. It becomes the operational nervous system of the factory, linking intelligence in motion with intelligence embedded in equipment.
A technical journey that demands strategic clarity
Perducat closed his session at Automation Fair 2025 with a return to the importance of decisions made today. The next decade will be shaped by leaders who design architectures that can learn, adapt and scale across global operations. Autonomy is not a futuristic ambition but an achievable path defined by clear technical principles and grounded application of AI and robotics.
“There is noise everywhere,” he said. “There is hype, there are countless possibilities, but what matters is the ability to bring all of this into reality. The foundation is here. The systems are real. What you decide to do with them will determine how well you navigate the next phase of industrial transformation.”
The message was both practical and directional. Industrial autonomy will not come from individual technologies. It will come from architectures that connect them. It will come from systems designed to evolve. It will come from factories that learn.
