Rockwell's AI Ambitions Test the Limits of Factory Automation

MILWAUKEE -- Rockwell Automation has unveiled industrial controllers and AI software meant to help make factories operate autonomously. But making AI play a more prominent part in production processes raises the question of how the machines would be programmed to act when automated decisions go wrong.

The American company presented its updated technology portfolio Oct. 27, prior to Automation Fair 2026, which features, among others, LogixEdge, an industrial controller said to unite traditional industrial control and information technology and AI workload capabilities for autonomous control.

It comes after FactoryTalk Conductor was presented Oct. 21 as a software designed to synchronize activities of AI agents in industrial engineering applications. The platform is set to be launched commercially in the first quarter of 2027.

LogixEdge and FactoryTalk Conductor mark two stages of the automation process. The latter begins with engineering tasks such as design, simulation, and testing. LogixEdge puts the AI processing capability closer to those systems that control machine operation.

AI assistance in designing a production process differs from the authorization of software to change machine behavior during operation.

Where AI Control Meets Safety Limits

Programmable controllers are currently employed to carry out instructions, monitor sensors, and coordinate machinery in factories. The use of AI would increase the flexibility of factory operations through the analysis of conditions and suggesting changes that go beyond set sequences.

However, the problem lies in the fact that such changes should stay within the safe operational limits.

Suggestions for combining AI with Operational Technology provided by Australian and international cybersecurity organizations list safety risks, reliability risks, and cybersecurity risks. They stressed that large language models should not be able to make any critical safety decisions.

Thus, there should be some sort of AI-independent protection. For instance, if the system suggests increasing the speed of operation, it should not result in dangerous machine actions.

Furthermore, the testing is complicated bedause the behavior of AI is conditioned by the current working conditions and input information rather than set instructions. Thus, manufacturers should know how their systems behave under conditions of incorrect sensor readings and unpredictable output.

The AI Risk Management Framework recommended by the U.S. National Institute of Standards and Technology provides further suggestions regarding monitoring and evaluation for human oversight.

Who Approves the Machine's Next Move?
In addition, connecting AI software to industrial controllers broadens cybersecurity issues.

ISA/IEC 62443 standards cover cybersecurity requirements for industrial automation systems with respect to access control, network security and system integrity.

In case of AI-based processes, manufacturers need to decide on authorized persons, record their decisions and put machines in a stable operating mode after a failure.

According to Rockwell, FactoryTalk Conductor will enable AI agents to interact within engineering software. The first integrations provided by Rockwell include Studio 5000 Logix Designer, Logix Echo, Emulate3D and FactoryTalk Optix. But it does not mean that AI agents can operate production machines independently.
At the same time, Rockwell's press releases do not clarify to what extent the new AI-based controllers will make decisions in everyday factory activities. Also, the manufacturer provided no comparative results concerning the performance of its autonomous control capabilities.