Managing the Rise of Agentic AI in Manufacturing
During AI Frontiers 2025, participants raised important questions about the growing role of agentic AI in manufacturing—covering topics such as regulation, interoperability, and the management of autonomous systems. While we did not have enough time to address all the questions during the live session, Peter Sorowka, CEO of Cybus, offers in this article a practical direction for manufacturers navigating this fast-changing field.
How can regulatory frameworks keep up with the pace of agentic AI adoption in manufacturing, especially in highly regulated sectors like pharma and energy?
In industries such as pharmaceuticals and energy, manufacturers have always been under high regulatory pressure. Agentic AI is now accelerating processes and decisions at a pace that traditional regulatory cycles can hardly keep up with.
The key lies in risk-based frameworks: regulators should prescribe fewer technological details and focus more on principles such as traceability, data integrity, and auditability. Manufacturers, in turn, need data frameworks and tools that enable transparency by design. We have already seen several manufacturing customers not only comply with regulatory requirements through complete data governance and traceable interfaces, but also operationalize them for a positive economic effect.
What metrics beyond ROI should manufacturers track to evaluate the success of agentic AI deployments?
ROI is important, but it falls short for agentic AI and is often too difficult to determine. Manufacturers should consider additional metrics, such as:
Time-to-decision: How much does the time from data input to action and solution decrease?
Compliance: Is regulatory accuracy maintained or even improved?
Employee satisfaction and acceptance: Does AI increase safety and reduce monotonous work?
Resilience: How quickly does the system recover from disruptions or deviations?
These qualitative factors are particularly crucial for a sustainable assessment – and can be measured objectively using an interoperable data infrastructure.
How do you ensure interoperability between AI agents from different vendors or platforms?
Interoperability remains one of the main challenges in industrial AI. Different vendors often rely on proprietary protocols, making it difficult for AI systems to work together effectively.
The way forward is through open, standardized data models and infrastructure capable of harmonizing information from any source. Using text-based configurations helps ensure that AI agents can interpret and act on factory data consistently.
A key step is building a neutral connectivity layer, known as a Unified Namespace (UNS). This foundation allows companies to connect different AI agents easily, maintain data governance, and avoid vendor lock-in.
What role will open-source ecosystems play in shaping the standards for agentic AI in industrial environments?
Open-source ecosystems will play a decisive role in determining which standards prevail. This has long been the reality in the IT world, and we are seeing the same trend in the industrial environment. Standards such as OPC UA and MQTT have gained momentum through open community work.
For agentic AI, this means that manufacturers benefit from relying on platforms that natively support open source standards. This allows them to retain the flexibility to integrate their own agents or connect to new ecosystems.
How can manufacturers prevent “agent sprawl” — having too many autonomous systems acting independently without proper coordination?
“Agent sprawl” – i.e., the uncoordinated proliferation of autonomous systems – is a real risk. The solution lies in three principles:
Centralized governance: All agents must be orchestrated via a common data and control layer.
Clear role model: Each agent should have a clearly defined function instead of doing “a little bit of everything.”
Monitoring & feedback loops: Companies must maintain control at all times by continuously monitoring and controlling agents.
We have seen customers creating an integration layer for all data flows – the basis on which the manufacturers can now control their agents before chaos ensues.
Building Trust in Industrial Autonomy
Agentic AI is not just a technological shift—it’s a change in how manufacturing systems operate and make decisions. Success will depend on clear governance, open standards, and reliable data structures.
Manufacturers that adopt these principles will be able to move confidently toward coordinated, intelligent, and accountable automation—a model where AI strengthens operations rather than complicating them.
Related articles:
