Designing Workflows that Link the PLC, Unified Namespaces and AI

In my last two articles (“Think Like an Operator: Building Human Intelligence into Industrial AI” and “Empowering Plant Floor Operators with LLMs Using Unified Namespace and On-Prem AI”, I explored how pairing artificial intelligence (AI) with a unified namespace (UNS) can empower industrial workers, and how augmenting AI with real plant data helps bridge the knowledge gap left by retiring experts. Both posts circled back to a central idea: AI is not here to replace operators and engineers, but to extend their reach and preserve the expertise embedded in our plants. 

This time, I want to shift focus down to the layer of automation that often gets overlooked in digital transformation conversations, the programmable logic controller (PLC). It may not feel flashy or innovative in today’s age of AI, but the PLC remains the steadfast workhorse of manufacturing. It’s the foundation of the control system, quietly anchoring every digitalization initiative. For decades, it has been the operational source of truth, executing the logic that literally keeps the lights on, production lines moving and safety systems engaged. The structure of that logic, formalized by standards like IEC 61131-3, provides something we often take for granted in the AI world: determinism, clarity and reliability.

Why the PLC still matters in an AI-driven era 

While modern architectures like UNS and protocols such as MCP (model context protocol) add flexibility and context, they don’t replace the certainty of hard-coded PLC logic. Ladder, Function Block and Structured Text may feel old relative to the advancements in AI, but they remain the backbone of safe, predictable manufacturing. If the code says, “close valve,” the valve closes, every time. That determinism reflects the engineer’s operational intent and is what ensures production stays consistent and dependable.

AI complements this by interpreting patterns and offering confidence-based suggestions. It’s invaluable for spotting anomalies and predicting issues, but it can’t replace the unambiguous logic that keeps production running.

Adding context with UNS and MCP

UNS coupled with MCP provides the foundation for practical implementation of generative AI on the plant floor. The UNS creates a shared language for operational data, while MCP offers a standardized way to describe what tools are available to a large language model and how to interact with them. Companies like HighByte have been especially good at demystifying these concepts, breaking down complex architectures into something plant teams can actually work with.

With that foundation in place, imagine workflows where the PLC still governs the control loop, but MCP-enabled agents can ask for context, such as: What does this signal represent? Where else is this tag used? Who should be alerted if something looks unusual? 

This is where AI stops being a black box and starts becoming a collaborative assistant.

Even with context and workflows, one truth remains: humans must remain in the loop. No AI should be left to unilaterally alter a safety interlock or push unvalidated code to a controller. 

Instead, think of human-in-the-loop workflows as guardrails:

• AI observes the data flow through the UNS.
• AI interprets patterns and anomalies using MCP to call the right tools or suggest actions.
• Human operators validate those suggestions before they reach the PLC.

This model preserves the determinism of control logic, while still leveraging AI’s strength in pattern recognition, prediction and knowledge preservation.

The workflow opportunity

Low-code workflow engines (think of drag-and-drop orchestration tools like n8n) provide a practical bridge. They allow teams to design repeatable AI-assisted but human-approved sequences that connect plant data streams, analytics and decision points without breaking the chain of accountability. These workflows don’t replace the role of PLC logic; they complement it. The deterministic control stays in place, while context-aware workflows give humans more insight and flexibility.

As manufacturers explore AI’s role in operations, it’s tempting to get caught up in futuristic visions of fully autonomous plants. The reality is both more grounded and more exciting: Traditional industrial automation will continue to anchor operations with deterministic control, UNS and MCP will provide the connective tissue for context, and humans will remain the critical decision-makers who approve, override and refine AI recommendations.

The future of industrial AI isn’t about handing the keys over to AI to run your plant. It’s about designing workflows built on the operational realities we’ve trusted for decades, while giving humans smarter tools to make better decisions, faster.

Dan Malyszko is vice president at Malisko Engineering, a certified member of the Control System Integrators Association (CSIA). See Malisko Engineering’s profile on the CSIA Industrial Automation Exchange.

• Why Batch Manufacturers Need Real-Time ERP and MES Integration 
• IT/OT Network Assessment: An Essential Guide to Securing Manufacturing Operations and Qualifying for Cyber Insurance
• The Strategic Shift Toward Vendor-Agnostic Control Systems