The 2025 AIChE Global Congress on Process Safety was filled with forward-looking ideas, but one particularly resonant insight came from Dr. Rainer Hoff in his presentation “AI-PSM: Progressing but Not There Yet.” His message was clear: while artificial intelligence in process safety is advancing, organizations are still refining how best to embed it into practical workflows. One compelling use case is the application of AI agents within modern management of change solutions and applications to assist in distinguishing between a Replacement in Kind (RIK) and a Management of Change (MOC)—a decision point that is deceptively simple but operationally critical.
The RIK vs. MOC Dilemma
In PSM-covered facilities, the question of whether a proposed change constitutes a true replacement in kind or requires a full MOC process is pivotal. A misclassification can lead to bypassed hazards, unvalidated assumptions, and a breakdown in safety defenses. Traditionally, this judgment is left to experienced engineers, maintenance supervisors, or project teams—each with varying interpretations and levels of documentation.
But what if an intelligent assistant could act as a neutral, informed advisor, nudging users to consider critical details they might otherwise overlook?
The Solution: AI Agents as Active Facilitators
Imagine an AI agent embedded into your MOC portal or change request system. Rather than just processing a checkbox form, the agent engages with the user in a dynamic conversation, tailored to the specifics of the proposed change. For example:
User: “We’re replacing a flow transmitter on the hydrogen line.”
AI Agent: “Is the new transmitter from the same manufacturer and with identical specifications?”
User: “Same specs, different manufacturer.”
AI Agent: “Does the new device have the same safety certification (e.g., SIL rating) and communication protocol?”
User: “It’s rated SIL2, like the original, but uses a different HART protocol version.”
AI Agent: “Based on that information, this may not be a Replacement in Kind. Let’s initiate a formal MOC review to assess integration and control system impacts.”
This kind of exchange demonstrates how AI agents can guide the user through context-specific questions, ensuring no critical detail is missed before classifying the change. It turns a passive form into an active dialogue, helping frontline staff make safer, more consistent decisions—especially in facilities with complex control systems or multiple layers of safety instrumentation.
Contextual Intelligence and Data-Driven Support
Beyond these conversations, an AI agent can cross-reference the proposed change with internal databases, historical MOCs, design specifications, and known risk scenarios. For instance:
- “Similar transmitter changes in this unit previously impacted loop stability.”
- “This equipment has a history of integration issues with DCS Version 8.5—initiate compatibility review.”
These insights are surfaced at the point of decision-making, supporting both experienced personnel and new hires alike.
Transparent Justification and Audit Trail
After guiding the user to an RIK or MOC decision, the AI agent can generate a traceable justification:
“Change evaluated: Flow transmitter replacement. Specifications match; manufacturer and protocol differ. Classification: MOC initiated due to potential integration and safety certification variance.”
This enhances documentation quality and strengthens compliance during internal audits and regulatory inspections.
Still Progressing—But Closer to “There”
Dr. Hoff reminded us that while the integration of AI into PSM is promising, challenges remain—particularly with trust, data completeness, and change management. But the use of AI agents in decision-support roles like RIK vs. MOC classification stands out as a low-risk, high-impact opportunity.
With the right training data, rulesets, and human oversight, these agents can improve consistency, reduce overlooked risks, and free up expert bandwidth for higher-value work. They don’t replace the engineer—they amplify their attention to detail and ensure alignment with policy.
Final Thoughts
The next generation of PSM tools won’t just record decisions—they’ll help make them better. AI agents that facilitate thoughtful, data-driven engagement are one of the clearest ways to achieve this today. As we build on the insights from industry leaders like Dr. Hoff, we move one step closer to truly intelligent process safety systems.
