Hi all,
I want to open a conversation that I think we've been dancing around in our chapters and LinkedIn feeds, but rarely discuss with the rigor this
community is known for.
Not "will LLMs change EAM work?" - they already are.
The honest question is where they land, what they break, and what we should refuse to automate.
I'd like to hear from planners, schedulers, reliability engineers, CMMS administrators, and asset managers in this group. Below are the
some already is going, possible or for real,
Where LLMs are already useful (today, in some solution):
- Work order quality. Free-text descriptions like "pump making noise again" become structured, classified, with failure code suggestions, asset references, and probable cause prompts, before the planner touches it. This alone fixes a decade of garbage-in / garbage-out in CMMS reporting.
- Job plan drafting. Generating first-pass task lists, labor estimates, and material requirements from an equipment type + failure mode, leveraging OEM manuals, internal history, and similar past work orders. Planner still owns the final plan; LLM removes the blank page.
- Backlog triage assistant. Conversational queries against the backlog: "Show me overdue PMs on critical assets with no spare parts in stock." No more waiting on a BI team for ad-hoc questions. Failure analysis support. Summarizing months of work order history per asset, surfacing recurring failure patterns a human would need hours to find.
- Knowledge capture. Senior technicians retiring - LLMs as a way to interview, structure, and preserve tribal knowledge into searchable, governed content tied to assets and failure modes.
- Training and onboarding. New CMMS users asking a copilot "how do I close a work order with partial parts return?" instead of waiting on a SuperUser.
Where I think we should be careful:
- PM optimization decisions. An LLM can summarize and suggest. It cannot own the call on whether to extend a PM interval on a critical asset. That's a reliability engineering decision, with consequences the model doesn't carry.
- Root cause analysis. LLMs are excellent at pattern association, not at causation. RCA without human judgment becomes a confident-sounding hallucination factory.
- Criticality and risk ranking. Subjective, business-context-dependent, and political. Don't let a model that has never walked the plant decide what's critical.
- Compliance and audit trail. Regulated industries, utilities, oil & gas, pharma, aviation, need traceability LLMs don't yet provide natively. "The model suggested it" is not an audit defense.
The CMMS/EAM vendors are racing to bolt LLMs onto their products. Most of what's being shipped is a chat window over the database, useful, but not transformational. The deeper opportunity is in the workflows around the solution, not inside it: how we plan, how we capture knowledge, how we make decisions.
I'd would like hear concrete experiences from this group.
------------------------------
Janse Romero Borçari
Asset Management Consultant
Maximo Matters
------------------------------