71% of Utilities Are Short-Staffed. That Makes Running Two EAM Systems Twice as Dangerous.

Black & Veatch's 2026 State of the Water Industry survey polled more than 600 water sector stakeholders and found that 71% identified staffing shortages as a significant barrier to operations — edging out aging infrastructure (66%) as the single most cited concern. That's a remarkable inversion. For years, pipes and pumps were the headline problem. Now the humans who understand them are.

The two challenges are not separate. They interact. And for utilities managing both a distribution network and a portfolio of treatment facilities, the interaction creates a risk that most EAM conversations still fail to name directly: when experienced people leave, they take with them the institutional knowledge of two fundamentally different asset worlds — and the tools those people used were never designed to share that knowledge in the first place.

Two asset types, two knowledge silos

Water infrastructure is genuinely dual-natured. The distribution network — mains, laterals, pressure zones, pump stations — behaves like a graph. Pressure propagates. A degraded segment affects everything connected downstream. Knowing the network means understanding topology: which valves isolate which zones, where the cascading exposure lies when a main fails at a junction, how seasonal demand shifts load across the system. That knowledge lives in GIS tools, hydraulic models, and the heads of the engineers who have been walking those corridors for twenty years.

The treatment side — plants, pumping stations, chemical dosing systems, clarifiers — behaves differently. It's hierarchical: a facility contains systems, systems contain equipment, equipment contains components. Failure modes follow parent-child relationships up and down that hierarchy. An experienced plant manager tracks OEE-grade telemetry, understands which maintenance deferrals are tolerable and which compound risk upstream in the process chain, and carries an intuition for the seasonal failure patterns of specific equipment classes.

Both types of knowledge are deep. Both take years to build. And in most utilities, they live in separate systems — a GIS-adjacent platform for the network, a traditional CMMS or facility-management tool for the plants — precisely because the asset types are so different that no single tool handled both well. The result is two institutional siloes that happen to employ people who sometimes talk to each other.

What attrition actually costs

The EPA estimated in 2020 that roughly one-third of the US drinking water and wastewater workforce would become retirement-eligible within the decade. That window is now open. When a network engineer with 25 years of institutional knowledge retires, the query patterns, the informal heuristics, the unwritten rules about which data to trust — none of that transfers automatically to the system they left behind, because the system was built to record work orders, not to capture reasoning.

The typical mitigation is knowledge transfer: documentation projects, shadowing arrangements, structured handoffs. These work, up to a point. But they presuppose that the outgoing expert can articulate what they know, and that the incoming replacement can absorb it fast enough to act on it before the next major incident. Both assumptions are optimistic when the staffing market is tight enough that utilities are hiring people without direct sector experience.

The deeper problem is that institutional knowledge of a water utility is inherently cross-domain. The engineer who has managed the system for twenty years doesn't just know the network or just know the plant. They know how the two interact — how a change in distribution pressure affects the load on treatment, how a planned maintenance outage on a pump station ripples into network pressure management the next morning. That cross-domain knowledge is precisely what disappears when the person who held it in their head moves on.

Why linear asset management and discrete asset management need to share a reasoning layer

Most EAM consolidation arguments focus on data: shared asset IDs, consistent taxonomies, a single source of truth for work order history. Those are real benefits. But for a utility facing a knowledge-transfer crisis, the more important question is whether the system can encode reasoning — not just data.

An AI agent that can reason over network topology can do something a work-order database cannot: it can surface the cascade exposure of a pending pipe repair in the context of current operating conditions across the zone. It can flag that the planned maintenance window for a pump station coincides with a period of elevated demand in the connected pressure zone, and that a similar coincidence three years ago produced a pressure exceedance event. That is institutional knowledge made persistent.

An AI agent that understands asset hierarchy can do the equivalent for the plant side: surfacing parent-equipment failure history when a child component generates an anomaly, recommending inspection scope based on the maintenance state of related assets in the same functional location, identifying which deferral decisions have accumulated enough latent risk to warrant escalation.

The problem is that these two reasoning modes — network-aware and hierarchy-aware — require different data models and different inference approaches. They are not the same problem wearing different clothes. Stitching them together via middleware or a nightly ETL produces something that can query across both systems, but not something that can reason across them. The cross-domain insight that a retiring engineer carried in their head requires a data model that was unified at the foundation, not integrated at the surface.

For a deeper look at the modeling differences, the linear asset management knowledge guide covers GIS-native data structures, dependency graphs, and cascade modeling in detail.

The case for an agentic platform built for both

The right response to knowledge attrition is not better documentation. Documentation captures what people did. AI agents can capture why — the logic, the context, the cross-domain dependencies — in a form that compounds over time rather than depreciating as the workforce turns over.

UniEAM's platform maintains separate native data models for linear and discrete assets: GIS-native graph reasoning for the network, ISO 14224-aligned hierarchy reasoning for the facilities. AI agents operate continuously across both, proposing actions, logging decision context, and surfacing cross-domain dependencies that would otherwise require a human expert to hold simultaneously in working memory. Every prediction carries a decision log: the data inputs, the model state, the inference chain. When the person who used to know why a certain pressure zone behaves strangely on Friday mornings is no longer in the building, that context doesn't disappear — it's in the system.

Over 60% of utilities report that their infrastructure is already operating beyond intended lifespan, according to AWWA's 2025 State of the Water Industry Report. The assets are aging. The workforce is thinning. Running two separate EAM systems under those conditions is not a neutral architectural choice — it is a compounding liability.

The utilities that come through this period intact will be the ones that treated knowledge retention as an infrastructure problem, and solved it with the same seriousness they bring to pipe replacement. That means AI agents embedded in the asset management workflow — not a chatbot bolted on afterward — and a single platform that can reason across both the network and the plant, because the engineers who used to hold both in their heads are counting down to retirement.

Curious how UniEAM handles both? Request a demo.