Utilities don’t need more AI use cases—they need a better starting point

4-minute read

Over the past two years, utilities have launched a wave of AI initiatives. In my conversations with utility executives, a consistent pattern emerges:

AI activity is increasing, but day-to-day operations look largely the same.

The use cases are not the problem. Pilots are running, and new capabilities are being added to existing systems.

But teams still pull data from multiple systems. They validate and revalidate before acting. Decisions move at the same pace, with more information layered into the process.

Many utilities are still deciding where to focus their AI investments. Nearly a third of utility executives say they don’t know how to prioritize investments or where to focus first.

When efforts do begin, momentum rarely carries beyond early experimentation. Organizations move past initial pilots without seeing meaningful operational impact, leaving them stuck in “pilot purgatory.”

Why starting with use cases keeps utilities stuck

Most AI efforts in utilities begin with use cases.

The logic is familiar: identify opportunities, prioritize them, and launch pilots tied to specific capabilities.

That approach produces activity. It almost never changes execution.

Use cases define what AI can generate—forecasts, classifications, recommendations. They do not define how that output is used within the flow of work.

As a result, new capabilities are introduced without being anchored in how work actually runs.

This is where progress starts to stall.

Article continues below.

Where execution falls short

In many cases, the models are doing exactly what they were designed to do. Data becomes easier to access. Forecasts improve. Reports are produced faster.

But the workflow around those outputs doesn’t change.

Teams still validate information, reconcile differences across systems, and move decisions through a defined sequence of steps. The work may look different, but it moves through the same process.

That’s where the disconnect becomes visible.

Utilities measure progress in reliability, cost, risk, and speed to decision. Improvements in data or analysis only matter when they change those outcomes. When the workflow stays intact, the impact stays limited.

The result is a familiar pattern: more insight, but no meaningful change in how work moves.

A better starting point: existing workflows

The utilities seeing measurable progress do something different: they begin with work already in motion, using tools and data they already have in place.

They analyze how work is currently done—where it slows down, where effort concentrates, and where decisions depend on pulling together information from multiple systems.

That lens points to core operational processes such as

• Work order management
• Crew scheduling and dispatch
• Inspection and maintenance cycles
• Regulatory reporting

These workflows sit at the center of day-to-day operations. When they change, the impact shows up immediately.

What this looks like inside real workflows

The difference becomes clear when you look at how core operational processes actually run.

Vegetation management: narrowing the field of focus

Vegetation programs already operate at scale. Crews collect data, analysts review it, and planners decide where to send resources.

The constraint is volume, not insight. Teams work through large datasets to identify a relatively small number of high-risk areas.

AI changes the starting point.

Instead of working through the full dataset, teams begin with a prioritized subset generated from risk signals across assets, conditions, and environment. Analysts still review and confirm, but they start from a narrower and more relevant set of inputs.

That change carries through the workflow. Crews spend more time on high-risk areas, lower-priority work falls out of the critical path, and planning becomes more targeted.

Regulatory reporting: reducing assembly work

Regulatory reporting follows a defined process, but much of the effort is spent assembling and validating data from multiple systems.

Each cycle repeats the same sequence: gather inputs, reconcile inconsistencies, and prepare outputs to meet reporting requirements.

Here again, AI changes the starting point.

Instead of assembling data from scratch, teams start with information that is already structured against reporting requirements, with discrepancies surfaced earlier in the cycle. Review and validation still happen, but less time is spent assembling information.

The impact is cumulative. Reporting cycles become more consistent, late-stage issues are reduced, and teams have greater confidence in audit readiness.

How value builds over time

When AI is applied within a workflow, the impact doesn’t arrive all at once.

It starts small. Individual tasks take less time. Information is easier to work with. Outputs become more consistent.

From there, the workflow itself starts to shift. Steps drop out. Handoffs become less frequent. Decisions move with fewer interruptions.

As those changes compound, the effect extends beyond a single workflow. What began as a localized improvement starts to influence how adjacent processes operate.

Organizations that try to scale too early run ahead of the underlying change. The workflows haven’t shifted enough to support broader adoption, so the impact doesn’t carry.

Where to go from here

The next step isn’t launching more AI initiatives. It’s changing how progress is measured.

Progress shows up in execution. Decisions happen with less delay. Work moves with less effort and fewer interruptions.

That’s the signal to build on.

Start with workflows already in motion. Look for evidence that execution has changed. Where that change is consistent, expand it. Where it is not, adding more AI won’t help.

Many utilities lose time by scaling activity instead of strengthening impact.

A narrower focus, applied with discipline, produces better results.

AI becomes meaningful when it’s embedded in how work runs every day. That impact comes from picking a small number of workflows and carrying changes to completion.

You might also be interested in …