Why The Next Decade Belongs To The Utility Enterprise That Operates As One

Frank Carnevale, Country Head, Canada for iGreenTree.ai | AI, Digital Innovation & Cleantech | Energy & Utilities.

getty

Every utility leader is buying AI, but very few are getting it to production. That gap is the most expensive thing in the industry right now, and almost no one has it on a budget line.

The problem was never the algorithm. McKinsey’s 2025 State of AI survey found that while nearly 90% of companies use AI, only about 7% have fully scaled it; value comes from rewiring workflows and the operating model, not a better model. Bolt AI onto a fragmented enterprise, and you do not modernize it. You automate the dysfunction, and you scale it.

The utility of the future is defined not by which technologies it buys, but by whether it can act as one system.

​A Coordination Gap, Not A Technology Gap

​Utilities are rich in operational data but often struggle to turn that data into outcomes. Smart meter networks, grid control systems, outage platforms, customer systems and geospatial tools generate terabytes of information every day, yet much of that value remains trapped in dashboards rather than translated into action.

Drop an AI model in, and it might dazzle in the demo, but then it falls over the moment it meets live data that was never integrated. In National Grid Partners’ 2025 survey of utility innovators, the top obstacles included talent and integration with existing systems, not the quality of the models.

I have a name for the drag behind this: enterprise friction, the operational, informational and organizational resistance that builds when work crosses functions never designed to run as one. It's the same asset recorded three ways in three systems. A decision that stalls because no single owner is accountable across the silos. Decades of operational knowledge that disappears when experienced operators retire. None of it appears as a line item, which is exactly why it goes unmeasured and unaddressed.​

Why This Is Urgent Now

​Utilities are experiencing what many enterprises are beginning to feel—only earlier and more intensely, because three pressures are converging at once.

1. Demand is surging. NERC’s 10-year outlook projects 224 gigawatts of North American summer peak growth, a 69% increase in a single year, driven largely by AI data centers.

2. AI is both the source of new load and a potential solution. The IEA expects global data center electricity consumption to roughly double by 2030.

3. Affordability is tightening. NEADA reports that the average U.S. residential electricity bill rose nearly 29% between 2021 and 2025, with roughly one in six households now behind on payments.​

The Operating Model: Converging Five Things

​Based on my research with iGreenTree's Utility Singularity operating model, there are five key domains that utilities should converge that the traditional model keeps apart:

1. Operations: One real-time view across generation, grid, field and customer

2. Data: One governed truth

3. Intelligence: AI inside workflows on trusted data, not side experiments

4. Governance: Risk, compliance and decision rights designed in

5. Workforce: Expertise scaled as a system capability, not trapped in individuals

The sequence is what most teams get wrong. Data and governance come first; intelligence depends on them. Trustworthy AI is not bought; it is earned with clean, governed data and clear rules. EPRI reached the same conclusion in its August 2025 readiness report: Data readiness is the precondition that turns AI potential into value, and most utilities are not there yet. Skip it, and you keep manufacturing pilots.

The 'Aha!' For Anyone Playing With AI

The fix is an operating model, not another pilot. Every form of enterprise friction has a mirror image that becomes value once it is removed: less rework, maintenance driven by an asset’s actual condition rather than a fixed calendar, a single forecast instead of multiple conflicting ones, faster response when failures occur and AI that can finally reach production because the systems beneath it are able to support it.

Where outcomes are directly incentivized, the results are measurable. After Great Britain introduced outcome-based regulation under RIIO, customer interruptions fell by roughly 19%. That said, the magnitude of impact will always vary by organization and context. It is ultimately a measurement exercise, not a fixed benchmark. Anyone promising a guaranteed enterprise-wide savings percentage is selling the very thing convergence is meant to replace.

Keeping The Human In The Loop

It is important to emphasize that the goal is not autonomy; it is decision intelligence with a human in the loop. On a power grid, an unsupervised error is unacceptable; in other domains, it may simply be costly. The real prize is not replacing human expertise but scaling it, especially as experienced operators retire. Nearly half of the utility workforce is expected to retire within the next decade, while the Center for Energy Workforce Development already reports a 76% employer-reported skills gap.

This capability cannot be purchased off the shelf; it must be built. Technology adoption only scales when workforce models, training pathways and institutional knowledge evolve alongside it. That is one reason I have been exploring the concept of a Utility Transformation Guild: an industrywide capability engine designed to strengthen how utilities develop, credential and transfer expertise across the workforce. Without modernizing the capability engine itself, the industry will struggle to fully capture AI’s potential while maintaining reliability and affordability.​​

Not Just A Utility Story

Now run the substitution. Replace “utility” with bank, hospital, insurer or manufacturer, and almost nothing changes. The pressures differ, but the diagnosis does not. The organizations getting durable value from AI are not those with the cleverest models, but those that did the unglamorous work of making the enterprise coherent enough to use them.

Ask your team, candidly: How many AI pilots reached production last year, and what stopped the rest? Do our systems agree on one version of the truth, or do we reconcile by hand? If those answers sting, you have found your enterprise friction. The good news is that friction is fixable, and removing it is cheaper than the next big build.

Bottom Line

The most important question on your AI road map is not which model to buy, but whether your enterprise can act on what it tells you. Treat this as one more IT project and you will lose the decade to those treating it as the main event. The future does not belong to the utility with the most technology. It belongs to the one that operates as one. The same is true of your business.​​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?