Artificial intelligence may be advancing rapidly, but electric utilities are taking a measured approach to deploying it in mission-critical grid operations.

During a panel discussion at the IEEE PES T&D Conference & Exposition 2026, utility leaders from Dominion Energy, Duquesne Light Company and Southern Company said the industry's success with AI will depend less on the technology itself than on building trust through transparency, repeatable results and human oversight.

Moderated by Vikhyat Chaudhry, co-founder and CTO of Buzz Solutions, the session focused on explainable AI (XAI), human-in-the-loop systems and AI-powered infrastructure inspections. The discussion centered on how utilities can safely scale AI across transmission, distribution and substation operations while maintaining reliability and resilience.

"Power utilities are mission-critical," Chaudhry said in opening the discussion. "It is so important to have trust and an explainable AI system."

Trust Starts With Integration

For utilities, panelists agreed, AI must fit naturally into existing workflows before employees will rely on it.

Matthew Rogers, who oversees transmission line operations and inspection programs at Dominion Energy, said utility employees need AI tools that deliver dependable results and integrate seamlessly into existing systems.

Rather than requiring workers to search for information in separate applications, Rogers said AI should be embedded within GIS and asset management platforms. Ease of use, reliability and accessibility are just as important as the underlying algorithms.

"It's got to be integrated with our GIS system. It's got to be part of the asset management program for it to be successful," Rogers said.

Repeatability Builds Confidence

For Erin Henley from Duquesne Light, trust comes from consistency.

Henley, who focuses on work enablement technologies and drone program expansion, said repeatable performance often matters more than achieving perfect accuracy because it gives field personnel confidence in AI-generated recommendations.

"I think for us it really boils down to repeatability over accuracy because it drives confidence down to the end user," she said.

The panel emphasized that explainable AI is especially important for inspection workflows, where utilities must understand why an AI system identified a potential defect before dispatching crews or making operational decisions.

AI Must Augment, Not Replace, Workers

Mike Kerrigan of Southern Company said utilities are more likely to embrace AI when it clearly enhances employee capabilities rather than replacing them.

Kerrigan, whose team evaluates emerging technologies and pilots enterprise-scale innovations, said successful AI deployments create greater value by combining human expertise with machine intelligence.

"When you really start to see adoption take hold," he said, is when AI demonstrates it can help employees achieve better outcomes together than either could independently.

The panel repeatedly highlighted the importance of keeping humans involved in validating AI outputs, particularly for transmission and distribution inspections where mistakes can have significant operational and safety consequences.

Growing Grid Demands Increase Urgency

Utilities also face mounting pressure to deploy new technologies as electricity demand accelerates.

Kerrigan pointed to the rapid pace of grid expansion underway at Southern Company, noting that in Georgia alone the company expects to add roughly as much new electric demand over the next several years as it took approximately a century to build.

That growth, combined with aging infrastructure and increasingly severe weather, is driving interest in AI-enabled aerial inspections, predictive analytics and other technologies designed to improve asset management while helping utilities scale operations.

Still, panelists agreed that widespread AI adoption will depend on proving that the technology is transparent, reliable and integrated into existing utility processes, not simply because it represents the latest technological advancement.