It’s natural to be drawn to shiny objects— we can’t help it. Something new, even if untested, presents a chance to cure our most baffling ailments. That’s especially true in the energy industry, which faces gargantuan challenges, like climate change, with many holding a sense of techno-optimism in their search for answers.
Artificial intelligence is the latest in a long line of revolutionary technology advancements that have captivated electric utilities. But this time there’s consensus that, at least one day, AI will transform power grid operations and customer engagement, emerging from the hype cycle that doomed its “shiny” predecessors.
AES, one of the largest vertically integrated energy companies in the world, shares the belief that AI will lead to a step change for electric utilities. The company developed an AI robot designed to build solar farms, deployed next-generation grid sensing technology to expand renewable energy capacity, and backrolled startups pioneering the latest AI grid applications. But while other utilities say they’re ready to break out of the “sandbox” of AI pilots, AES is taking a strategic approach before making the leap.
While AES’ global division is exploring AI tools for both its generation and utility portfolios, local operations like AES Indiana and AES Ohio are taking a phased approach. The Midwest utilities prioritized a foundation for data organization, quality, and analysis before advancing further into AI, the company’s data team said.
“We needed to walk before we ran,” said Scott White, a senior data scientist at AES, told POWERGRID in a recent interview.
Utilities, like National Grid and Entergy, have implemented AI software applications to streamline core responsibilities like vegetation management, with some boasting double-digit reliability gains as a result. AES, however, wanted to find out if similar progress could come from better utilization of legacy utility data.
A few years ago, AES Indiana identified opportunities for reliability improvements in specific neighborhoods, where dense tree cover and older infrastructure created localized hot spots. “People like to live by trees,” White noted, adding that a standard four- to six-year trimming cycle wasn’t enough to address problem areas. The result: costly outage responses to outages that could be avoided with proactive planning.
To identify where and when to trim, AES turned to its existing data sources, integrating information on outages, assets, and vegetation density to create a “global picture” of problem areas. The result was a hybrid approach combining traditional cycle-based trimming with condition-based targeting informed by predictive models. According to
White, the new strategy has led to an estimated 10% to 20% improvement in reliability compared to a pure cycle-based approach for those problem circuits.
While tools like satellite imagery and AI-driven analytics are on the horizon (AES has had some conversations with Neara), AES is proceeding carefully.
“With our data, we know certain things are very good and certain things are very bad,” White said, citing challenges like spatial resolution and cost allocations. For example, pinpointing the exact location of outages within long circuits remains a blind spot. Without improving its GIS and asset management systems, more advanced tools would only add complexity without delivering clear value.
The caution is not unwarranted. As Casey Werth, IBM’s global energy and utilities lead, noted, utilities often struggle with “blackhole” data—information that exists but remains incomplete, outdated, or inaccessible. “Without a strong data foundation, even the most advanced AI tools will fail to deliver actionable insights,” Werth explained. AES’ approach aligns with this warning, ensuring that their data is optimized and actionable before investing in AI-driven tools.
“It takes a concrete (data) foundation,” said AJ Hall, portfolio management director at AES’ innovation arm, AES Next. “So that when we get to higher level decision-making with AI, it’s immediately adding value.”
A cultural shift toward data-driven decisions
AES’ new focus on data stems from a broader organizational shift. The company’s recently established data science and analytics team, led by Norvin Clontz, is tasked with improving decision-making across all aspects of utility operations, from storm response to reliability KPIs.
“We’re generating most of our benefits with traditional predictive models,” Clontz said. Techniques like regressions and random forests are already providing actionable insights. For instance, AES is now able to quantify the impact of its vegetation management programs and prioritize investments that deliver the greatest reliability improvements.
AES has also adopted a “medallion” approach to data warehousing, transforming raw source data (the “bronze” layer) into curated datasets (the “silver” layer) and finally into purpose-built applications or dashboards (the “gold” layer) for decision makers. This ensures that data is accessible and actionable without burdening operational systems.
“There’s a challenge for that last mile, the consumption layer, to actually get into the hands of the decision-makers in a digestible way,” Clontz said.
Looking ahead
While AES remains focused on foundational improvements, the utility is already exploring advanced tools for the future. Hall pointed to digital network models, which integrate satellite, GIS, and LiDAR data to simulate grid conditions and identify risks like encroaching vegetation. “To do that, you need a good foundation,” he emphasized.
AES’ approach reflects a growing recognition in the industry that advanced tools are only as good as the data feeding them. By taking the time to clean and optimize its data, AES is building a platform for sustained reliability and efficiency gains—without rushing into costly, overhyped solutions.
