Report: Inside the utility mindset on AI deployments

A report out today from Latitude Intelligence and Indigo Advisory Group examines where and how the utility sector is integrating artificial intelligence tools — and what will define the market’s near-term evolution.

• The top line: Utilities are entering a new era of tech deployments, and are in particular experimenting with AI’s potential to solve pressing issues around reliability and resiliency. For now, though, most utilities say they lack a central, holistic vision of AI deployment. For solutions providers, that means navigating a still-shifting market as data availability and accuracy improves and utilities hone how to prioritize workloads and AI applications.
• The market grounding: AI and other advanced technologies aren’t entirely new to the utility sector — many utilities have been working with machine learning for tasks like load forecasting for quite some time, and there has long been a trend toward digitalization. However, the focus in prior decades was on sensor deployment, communications networks, and smart meter installations, all of which serve as the foundation for today’s turn to more automation. As the technology is incorporated more widely into grid systems, utilities’ priority use cases for AI include load forecasting, DER orchestration, and predictive maintenance and grid health optimization.
• The nuts and bolts: The report examines both the supply and demand sides of the utility AI market, relying on data from utilities and vendors alike to assess how the power sector is evaluating and adopting AI. Researchers for Latitude Intelligence — which is the research arm of Latitude Media — and Indigo Advisory Group interviewed 15 utilities and more than 20 vendors, and tracked data across financial statements, IRPs, FERC filings, and announcements.

The report examines the nuances of strategic AI deployment and evolving market conditions, and highlights the approaches that have vaulted certain utilities to the forefront of AI innovation.

Today, utilities are experimenting with AI across a broad swath of use cases, including everything from fault detection and predictive maintenance to energy forecasts, vegetation management, and automated report generation. 

Despite those early deployments, though, the report found that most utilities lack a central AI strategy. Some rely on department-level experimentation and pilots, while others are taking a wait-and-see approach to investments in the technology.

The handful of utilities that are taking a more organized approach spanning business units and use cases tend to see the technology’s adoption as a competitive advantage. National Grid, for example, is leveraging AI within advanced metering infrastructure to enable sub-second data sampling and real time load disaggregation. Avangrid is similarly invested in integrating AI into its operations, having hired an in-house AI development team that’s building out AI systems to do things like forecast grid performance and assess equipment health.

One of this camp, dubbed “innovators” in the report, said that “the integration of operational technology and IT, and everything driven by artificial intelligence, machine learning, data analytics — I think that's just the beginning.”

However, according to the report, most utilities currently sit on the other end of the spectrum. Around 70% of utilities are still in the early stages of understanding and adopting AI, and are prioritizing a more calculated — and therefore slower — approach to major strategic investments in the technology; one of those “observers” told the researchers that “major AI strategic investments” are “probably five years out.” 

Within those utilities, AI is being managed as an emerging technology, with a focus on market monitoring and knowledge acquisition, as well as customer engagement or operations pilots.

While the current approach of department-level experimentation will help those utilities prove the value of AI applications, the report found that a centralized strategy and approach will be a key step to broader-scale deployment and integration. That shift will ultimately define the market’s near-term evolution, easing more reticent utilities into low-risk use cases like vegetation management and allowing them to lay the groundwork for longer-term, higher impact deployments.

The pace and direction of that evolution, and the success of vendors targeting the power sector, will hinge largely on three factors: 1) data availability, a widespread challenge when it comes to AI’s potential in the power sector, 2) risk, i.e. the likelihood of snags like data biases and system failures, and 3) upside, i.e. whether the potential outcomes justify the upfront investment in data collection and implementation, and outweigh the risks.