.svg)
AI data governance was a central theme at Utility Analytics Week 2025, which brought together leaders from across the energy and utility sector. The event explored how AI-powered systems, data governance, and analytics are reshaping decision-making in the industry.
From human-first change leadership to practical applications of AI analytics, these are our top takeaways and what they mean for the future of AI utilities.
If your AI initiative isn’t landing with your teams, the issue may not be technical. As Timothy Krall (Exelon) shared in his session on leading analytics through change, adoption is human first, technical second. Whether it’s an AI-enabled chatbot or outage prediction dashboards, success hinges on integrating change management from the start.
Key takeaway: Build with stakeholders, not just for them. Measure success by usage, not delivery.
AI without governance is a fast car with no brakes. Trust in AI doesn’t come from the algorithm, it comes from how teams govern, communicate, and validate it. In the AI at the Crossroads panel, speakers emphasized that AI data governance needs both structure and community. Samantha Grant (Dominion) shared how careful, phased efforts with co-pilot AI and transparency helped build trust prior to scaling.
A major theme was context engineering — ensuring AI fits into real-world business workflows and isn’t just used blindly. That includes human review of AI outputs, especially in safety-critical or customer-facing processes.
Strong AI data governance enables safer, more scalable deployment of AI systems, particularly in high-risk operational areas.
Key Takeaway: Governance isn’t just policy, it’s trust. You need it in place before you scale AI.
Utility safety programs have always been data-driven, but now they’re evolving into something more predictive and intentional. In the Community Conversations: Safety Analytics session, CPS Energy showed how increasing safety observations correlates to fewer injuries.
But it’s not just about logging more data. What matters is why teams report, not how often.
With AI-enabled predictive analytics, safety teams can identify patterns and act on them before incidents occur.
Key Takeaway: Culture drives data quality. Incentivize engagement, not check-the-box metrics.
Utilities are shifting from centralized, slow-moving systems to AI-powered edge intelligence — using next-gen meters, sensors, and AI to act on data faster and more locally where the action happens. Landis+Gyr shared how they now resolve network constraints and data issues in under 8 weeks thanks to fast-lane innovation strategies and closer-to-the-source analytics. This new model also supports AI-driven automation, allowing utilities to adapt more quickly to grid demands and disruptions.
The edge-to-cloud model offers speed and flexibility, but it also raises the bar for reliability and coordination. Teams can’t wait weeks to act on insights — the grid needs real-time awareness and responsive tools.
Key Takeaway: Grid analytics are moving to the edge, and utilities need infrastructure and culture ready to move with them.
There’s a growing focus on agentic AI and automation, but human oversight isn’t going anywhere. In this deep-dive session, teams walked through their model selection, testing, and validation processes. They emphasized the importance of:
Teams are using advanced AI models and machine learning algorithms to optimize everything from predictive maintenance to customer satisfaction.
Security and ROI are also front and center. Some AI tools cost thousands per day to run, so teams run detailed cost-benefit analyses before scaling. And they design every pipeline with built-in validation, so flawed outputs don’t reach the customer or the grid.
The decision-making process now depends on a blend of human expertise and AI-driven insights, grounded in governance and oversight.
Key Takeaway: Deploying AI at scale isn’t about doing more. Rather, it’s about doing it safely, securely, and with a clear ROI.
AI has the potential to significantly strengthen AI data governance by optimizing how data is understood, monitored, and maintained. AI can:
It also helps modernize data practices by providing continuous monitoring and intelligent automation.
The key is pairing AI with strong human oversight and clear policies. AI can reduce manual overhead and bring visibility to sprawling data environments, but it still needs well-defined rules and ownership to be effective.
To measure AI ROI, utilities should look beyond technical metrics like model accuracy and focus on business outcomes.
Key indicators include:
AI utilities benefit most when ROI includes both operational performance and improved customer satisfaction. Start by identifying the problem the AI is meant to solve, then measure what changed, whether operationally or financially, after implementation. ROI is about efficiency and impact.
Start with the understanding that not everyone needs to become a data scientist — but everyone should be able to understand and use data in their role.
Proven strategies include:
Integrating AI training into workforce development helps teams work confidently with AI systems and interpret valuable insights.
Tailor the approach by department and focus on outcomes, not just tools. Data literacy builds confidence, reduces resistance, and creates a stronger foundation for analytics adoption.
Across every session at UA Week, a common thread emerged: successful innovation isn’t flashy, it’s grounded. It’s legal teams co-creating with data scientists. It’s field crews shaping the training curriculum. It’s governance structures that support creativity instead of slowing it down.
For utilities, digital transformation isn’t optional. But making it sustainable means treating change as part of the product, not as a separate process.
What challenges are you solving on your analytics journey? Whether you're building a modernized data foundation or deploying AI into mission-critical systems, Launch is here to help.
Connect with a Navigator and start using AI more effectively — smarter, safer, and with real business value.
