Perspectives
“AI-powered” is meaningless: why energy networks need a sharper strategy
This article was originally published in Utility Week
As system operators accelerate their use of AI, from NESO’s Volta programme in the control room to SSEN T’s community consultation, the conversation across the wider energy sector has yet to catch up. “AI-powered” is now used to describe everything from rules-based automation and predictive machine learning to generative AI, while revealing little about what’s actually being proposed. The result is a maturity gap between how AI is being discussed and how it can be used.
In a safety-critical, regulated industry, that gap makes AI harder to adopt, govern and deploy where it could genuinely deliver value.
At the same time, power networks are being asked to solve ever more complex problems. From connections to control rooms, operators are expected to make critical decisions, faster, often with incomplete data.
If the sector is serious about maintaining resilience and reducing bills whilst scaling to support the transition to net zero, it needs a more precise, problem-led and governed approach to AI. Three areas stand out:
1. “AI-powered” masks what matters
Widespread consumer adoption of Large Language Models has made AI feel both transformative and immediately applicable everywhere. But consumer familiarity is also inflating expectations, creating the impression that all AI is equally capable and ready for live operational use. ChatGPT is highly effective around language-based tasks but it’s not going to design networks. The MCP servers that accelerated agentic AI are impressive but agents are not about to run critical national infrastructure.
What’s more, the term “AI-powered” collapses fundamentally different approaches into a single category. Machine learning, generative AI and agentic AI all operate in distinct ways in live operations, carrying different risks, data requirements and governance needs. Treating them as interchangeable makes it harder to assess whether a tool is appropriate for the decision it is meant to support, and allows vague terms to enter the conversation without being tied to a specific trade-off or outcome.
2. AI’s value: better decisions
A more effective approach is to start with a valuable business problem and ask how emerging technology, including AI, can help solve it. In energy networks, some of the most intractable problems lie in system planning and asset management.
For planners, connections are the most urgent challenge. Demand for increased grid capacity is rising fast, but reinforcing the network takes years. That leaves planners under pressure to get more from the existing system while weighing flexibility, demand, curtailment and risk. In this context, data and AI can help planners model the network, account for market dynamics and make decisions that are tied to outcomes such as faster connections or lower system cost.
In control rooms, operators are responsible for operational resilience, but now they have to consider financial efficiency too. Supporting them includes bringing together fragmented data, from SCADA to dispatchable flex assets, and boosting their decision-making with scenario modelling based on past performance and future predictions.
A useful approach is starting to emerge. UK Power Networks’ DSO is exploring how network constraints could be alleviated using AI. When forecasting tools show that generation is likely to push part of the network above its threshold, operators often have to decide manually whether to deploy a flexibility service or carry out switching operations. The team’s aspiration is to understand whether AI support for day-ahead optimisation would improve reliability and help address longer-term planning challenges. The value here lies not in adding an “AI-powered” label, but in applying the right tools to real operational decisions in an environment where resilience is critical.
3. Governance turns AI into impact
If “AI-powered” fails to articulate what a system does, weak governance obscures whether it should be used at all - and whether it will deliver value in practice.
As organisations move from experimentation to operational use, AI goes from being a technology problem to becoming one of adoption, cost and value. This shift requires a coherent framework of policies, processes, accountability and oversight that surrounds every AI deployment.
Governance must answer three practical questions:
- Where can AI be used safely in a regulated, safety-critical environment?
- Is this worth building - will it change a decision or behaviour?
- Can it be trusted to operate reliably, at scale and over time?
Without this clarity, even well-performing models fail to translate into better outcomes because ownership, controls and adoption are not in place.
Done well, governance is an enabler, not a constraint, on innovation. Northumbrian Water is known for innovating across infrastructure networks including using AI in its smart sewer project to help prevent spills. Now they are getting ahead by investing in AI governance too. In practice, that means being clear about strategy, governance and capability: strategy sets direction and the course of action, governance defines boundaries and decision rights, and capability determines how systems are built and run.
Final thought
In energy networks, AI is often presented as the answer, but that framing is backwards. The starting point isn’t the technology, it’s the business problem, the decision that needs improving and the value the organisation is trying to create. The real solution is building a resilient system that brings data together to solve a meaningful business problem, with AI as part of that system not as the strategy itself.
The organisations that get this right will be the ones that use AI precisely, govern it clearly and trust their systems to work reliably, at scale and over time.
Marcus Gardiner
Client Director - Energy and Utilities
24 June 2026
Subscribe
to our monthly newsletter for our latest expert content.
About the AuthorMarcus Gardiner
Marcus is a technical leader responsible for clients, strategy, growth and technical vision in Softwire’s Energy & Utilities sector. Having had a first career as a business consultant before becoming a software developer, Marcus has a unique view of how to change what’s possible in businesses through people and technology. From open data for energy trading decisions to personalised support for vulnerable customers, Marcus is immensely proud to lead teams that securely handle vast volumes of live data, build resilient 24/7 systems with care, and deploy AI in ways people understand and trust. Marcus speaks regularly at industry events, co-hosted LeadDev London 2024, sits on the LDX3 programming committee, and enjoys sharing ideas with others in the tech and energy communities.
Read more
