AI for Energy: Intelligence at the Heart of the Transition

The global energy system is undergoing a transformation unprecedented in both speed and complexity. Decarbonisation, decentralisation, electrification, and digitalisation are reshaping how energy is generated, distributed, and consumed.

Artificial intelligence is increasingly shaping this transition. The critical question is no longer whether AI has a role to play, but how it can be integrated efficiently, safely, and at scale. The webinar AI for Energy – Intelligence at the Heart of the Transition brought together perspectives from technology implementation and risk assurance to examine how AI is already being applied across the energy sector, where it is delivering tangible value, and what must change to enable wider adoption.

Energy systems are becoming fundamentally more complex. The growth of renewable generation, electrification of demand, decentralised assets, and flexible consumption is introducing volatility that traditional, manually operated systems were never designed to manage.

In this context, AI has a clear role to play in energy optimisation through high-frequency demand forecasting, real-time supply–demand balancing, automated optimisation of grid and plant operations, and predictive maintenance and asset performance management.

“The question isn’t really whether AI can be used to optimise energy and industry, but more how we do this effectively and safely, and in a way that doesn’t contribute more risk to energy security” - Harry Morphakis, Partner at Artefact

This shift reflects the operational reality now facing energy operators. Systems designed around centralised, dispatchable generation are increasingly required to manage intermittent renewable supply, more dynamic demand patterns, and a growing number of decentralised assets. Human-led decision-making, operating on hourly or daily cycles, is no longer sufficient to manage this level of complexity.

Supply-Demand Balancing

Across the discussion, supply–demand balancing emerged as the most critical and impactful application of AI in the near term. As renewable penetration increases, balancing the system becomes both more challenging and more valuable. Weather-driven generation, electrified transport and heating, and energy-intensive digital infrastructure such as data centres are all contributing to sharper peaks and troughs in demand.

Harry Morfakis highlighted that forecasting and optimisation are not experimental capabilities but largely solved technical problems. He pointed to real-world examples where AI-enabled forecasting has enabled dynamic tariffs and near real-time grid optimisation. In Sweden, for example, AI-driven models have reduced calculation times from minutes to seconds, allowing pricing and balancing decisions to be updated every fifteen minutes.

What makes this particularly significant is that improved forecasting does not merely reduce costs; it underpins energy security. Better anticipation of demand and supply enables operators to prevent system stress before it occurs, rather than responding after the fact. In an increasingly volatile system, this capability is becoming indispensable.

When asked to identify the single AI use case most likely to improve system performance by 2030, both speakers converged on the same answer: demand forecasting and supply–demand balancing. These applications offer immediate, system-wide benefits while laying the groundwork for more autonomous operation in the future.

“The energy sector is undergoing two historic transitions at the same time: the energy transition itself, and the digital transition. Together, they are increasing complexity and uncertainty across the system” - Matthew Celnik, Team Lead for Digital Trust, DNV

Despite the sophistication of modern AI tools, both speakers emphasised that technology itself is rarely the primary constraint. Instead, barriers arise from organisational readiness, legacy infrastructure, and cultural resistance to change. Energy systems are designed to prioritise stability and reliability as it is a matter of national security.

AI and Risk

However, this risk-averse culture can slow the adoption of technologies that require new ways of working and new forms of trust. Legacy operational technology, often decades old, complicates integration. Data may exist, but not at the frequency, quality, or reliability required for AI-driven decision-making. Even where sensors are installed, confidence in their calibration and consistency is not guaranteed.

AI systems cannot function without high-quality data, but just as importantly, operators must trust both the data and the outputs derived from it. A recurring challenge identified in the discussion was the gap between data engineers and energy domain experts, which can hinder both implementation and adoption.

Transferable AI skills

Perhaps the most structural challenge highlighted was the shortage of cross-disciplinary skills. The energy sector increasingly requires professionals who understand energy systems, data engineering, AI, and risk management simultaneously.

Matthew Celnik emphasised that this skills gap exists not only within industry, but also among regulators and public bodies. Governance frameworks cannot keep pace with technological change if those responsible for oversight lack the necessary expertise. Regulation, he argued, should act as an enabler of adoption—but only if it is informed and proportionate.

This shortage of hybrid expertise represents a long-term constraint. Addressing it will require sustained investment in training, education, and organisational design, rather than reliance on isolated technical specialists.

As AI systems move from advisory roles toward greater autonomy, questions of risk and accountability become more acute. The energy system is, by definition, safety-critical. Failures can cascade rapidly and affect millions of people.

Such failures rarely stem from a single cause. Instead, they emerge from complex interactions between technology, governance, human oversight, and organisational incentives. AI introduces new forms of systemic and emergent risk that must be managed proactively.

Crucially, neither speaker argued for risk avoidance. Instead, they advocated applying the same principles that have governed high-risk engineering systems for decades: independent assurance, clear boundaries, defined human–AI interfaces, and continuous monitoring.

Conclusion

The discussion made one point unequivocally clear: artificial intelligence will be central to the future performance, resilience, and security of energy systems. As energy networks become more decentralised, weather-dependent, and digitally interconnected, the level of complexity they must manage exceeds what traditional operating models were designed to handle. In this context, AI is not a future enhancement but a necessary capability.

However, the path to effective adoption is not primarily a technological one. The tools required to forecast demand, balance supply, and optimise assets already exist. The real challenge lies in embedding these tools into live, safety-critical systems in a way that is trusted by operators, supported by robust data, and governed by clear assurance frameworks. Without confidence in data quality, decision logic, and accountability, even the most advanced models will remain underutilised.

Join us for further discussions on the UK’s Net Zero ambitions and the future of the energy transition at The Foresight Event 2026 this February, and connect with industry leaders shaping the next phase of energy system transformation!

---

---