AI in Energy: People, Structure, and Scale

AI adoption in energy operations has moved past debate. The real challenge, as Natalia Klafke, Executive Vice President of Energy and Sustainability at Radix, observed at Impact 2025, is execution: many companies have pilots and dashboards, but few have built the data infrastructure, organizational alignment, and change management practices needed for sustained results.

Natalia Klafke, Executive Vice President of Energy & Sustainability at Radix, has spent over a decade helping energy companies modernize complex operations. At Impact 2025, she noted a clear change in the air:

“The event is way bigger this year. The maturity of discussions is getting more interesting. We see concrete achievements and way bolder goals.”

That maturity, however, comes with a reality check — progress is uneven. Many companies have pilots, dashboards, and proofs of concept, but few have built the foundation for sustained impact.

Why Do AI Pilots Fail to Scale in Energy?

AI pilots fail to scale in energy because implementation speed outpaces foundational readiness. Organizations that achieve meaningful results first invest in clean, connected, and contextual data as a shared operational resource, and anchor every digital initiative to a defined goal such as lowering energy intensity, improving throughput, or reducing unplanned shutdowns.

Strong data infrastructure is the starting point. Without it, no amount of modeling or automation can deliver consistent value. AI systems depend on data that is clean, connected, and contextual. Companies that treat data as a shared operational resource — not a departmental by-product — create a base for improvement that scales across sites and functions.

Equally important is strategy. Every digital initiative should serve a defined operational goal: lowering energy intensity, improving throughput, or reducing unplanned shutdowns. When the goal is clear, data teams and operations can work toward the same target instead of running parallel experiments.

What Causes Organizational Misalignment in Industrial AI?

Organizational misalignment is the most common reason technically sound AI projects stall in energy companies. According to Natalia Klafke of Radix, a new analytics tool can be deployed in days, but aligning departments around its purpose and use can take months, delaying adoption, eroding trust, and fragmenting data systems.

Energy companies operate across layers of safety, engineering, IT, and business management. Each has its own priorities, processes, and pace. When these groups move in different directions, even a technically sound project can stall.

Natalia Klafke has seen this repeatedly. A new analytics tool can be deployed in days, but aligning departments around its purpose and use can take months. The misalignment is costly — it delays adoption, erodes trust, and fragments data systems that were meant to integrate them.

Organizations that handle this well establish cross-functional teams from the beginning. They define a shared set of metrics, give decision-making authority to mixed teams, and make alignment a continuous process rather than a one-time meeting.

How Is AI Adoption Measured in Energy Operations?

In industrial energy operations, adoption is the only valid measure of AI success. A system ignored by operators, engineers, and technicians delivers no value regardless of technical performance. The most effective leaders track adoption as a production metric: daily active users, facilities rolled out, and whether insights are influencing real operational decisions.

In plants and refineries, where processes run 24/7, tools that are ignored or underused might as well not exist. High adoption signals that the people closest to the process — operators, engineers, technicians — see real value in the system. Low adoption, even with perfect code, shows that the solution doesn’t fit the workflow.

For this reason, the most effective leaders track adoption like a production metric. They ask: How many users rely on the system daily? How many facilities have rolled it out? Are the insights influencing real decisions? When those numbers rise, the rest — productivity, efficiency, safety — tends to follow.

Why Is Change Management Critical for Industrial AI?

Change management is a core operational function in industrial AI programs, not an internal communications task. Energy companies face a dual workforce challenge: retiring staff carry deep process knowledge, while incoming digitally native professionals bring data fluency. Bridging both groups through training, mentoring, and co-designed workflows directly affects system use, morale, and operational safety.

Energy companies face a dual challenge: a wave of retirements among experienced staff and the influx of younger, digitally native professionals. Each group has strengths — one brings deep process knowledge, the other fluency with data tools — but bridging them requires intention.

Companies that succeed view change management as operational work, not internal communications. They dedicate time to training, mentoring, and co-designing workflows that balance automation with human oversight. The result is not just better system use, but higher morale and safer operations.

How Does People Alignment Build System Resilience?

System resilience in energy operations depends on organizational alignment. Once teams can question data, adjust assumptions, and collaborate across disciplines, digital tools remain effective even as conditions change. The renewable energy sector demonstrates this in real time, where shifting weather, variable generation, and evolving grid demands test even well-designed systems continuously.

This is where the renewable energy sector offers valuable lessons. Renewable operations are dynamic by nature — subject to shifting weather, variable generation, and evolving grid demands. They show, in real time, how fragile even the best digital systems can become without context and adaptability.

How Do Renewables Manage Data Overload?

Renewable energy operations face a data overload problem, not a data scarcity problem. Turbines, inverters, and grid systems generate continuous streams, but without context, the data becomes noise. Leading operators are responding by recalibrating models more frequently, combining AI forecasts with human domain expertise, and prioritizing adaptability over point-in-time prediction accuracy.

Forecasting renewable generation now involves more volatility than ever — weather behavior, load shifts, and the rising power demand from AI data centers all interact in unpredictable ways. Models built for yesterday’s stability are struggling to keep up with today’s variability.

Leading operators are adapting by recalibrating models more frequently, combining AI forecasts with human domain expertise, and focusing on resilience rather than prediction alone. The emphasis is shifting from accuracy in a snapshot to adaptability over time.

The same principle applies beyond renewables: resilient systems require resilient organizations. When teams are equipped to question data, adjust assumptions, and collaborate across disciplines, digital tools can remain effective even as conditions evolve.

How Is AI Increasing Electricity Demand?

AI is now a significant driver of electricity demand. The large-scale computing centers that power modern AI consume vast amounts of energy, adding a new and growing variable to the same grids that energy operators are trying to optimize. This creates a feedback loop where the tools used to manage energy supply are also increasing the load those systems must serve.

The same technologies being used to optimize operations are driving a surge in electricity demand. Large-scale computing centers that power modern AI now consume vast amounts of energy — sometimes comparable to industrial plants.

This growing interdependence between the energy and technology sectors is forcing a reassessment of what “efficiency” means. Improving algorithmic performance or data center cooling isn’t just a technical issue; it’s now part of global energy planning. Digital efficiency and energy efficiency can no longer be managed in isolation.

No Way Back

AI’s role in the energy sector is no longer experimental. It’s part of the operating fabric — guiding maintenance, predicting failures, balancing grids, and informing strategy. The question is no longer whether it should be used, but how well it is governed.

The companies moving forward understand that technology alone doesn’t change outcomes. Structure, clarity, and trust do. AI’s potential in energy will be realized not through faster adoption of tools, but through more deliberate alignment of people and systems.

As Natalia Klafke put it: “Technology isn’t the challenge. Agreement is.”

1. What does Radix’s Natalia Klafke say is the biggest barrier to AI success in energy?

According to Natalia Klafke, Executive Vice President of Energy and Sustainability at Radix, the biggest barrier is organizational misalignment, not technical failure. Aligning departments across safety, engineering, IT, and business management around a shared purpose takes far longer than deploying the tool itself. Companies that address this by forming cross-functional teams with shared metrics and real decision-making authority see the strongest results.

2. How should energy companies measure AI adoption in industrial operations?

Energy companies should measure AI adoption the way they measure production metrics. The relevant questions are: how many users rely on the system daily, how many facilities have deployed it, and whether the insights it generates are influencing actual operational decisions. When adoption numbers rise, downstream outcomes including productivity, efficiency, and safety tend to follow, according to the practices described by Radix.

3. How does AI adoption in renewables differ from fossil fuel energy operations?

Renewable energy operations face greater data volume and variability than traditional fossil fuel settings. Turbines, inverters, and grid systems stream continuous data subject to weather shifts, load changes, and surging power demand from AI data centers. Where fossil fuel operations often focus on process stability, renewable operators must prioritize model recalibration, human domain expertise alongside AI forecasts, and resilience over static prediction accuracy.

4. Why is data infrastructure the starting point for AI in energy companies?

Data infrastructure is the starting point because AI systems depend on data that is clean, connected, and contextual. Without it, modeling and automation cannot deliver consistent value regardless of how sophisticated the AI tools are. Energy companies that treat data as a shared operational resource across sites and functions, rather than a departmental by-product, create a foundation that supports scaling across the organization.