Why energy and infrastructure leaders need less AI hype and more operational value

How much of AI’s promise is real and how much is just talk? For engineering and operations leaders, the real test is whether AI can drive measurable impact on asset performance.

Scepticism about AI in asset management is reasonable. For many engineering and operations leaders, the reality so far has been a lot of noise, a few pilots and not much measurable value. If that’s your experience, you’re not wrong to want more before committing further.

In asset-intensive environments, the stakes are too high for vague promises. Decisions affect safety, uptime, compliance and capital performance. Trusting a black box because it’s labelled ‘AI’ isn’t good enough. You need to know what the technology is actually doing and have confidence that it will improve operational performance and deliver the commercial outcomes you want.

That means working with people who will be honest about what AI is and isn’t, understand your operational context and can bridge the gap between data science and engineering. Ultimately, it comes down to trust.

Cutting through the AI noise

Much of the confusion around AI comes from treating it as a single thing. It isn’t. There are three distinct types and they do very different jobs:

Traditional AI – often called machine learning – works with structured data to find patterns and support forecasting and classification. In asset management, that means predicting failures or surfacing risk factors that are difficult to spot manually across large datasets. Many organisations are already using some version of this.

Generative AI – which combines large languages models (LLM) and deep learning – works with the messy, unstructured information that traditional AI can’t handle. Things such as inspection images, technician notes, maintenance histories, project reports and manuals. It can interpret a field photo, summarise years of maintenance records or make dense technical documentation actually usable. In short, it unlocks knowledge that would otherwise stay buried.

Agentic AI goes further still. Rather than producing insight for someone to act on, it connects workflows, monitors conditions, triggers alerts and suggests next steps across systems. It doesn’t just flag a reliability issue – it pulls together the relevant data and proposes a prioritised response.

Value is cumulative

These three AI types aren’t competing – they build on each other. Traditional AI predicts, generative AI interprets and agentic AI helps you act. The real impact comes when they’re combined in a coherent decision framework. The organisations seeing genuine returns are those using a layered approach – turning data into insight and insight into timely action across the asset lifecycle.

Take a major energy operator. The organisation wanted to reduce O&M costs without sacrificing availability. We used all three types of AI to develop a solution:

• Traditional AI analyses inverter performance data, weather-adjusted generation and historical fault logs to rank assets not by age – which is what most operators still do – but by degradation risk.
• Generative AI then works through five years of unstructured maintenance notes and inspection reports for those flagged assets, surfacing recurring fault signatures that never made it into structured fields and would otherwise require a senior engineer to review manually.
• Agentic AI connects that output to the scheduling system – proposing bundled maintenance visits optimised by geography and contractor availability, with the draft work scope already populated.

Where AI delivers for asset-heavy organisations

The real test of AI in energy and infrastructure is simple. Does it improve everyday operational decisions? In asset-intensive environments, value shows up in a handful of repeatable ways. Catching equipment failures early, before they cause downtime or safety issues. Spotting performance gaps across large asset portfolios, so effort and budget go where they’ll have the biggest impact. Bringing together operational, financial and environmental data so leaders can make more balanced decisions about performance, cost and long-term asset value. And giving engineers structured, decision-ready insights within their existing workflows, rather than making them hunt across systems for the information they need.

The gap that holds organisations back

Turning AI insights into action is where many asset-intensive organisations get stuck. They invest in data platforms and AI models, and those models do start generating useful outputs. But the insights often sit in dashboards that engineers rarely open. Reports get generated but not acted on. The models run alongside operational processes rather than inside them.

The problem isn’t usually the technology. It’s the gap between insight and action. If intelligence doesn’t reach the right people at the right moment, fit naturally into how decisions are made and connect to real operational priorities, it rarely changes behaviour.

The solution lies in ensuring that dashboards and insights are decision led. That means asking right at the start: What decision are we trying to improve? Who makes that decision and when? What information do they need to make a better decision? What format should that information be in? Get those answers right at the start and the rest of the implementation has something real to build on.

What getting it right looks like

The organisations that succeed with decision intelligence see it as an engineering challenge, with technology as the enabler. That mindset matters because engineers and technicians are sceptical of AI for good reason. They’ve seen technology programmes overpromise before and carry real accountability if something fails. They also hold critical knowledge about their assets that never makes it into any dataset.

Earning their trust takes more than good intentions – it requires deliberate change management. Involve sceptics early in the design process, focus pilots on genuine business problems and keep humans in the loop so AI never triggers actions without sign-off in the early stages. Just as importantly, make the model’s reasoning visible so decisions can be understood, challenged and trusted.

From hype to practical application

The truth about AI is that it won't fix a poorly defined problem or act as a substitute for operational judgment. What it can do – when applied to the right decisions, with the right data and people who understand both the technology and the asset – is meaningfully improve asset performance and deliver the commercial outcomes that matter to your business. If that's what you're working towards, let’s talk.