Most conversations about AI in regulated industries still start from the wrong assumption: that the hard part is getting the technology to work, while the reality differs quite a bit. Across every energy, utilities, and industrial organization we work with, the model, the chatbot, the anomaly detector, almost always works. The harder problem shows up right after, when someone tries to take that one working pilot and run it across the enterprise, and it stalls. That’s the point where a lot of pilots stall, working fine on their own, but with no path to running at scale.
Scaling AI beyond pilots: the real barriers leaders named in Cairo
We saw this play out a week ago, in a room in Cairo with around 40 senior leaders from Egypt’s energy and utilities sector: general managers, CIOs, ICT directors, AI leads, and people running exploration and production for organizations that keep a country running. Before they walked in, we asked them one question: what is the biggest thing standing between your organization and scaling AI beyond pilots?
Almost nobody said they were still deciding whether AI was worth pursuing, and almost nobody was asking us to explain what a large language model actually is. The challenges people raised were about what happens after the first AI project works:
- security and data confidentiality
- the difficulty of moving beyond a pilot once legacy systems and years of accumulated infrastructure get involved
- proving return on investment to a board
- getting data clean and connected enough to trust (AI-ready data)
- the absence of any real AI governance or trust framework once an AI agent is making decisions on its own.
That pattern is not unique to Egypt, and it is not unique to energy. It shows up in every regulated and safety-critical environment we work in, built as they usually are on decades of systems that were never designed to talk to each other, and now being asked to run autonomous and AI-driven operations on top of that foundation. Cairo just made the pattern visible, because the room was full of people already past the experimentation stage and asking the harder questions early about AI governance, data quality, and how to move from pilot to production without multiplying risk.
Why a working pilot doesn’t always mean it’s ready to scale
Here is the pattern we keep seeing, in Egypt and everywhere else we operate. An organization runs a successful AI pilot: a chatbot that resolves tickets, a model that flags anomalies, an agent that drafts a report, and it works well in isolation, on a narrow task with a clear boundary around it. Then someone tries to take it enterprise-wide, and it stalls, usually because there is no single place where that AI can see the data it needs, follow the workflow it’s supposed to follow, and get governed the way a regulated business requires. Five systems, three data silos, and no consistent way to say who or what approved a given action tends to be the real story behind a stalled rollout, far more often than the AI itself falling short, and the platform sitting underneath the AI is usually where that barrier lives.
This is why we think about AI differently than a lot of the market currently frames it, and it’s worth being specific about what that means in practice. Treating AI as a feature bolted onto an existing technology stack tends to be where the trouble starts, because it leaves every agent operating with its own rules and its own blind spots. We think of it instead as a capability that has to sit on top of a single platform that already connects an organization’s data, workflows, and people, so that when an AI agent acts, it acts inside the same structure of approvals, audit trails and accountability that governs everything else in the business. Skip that foundation data governance, identity, and AI agent governance and scaling AI mostly means multiplying risk faster than value, which is the opposite of what any board wants from the investment.
Questions to ask an AI partner in a regulated industry
If you’re evaluating who to bring in for this kind of work, the more useful question to ask goes beyond whether a partner can implement an AI tool. Whether they have already done the harder work of governing AI at scale, in an environment as regulated and complex as yours, and can show you what that looked like, tells you far more.
A few things worth pressure-testing in any conversation like that:
- Has the partner run AI in production, at scale, in a regulated industry, or is the story still mostly about pilots and proofs of concept?
- Does their model start with your highest-priority business problem, or with a list of modules they’d like to sell you?
- Do they talk about governance, identity, and risk management as a core part of the AI conversation, or as an afterthought once something goes wrong?
- Are they committing to your market for the long term, with local presence and accountability, or passing through on a project basis?
We’ve built our own answer to those questions over more than a decade of ServiceNow-only delivery across Europe, the Middle East and beyond, including in energy organizations that operate under exactly the kind of regulatory and safety pressure this sector knows well. Our commercial model reflects the same thinking: we start with the use case that matters most to your business, prove the value there, and expand from a position of demonstrated trust. It’s the same approach we brought to the room in Cairo, and the same one we bring to every market we enter.
Where this leaves the conversation on AI governance and scale
Whether you were in that room in Cairo or are reading this having never heard of the event, the underlying question is the same. The organizations asking about governance, ROI, and scaling past the pilot aren’t behind, they’re asking the right questions earlier than most markets do. The real differentiator over the next few years will be who built the governed foundation to scale AI safely, and who chose a partner able to prove they’ve done it before.
That’s the conversation we’re here to have, wherever you happen to be in it right now, and however far along your organization already is.
Need additional advice or information? Contact our experts who will be happy to discuss your specific use case.
