Nearly 95 per cent of AI initiatives reportedly fail, not because of technology, but due to broader systemic challenges. What are the most common root causes behind these failures?
The first and biggest issue is data. AI is only as good as the data it’s trained on and connected to. In large enterprises, data is often fragmented across multiple systems, inconsistent, poorly governed, and siloed. When AI is deployed on top of that environment, it produces unreliable outputs or “hallucinations” because it’s drawing from conflicting or low-quality inputs. Without a unified data strategy, across both structured and unstructured data, AI cannot deliver meaningful results.
The second major cause is that organisations treat AI purely as a technology deployment, when in reality it represents a business model shift. Many initiatives are built bottom-up, starting with tools and pilots, instead of being driven top-down by clear business outcomes. AI needs to be anchored in a well-defined data strategy and aligned with how it will create measurable value for the business.
If you ask the wrong question or feed in the wrong data, you won’t get the right answer, and that’s exactly what’s happening in many enterprises today.
How should organisations think about using AI?
For the past few years, most AI conversations have centred around copilots. A copilot is simply a system that assists a human, it augments decision-making but keeps the human in control. That’s valuable, but copilots can still hallucinate, so they must be carefully designed, governed, and connected to reliable data.
The more strategic conversation today is about agentic AI Agents go beyond assistance, they can take action. In the agentic model, there are five levels of automation, with the highest level representing full autonomy. While many talk about full automation, very few organisations are operating at that level today.
The key realisation for organisations, including those in the UAE seeing success, is that there is no such thing as a single, universal agent. You need a fleet of specialised agents: one for finance, one for HR, one for security, one for executive workflows, and so on. These agents must interoperate seamlessly, just as human teams do. And they must be intelligent and well-governed, because if one agent hallucinates and passes flawed output downstream, the entire chain of decisions becomes compromised.
Organisations should therefore think of copilots as level one, augmentation, and agents as part of a broader automation and transformation strategy. Moving to agents is not just a technical upgrade; it’s an organisational shift. It impacts workflows, accountability, skills, and workforce planning.
Just as robotics gradually transformed factory floors and surgical theatres, agentic AI is now beginning to reshape white-collar work. Routine, level-one tasks will increasingly be automated. That’s not inherently negative, but it does mean cognitive expectations for human roles will rise.
Each organisation must decide: Where do we want augmentation? Where do we want autonomy? And how does a fleet of agents align with our long-term business strategy?
Where does Anomali fit into this agentic AI landscape?
Anomali sits at the intersection of big data and agentic AI. We are fundamentally in the big data business, helping organisations unify their data in a single cloud-native platform. When data is better managed, AI performs better, and that’s the foundation for everything that follows. AI maturity evolves in stages; while many customers are still at the early levels, others are progressing quickly. Our role is to support them across that entire journey.
We provide a comprehensive agentic SOC platform that can be deployed flexibly. While our primary focus is security operations, the platform’s capabilities can extend into IT use cases as well. Customers can adopt the full platform or leverage us specifically for intelligence, the model is designed to be adaptable.
Looking three to five years ahead, I see Anomali operating within an open ecosystem of intelligent agents. Organisations will bring their own agents, and we will enhance and enrich them. Others will do the same. The future is collaborative and interoperable, and that shift is not optional, it’s simply the direction the industry is moving toward.
You emphasised on clean data. What should organisations actually do about their data?
The key is to have a single, unified data lake that manages all your data. This works best in the cloud, but even if you’re on-premises, hybrid, or air-gapped, you can still optimise and clean your data. It may not be perfect, but it will be sufficient to drive real value.
When I say “cloud” in the UAE context, I mean sovereign cloud. It doesn’t have to be a US-based public cloud. For example, we’ve just launched a cloud instance with AWS in the UAE and will do the same in Saudi, with more instances planned across other hyperscalers.
The first step for any organisation is a clear data strategy. If you can adopt cloud or sovereign cloud, you’re on the right path. If you already use an open data lake, that’s excellent, we can enhance it, make it smarter, and more secure. There are multiple approaches, but it always comes back to having a deliberate data strategy that works for your organisation, not someone else’s.
Given the high failure rate of AI projects, what should CISOs and CIOs do differently?
The common mistake is focusing on tools and technologies in isolation, without a clear view of business goals. Optimisation starts at the top, with strategy, not the middle with processes.
My advice is to think less about “CISO” and “CIO” as separate roles, and more about the business itself. Where is the business headed? What are its strategy and data priorities?
Once you understand that, bring it into your organisation. In some cases, it even makes sense to unify the CIO and CISO functions, they’re both working with the same data. The goal is to align data management and security with the company’s strategic direction. From there, you can ask: “Can we create a single, unified data lake?” If not, what smaller steps can we take to optimise data use?
What is Anomali doing in the GCC, and what does the future hold for the region?
We’ve made significant investments in the GCC, working with some of the region’s largest public and private sector organisations. Our approach is to co-innovate with customers rather than impose a one-size-fits-all solution — we want to hear their problems first and explore solutions collaboratively.
One of the region’s biggest advantages is the absence of legacy technology debt. Enterprises here are modern, with a young, educated workforce built on the latest cloud architectures. The UAE has clear ambitions in AI, and the region’s energy resources make large-scale deployments genuinely feasible. That’s why we’ve deployed a local cloud instance and plan to expand further across a multi-cloud model.
That said, the local threat landscape is intense, which makes securing AI even more critical. The hype often focuses on “AI native”, chips, data centres, infrastructure, but the real enterprise impact comes from “AI enhanced,” where AI is applied in daily operations.
Looking ahead, the shift for CISOs and CIOs globally will be from multiple point solutions to a single, unified data fabric. Intelligence will increasingly come from smart agents, and organisations will be expected to do more with less as data volumes grow and threats become more sophisticated. It’s a chess game against increasingly sophisticated attackers.
For Anomali, the focus remains on data intelligence – filtering and managing data to deliver measurable outcomes. One client improved critical incident detections by 90% year over year with our support. Over the past few years, we’ve more than tripled our business in the GCC, and our goal is to triple again, or achieve 10x growth. With major enterprises, oil companies, and government entities in the region, I’m highly optimistic about what’s ahead.
Discussion about this post