By Mahesh Paolini-Subramanya, Founder and CTO, BKN301
AI is often described as the next operating system (OS) for financial services. Today, it is already central to operations, embedded in scoring transactions, routing customer interactions, flagging risk…the list goes on.
But there’s a misconception here arising in the industry. AI is in fact not the OS. Data is. Many financial institutions are trying to run cutting-edge applications on an OS that wasn’t designed with real-time intelligence in mind.
In the Middle East, the UAE, and Saudi Arabia consistently rank among the world’s AI leaders, but the region’s ambition has moved beyond adoption. In true form, it is looking to lead and shape how AI is built, governed, and operationalised globally.
That ambition requires more than bigger budgets and faster models alone. It requires a data layer stable enough to support them.
The Middle East’s advantage has always been long-term vision, with national strategies measured in decades rather than quarters. But lasting leadership cannot be built by upgrading applications while leaving the OS untouched. So, if the region wants today’s AI momentum to translate into lasting influence, it must invest early in the foundations that make intelligence trustworthy.
AI Is Scaling Faster Than Its Data Foundations
PwC research has reflected this pace of adoption. More than a third of Middle East CEOs report embedding AI directly into products and services, which is nearly double the global average. Yet beneath this progress, AI performance is starting to plateau, and it’s not because models have reached their limits, it’s more to do with the fact that the data they depend on has.
Adding to this, while 70% of CEOs in the Middle East report having a defined AI roadmap, results remain somewhat inconsistent. Some institutions see measurable gains in efficiency and responsiveness while others encounter erratic outputs, limited explainability, and increasing tension with risk and compliance teams.
I suspect the constraint here isn’t the AI-driven modern infrastructure itself, but the data sitting underneath that isn’t nearly as modern. We have to remember that AI does not ‘fix’ weak data foundations, instead, it holds up a magnifying glass. Running advanced models on fragmented data is like deploying modern software on an unstable OS.
Processes technically execute, but crashes, conflicts, and unexpected behaviour are inevitable.
More from Artificial Intelligence
Better Models Aren’t Fixing The Problem
If you’re newer to the AI tech stack, this dynamic is easier to understand through a more familiar lens. So, for example, no technology team would expect reliable behaviour from an application built on undocumented, constantly changing APIs. Developers would struggle to debug issues and observability and risk management would likely break down.
The issue here doesn’t have anything to do with the application logic, because the problem lies in the instability of the interfaces beneath it.
Now, shifting gears back to AI, data plays the same role. Models depend on consistent schemas, shared definitions, and reliable lineage. When those interfaces vary across systems or drift over time, AI outputs can be full of holes and difficult to explain, which is a recipe for loss of confidence in a technology already receiving significant skepticism.
Yet many financial institutions still operate data environments designed for a pre-AI era. Core systems were built for transaction processing and periodic reporting. Over time, digital channels, analytics tools and regulatory solutions have been layered on top. Data is copied, transformed, and redefined repeatedly to serve immediate needs.
The result is a patchwork data operating system. It works, but can it truly support today’s continuous, real-time intelligence?
Real Time Intelligence Also Means Real World Accountability
When we’re talking about real-time intelligence, we’re referring to AI systems that run continuously and influence decisions as they happen. This also means that they are drawing much closer attention from regulators. As automated decisions increasingly affect areas like credit approvals, fraud detection and pricing, these authorities are looking beyond model performance.
They want to know where the data comes from and how it is governed. In the UAE, regulators including the Central Bank, the Dubai Financial Services Authority (DFSA), and the Financial Services Regulatory Authority (FSRA) have made it clear that strong data governance and model risk controls are essential for using AI in financial services.
At the same time, expectations from customers are rising. Across the Middle East, people are used to fast, seamless digital services and are far less forgiving of delays, errors, or unexplained outcomes.
As a result, financial institutions are feeling the pressure from every direction. At the minimum, they need to be able to explain why a decision has been made and yet, for organisations with data fragmented across systems, this kind of transparency control is difficult (and expensive) to achieve.
Data Foundations Will Define The Middle East’s AI Leadership
For the Middle East to reach true, sustained global AI leadership, the next phase of AI must be modernising the data operating systems financial institutions run on. Data that was once good enough for reporting now needs to be reliable enough for automation with accuracy, traceability, and reproducibility fundamental to how effectively AI will operate at scale.
It’s perhaps no surprise that systems fail at their weakest layer and in the world of AI, that layer is almost always data structure, governance, and lineage, rather than model sophistication. The institutions that have already put the work into fixing this are now moving AI out of small pilots and into real, day-to-day operations. So, with the right partners, this is entirely possible.
Leaders in the region are in a unique position right now, not just to use AI at scale, but to future proof the foundations it will run on for decades to come.
