The Gulf's AI Maturity Test, According to OpenAI

Enterprise software vendors have spent years telling Gulf regulators that their data would stay exactly where it was told to. That promise has rarely been enough on its own for regulated industries like banking, where compliance teams tend to ask a second question once the first is answered: where is the data actually processed.

Closing the loop between storage and processing

OpenAI addressed that gap this year in two parts, first with in-region data residency for ChatGPT Enterprise in the UAE, then with inference residency, closing the loop between storage and processing within the country. Farouk El Hamzawi, GTM Director at OpenAI, said local storage alone was never going to be enough to satisfy regulated organisations across the Gulf.

In-region data residency was a prerequisite, he said, but local storage alone was not the full answer. For many organisations, the missing piece sat one layer deeper: inference residency, meaning where the actual computation happens, rather than simply where the data sits.

That gap has since closed, with inference residency now available in the UAE, meaning data can be stored and processed there on dedicated GPUs rather than infrastructure elsewhere. This creates a closed regional loop for both data storage and processing, El Hamzawi said, allowing OpenAI to work with a much broader set of highly regulated organisations. He was careful to frame this as an early milestone rather than a finished story, noting there is more to share soon about what the capability actually enables.

Three layers of adoption, moving at different speeds

Enterprise adoption across the region breaks down into three distinct layers, according to El Hamzawi, each moving at a different pace but reinforcing the others. The first is workforce adoption, giving employees tools that improve everyday work. Organisations across the UAE, Saudi Arabia, and Morocco are already deploying ChatGPT Enterprise and Codex at scale, he said, helping teams improve productivity across coding, research, and documentation.

The second layer moves further into how organisations actually operate, using AI agents to reimagine entire workflows, with agents working alongside employees rather than simply assisting them, automating routine tasks until they become part of the operating model itself. The third layer goes further still: building AI-native products and services that create new customer experiences and business models across an industry.

None of this, he argued, is really about adoption for its own sake. The real opportunity is not simply adopting AI, but embedding it into how organisations operate, which takes leadership, change management, and trust, not just the technology itself. That framing closely mirrors how Gulf enterprises more broadly have come to treat AI adoption as a matter of national competitiveness rather than a purely internal productivity exercise, with government mandate and enterprise rollout increasingly moving in step with one another rather than the latter lagging behind the former.

Why inference residency unlocked what data residency alone could not

He pointed to what an in-country deployment model actually unlocks for a bank specifically: keeping data processing within UAE jurisdiction, aligning with local regulatory and localisation expectations, reducing latency for real-time use cases, and operating within enterprise-grade security frameworks. Taken together, he said, this lets a bank deploy generative and agentic AI workloads while still meeting the region's stringent requirements around data governance, privacy, and regulatory compliance.

That expansion, in his account, is tied directly to regulatory readiness rather than running ahead of it, and he expects the region's presence to accelerate considerably now that the foundational infrastructure is in place, a dependency that echoes how the broader UAE-US AI partnership has treated physical and regulatory infrastructure as the prerequisite that unlocks enterprise scale deployment, rather than infrastructure catching up after adoption has already outpaced it.

Cybersecurity keeps surfacing as the common thread

Rather than singling out laggard sectors, El Hamzawi pointed to a broad shift underway across government, banking, energy, and healthcare, all moving past pilots into real deployment. One factor kept surfacing across nearly every sector he mentioned: cybersecurity. Organisations across the region are focused on ensuring their systems remain resilient against evolving threats, a priority serious enough that OpenAI launched a dedicated initiative for it.

Through the Daybreak program, he said, OpenAI has been able to partner closely with customers on that specific priority, reflecting how tightly enterprise AI adoption and security posture are now linked in the minds of Gulf compliance and technology leadership.

Where the push for adoption is actually coming from

Asked whether momentum was coming from employees or from leadership, El Hamzawi declined to pick a side, arguing the balance shifts depending on sector and use case. What gives ChatGPT a specific advantage, in his view, is familiarity. Millions of people already use it in their personal lives, and that existing comfort translates into genuine bottom-up demand at work, with far less resistance than typical enterprise software faces.

The bigger shifts, though, happen once leadership gets involved. The biggest transformations happen when CEOs and boards enter the conversation, he said, pointing to enterprise-wide priorities like workforce enablement, upskilling, and cybersecurity as things that only scale successfully with real commitment from the top. The strongest outcomes come when both forces move together: executives committed to transformation, employees already eager to use the technology, a dynamic that lines up with separate findings that AI adoption in the Gulf tends to widen the scope of what individual employees take on rather than simply making existing tasks lighter, making genuine leadership buy-in more consequential than a purely bottom-up rollout would achieve alone.

Pushing back on the Arabic-first models narrative

El Hamzawi gently pushed back on the premise that Arabic-first models are the missing piece for the region. The conversation itself is not new, he noted, but the data has not shifted in that direction. Across benchmarks, OpenAI's general-purpose models have continued to perform better than most specialised models, he said, comparing it to a debate that arises in nearly every industry vertical over whether a company actually needs a finance- or healthcare-specific tool rather than a strong general-purpose one.

As frontier models keep improving, he argued, they tend to outperform narrower, specialised alternatives built around a single use case or language. OpenAI's approach follows the same logic: building the strongest possible models and ensuring they perform well across as many languages as possible, Arabic included, rather than building separate regional versions.

What the next eighteen months look like

The region's trajectory looks strong, in El Hamzawi's view, though with a specific caveat about how it gets there. Over the next eighteen months, the Gulf has the potential to overtake many other regions in enterprise AI maturity, he said.

The risk sits in exactly the kind of careful groundwork organisations have been laying, isolated ecosystems built to be thorough that end up, almost accidentally, delaying full-scale deployment instead. Part of the problem is speed, since model development moves fast enough that a static foundation, or a pilot that drags on indefinitely, struggles to keep up with what is actually available.

The better approach, he argued, is deploying agents directly into production, scaling them quickly, and learning from real usage rather than theoretical readiness. As general-purpose models improve, some highly specific agents or use cases can become obsolete very quickly, he said, arguing that the most mature enterprises will be the ones that keep learning inside production rather than holding out for a perfect pilot. That approach, he added, is already playing out with organisations across MENA that OpenAI is working with directly, deployments he expects to become reference points for the rest of the market.