The share of UAE businesses using artificial intelligence rose from 53% to 72% in a single year, growth of 36% that comfortably outpaced Europe, where adoption climbed 29% to reach 54%. The figure comes from “Unlocking the UAE’s AI Potential 2026”, research carried out by Strand Partners for Amazon Web Services in collaboration with the UAE Artificial Intelligence Office and launched at the AWS Summit in Dubai this week.
The numbers that matter more to anyone holding an enterprise technology budget sit further into the report. Its survey of 1,000 business leaders and 1,000 citizens found that 47% of AI adopters are still using public chatbots and ready-made tools for incremental efficiency gains, while only 31% have reached the stage of custom systems, multi-model architectures and autonomous agents, which is where the report locates the real productivity value.
Chris Erasmus, Managing Director for AWS, UAE & MENA Growth Markets, attributed the speed of the climb to national ambition and a government that has set the pace. “I absolutely believe it’s ambition. The fact that you have the government leading the way is another big part of that, and so is the urgency,” he said.
The report supports his reading with Abu Dhabi’s pledge to become the world’s first fully AI-native government by 2027. Erasmus was equally direct about where most of that adoption still sits. “I think it’s still very much sitting at the very fundamental level: can I ask a question, can I get a response?” he said.
Large companies are slipping into AI’s lower tier
The sharpest divide in the data runs along company size. Among startups, 88% have adopted AI and 53% of those adopters are already working with its most advanced applications, while 61% report transformative or significant productivity gains. Large businesses sit at 51% adoption, behind both the startups and SMEs, whose 68% trails the national average only slightly.
The report describes the result as “an emerging two-tier AI economy”, in which AI-first companies such as Property Finder, Maqsam and Careem scale at speed while many established organisations struggle to move beyond pilots. The gap is widest on agentic AI, where only 32% of businesses say they have heard of the term, and among those that have, 9% have fully deployed it and a further 9% are piloting.
Erasmus said the pressure to experiment had been the right instinct, and that the next stage demands something different from organisations. “There’s an urgency to move quickly, to try the technology, to find the outcome, and that is critically important,” he said.
What has to follow, in his account, is a change in how companies talk about the technology internally. “What we have to see is this move from ‘I want to test the technology’ to ‘it’s tried, it’s trusted, it’s proven. I have very specific outcomes for my customers and my internal processes that I now want to apply it to and scale,'” he said.
When GEC Newswire asked whether failed pilots and disappointing results had slowed that shift, he described organisations returning to first principles after an initial phase spent learning what the technology could do. “A lot of organisations are now going back and asking what does my data strategy look like? Do I have all the guardrails I need? Do the skills exist today within my organisation, and do I have the mandate and the culture that allows us to go?” he said.
A 39% pay premium is masking a training gap
Skills are the barrier the report returns to most often. Nearly half of organisations (48%) cite a lack of digital skills as a major obstacle, and it takes an average of 5.8 months to hire someone with the right expertise. In hospitality, 71% name the skills shortage as their biggest barrier.
Businesses are responding mainly by bidding for talent. Some 81% say they would raise salary offers for candidates with the right digital skills, by an average of 39%, while only 31% of employees have received any digital skills training at all.
Erasmus located the weakness in what happens after people have been given access to learning material. “Many organisations have made that easy to access. You can get access to the curriculum, to the model, you can learn about something. Marrying that with the ability to play with it is the bit where the balance has to get better,” he said.
The second gap, he argued, appears when trained staff return to their desks with no formal way to use what they have learnt. “You upskill individuals, and then you have to give them the opportunity to drive the outcome with it as well. Access to knowledge, plus the ability to go and drive something physically, is what we focus so much on,” he said.
He described the role of the tools themselves in similar terms. Pointing to Kiro, AWS’s agentic development environment, he said such products exist to support the people using them. “Ultimately, that is there to supplement what a software development engineer is doing,” he said.
Curiosity, in his view, does the rest once the tools are in people’s hands. “I often speak about the natural curiosity of picking up the tools and doing something with them,” he said.
Businesses want sovereignty without paying for it
A third of businesses (34%) say a lack of clarity and confidence around data control, access and governance is holding back their expansion of AI. In financial services, the sector furthest ahead with 45% at advanced adoption, the leading barrier is the complexity of operating AI and digital services across borders (48%).
The same businesses want sovereignty without isolation. Some 87% say access to global technology companies is important for AI adoption, 89% say it is important for scaling quickly, and only 4% would support greater public investment in AI and cloud infrastructure if it raised their own costs.
GEC Newswire put it to Erasmus that applying governance to a technology evolving this quickly resembles jumping onto a moving train. He began with geography. “We know that we should bring infrastructure as close as possible to our customers,” he said.
The second element, he said, is the controls AWS builds into services such as Amazon Bedrock, its managed service for foundation models. “We provide all the controls and build it in a sovereign-by-design way that allows an organisation to control where their data sits and how their data is accessed at any point,” he said.
Those controls do not wall customers off from the models they want to use, he added, and AWS works with each organisation on the processes that sit around them. “The way we architect our services and present them to our customers gives them that control, and then we work with them to make sure the right governance process is in place and is mature enough,” he said.
Asked whether that makes regulation a brake on adoption, he said it did not. “I don’t see it as a blocker. I see it as a key part of the fundamentals that need to be put into place in order to scale really quickly,” he said.
He also pushed back on the idea that governance is being bolted onto a new technology after the fact, pointing to Amazon’s own history with machine learning. “Take Amazon as an example. As early as 1998, we were providing personalisation,” he said.
What changed with generative AI, he said, was that the technology reached the average consumer almost overnight, leaving enterprises to catch up on the controls. “Organisations want to know that they have that operational maturity in place before they scale,” he said.
The practical expression of that argument in the UAE is the Sovereign Launchpad, built by AWS and e& and endorsed by the UAE Cybersecurity Council, which since November 2025 has allowed government agencies and regulated customers to run all workloads except those classified Secret and Top Secret. It sits within a $1 billion strategic partnership between AWS and e&, and alongside AWS’s commitment to invest up to AED 20.1 billion ($5.47 billion) in its UAE region through 2037.
Rising AI bills are pushing buyers back to basics
Around 31% of organisations say insufficient internal financial resources are preventing them from expanding AI, and 39% of hospitality businesses cite an unclear business case or return on investment. With token consumption driving up bills and questions about stalled pilots growing louder, GEC Newswire asked Erasmus how AWS answers buyers who doubt the spending is worth it.
He said the conversation inside companies has moved on from whether to spend. “I’m seeing organisations starting to realise that it is a necessary investment, as part of the experience and the capability that they put out,” he said.
The discipline, he argued, lies in holding two time horizons together. “Balancing what is my immediate return, and a strategic impatience, with the long term is going to be a critically important thing for many organisations,” he said.
Sequencing matters as much as patience, since companies that begin with a pressing business problem and a clear driver for change, and only then choose the technology, tend to get further, and Erasmus compared it to the working backwards process AWS applies to everything it ships. “Anything we’ve launched into the market, we always take through that process to make sure we know, when the end customer picks it up, whether they know what they’re getting in terms of value,” he said.
The example he chose was The ENTERTAINER, the Dubai-based lifestyle and offers platform, which has built a multilingual assistant called H.A.P.I. on Amazon Bedrock. According to the report, it holds conversations in English and Arabic and orchestrates several AI agents, including a planner for personalised itineraries, a search assistant, a support agent, and an escalation assistant that hands users to a human when needed.
Erasmus said banks were applying the same thinking to customer experience and voice, and that the pattern repeats across sectors. “We’re finding lots of different areas where organisations believe they can improve the way a customer interacts with them,” he said.
AWS answers March’s strikes with multi-region design
Resilience carries more weight in the region this year than it did 12 months ago. On 1 March, during Iranian missile and drone attacks on UAE infrastructure, objects struck an AWS facility in its ME-CENTRAL-1 region, causing a fire that took one availability zone offline and significantly affected a second.
Asked how AWS now views security across both physical infrastructure and AI-driven cyber threats, Erasmus began with the company’s footprint. “We’re going to be committed to the region,” he said.
He pointed to how AWS designs its infrastructure to absorb failure. “We continue globally to look at how we evolve our infrastructure deployments: what that looks like from a physical perspective, how we think about our AZ deployments, how we think about multi-region as well, because we have this global consistency,” he said.
That consistency, he explained, is what lets customers move or replicate workloads without rebuilding them. “That allows organisations to deploy in a very consistent way in any region they choose to,” he said.
On cyber risk, he described security as Amazon’s highest priority while stressing that the responsibility is shared with customers. “We always talk about responsible adoption. Whenever we work with our customers, we talk about responsibility: we will bring all the capability, but there’s also responsibility on the customer,” he said.
The report offers a measure of how much distance remains between intent and execution. Once agentic AI was explained to them, 57% of businesses surveyed said they planned to use it or were considering it, a far larger group than the 9% of informed businesses that have deployed it. How quickly that intent turns into production will be decided company by company, in the data strategies, guardrails, skills and mandates Erasmus listed, with Abu Dhabi’s 2027 deadline for an AI-native government already setting the pace for everyone else.
