How AWS solutions facilitate the transition toward Agentic AI in the UAE – Special interview with AWS
As the UAE seeks global leadership through its National Strategy for Artificial Intelligence 2031 and accelerates its shift toward ‘Agentic AI’ in government sectors, AWS emerges as a key player delivering the solutions needed to realize these ambitions. MENA Tech recently sat down with Mats Carrgard, Solutions Architecture Leader, UAE at Amazon Web Services (AWS). In this interview, he shared how AWS solutions help companies move from testing AI to fully executing it, and explored how strategic partnerships are equipping local talent to build the future.
The UAE is actively leading the region with its National Strategy for Artificial Intelligence 2031, aiming to become a global hub for AI. How do solutions like Amazon Quick and Kiro align with this vision to transition local businesses from AI experimenters to true AI-driven organizations?
The UAE’s National Strategy for AI 2031 – launched in 2017 as the world’s first national AI strategy of its kind – set out to make the country a global leader in AI. What’s changed since is that the market has moved past the pilot stage. According to the Amazon Web Services (AWS) and IDC report “Unlocking the Full Potential of AI in the Middle East,” 98% of organizations in KSA and the UAE believe AI will profoundly transform their businesses, and 88% of those that have already invested report improved business performance. With 28% currently investing and another 50% planning to, the region is clearly transitioning from experimentation to execution.
That’s exactly the shift Amazon Quick and Kiro are built for. The UAE government has set an equally ambitious pace of its own: Sheikh Mohammed bin Rashid Al Maktoum, Vice-President and Prime Minister of the UAE and Ruler of Dubai, announced in April 2026 that 50% of UAE government sectors, services, and operations will run on agentic AI within two years. Organizations need tools that match this pace. Amazon Quick is an agentic workspace – a set of “agentic teammates” that answer questions grounded in your business data and take action through research, business insights, and automation. It gives anyone, technical or non-technical, the ability to put AI to work securely using their own documents and information within their own environment. Kiro is AWS’s agentic IDE (Integrated Development Environment) for engineering teams – it turns a plain-language prompt into structured specifications (requirements, design, and sequenced tasks) before writing code, helping developers build production software with AI as a disciplined collaborator.
Together, they democratize AI across the organization. Unlike general-purpose tools, which don’t operate on your own data by default, Amazon Quick and Kiro provide a wider variety of secure interfaces into the AI agents that organizations are building. One helps business teams operationalize AI in daily work; the other helps engineering teams build with speed and rigour.
AWS is deeply committed to being a long-term partner in the UAE’s high-tech acceleration. A key example is the $1 billion alliance between AWS and e&, which includes the AI Nation – Afaaq program – a nationwide AI upskilling initiative designed to train 30,000 individuals in cloud and AI technologies, directly addressing the AI and digital skills gap in the UAE.
When Amazon Quick helps teams draft localized content, how does it ensure the context remains regionally relevant and culturally aligned?
The AI models used in Amazon Quick and Kiro are, in many cases, the same models used across the industry. The model itself is, to some extent, a black box. So the way you control outputs – ensuring they are grounded, relevant, and correct – comes through two key interfaces: first, it’s about making sure the inputs to the model are correct. Both Amazon Quick and Kiro allow customers to use their own information – their own documents, emails, Teams messages – to ensure inputs are grounded and accurate. Instead of giving a vague outline, you provide the document itself, so the input is accurate by construction. Second, after you have the output, it’s about having checks and balances in place. AWS provides tools for evaluation of agents and observability. As an organization, you can control inputs, make sure they’re stored securely, and check outputs to ensure they’re correct and according to your guidelines. We have guardrails, evaluations, and observability tools to manage that.
Additionally, Amazon Quick learns gradually. It adapts through a knowledge graph based on how you interact with the tool, and becomes more and more helpful and personalized over time.
What kind of tangible productivity gains and performance boosts are you projecting for teams adopting these agents?
Just to be precise – it’s not a specific branded feature called “Morning Briefing,” but Amazon Quick’s desktop app does proactively surface exactly that kind of prep before a meeting. Everyone – individuals and professionals – processes a lot of documents. You interact with your emails, internal messaging whether it’s Teams or Slack, and many other tools. What Amazon Quick does is bring all those different pieces of information and different information screens together under one roof. It allows organizations and individuals to build their own knowledge bases, graphs, and workflows that let them interact with information they otherwise would have had to process manually.
As a specific example: instead of opening your email inbox and reading 100 emails, then reading all your messages, then going to your calendar and looking at your next meeting – you can integrate Amazon Quick with those workflows and simply ask, “Can you help me prepare for this meeting?” Amazon Quick will extract the relevant emails and messages and provide a briefing you can use to prepare. Instead of spending an hour or even two reading all documents and preparing a brief yourself, you can now do that in minutes.
It also allows you to do more within budget – cost efficiency becomes a big part of it, where you can use AI agents as an extension to the people in your organization. And it allows you to be more robust and maintain quality standards. AI agents can review your work – whether that’s for technical security, architecture, grammar, or errors – and serve as a safety net for whatever you’re working on.
We are entering an era of independence for AI agents. What is your advice to UAE technology leaders on how to prepare their workforce and IT architectures for this shift?
Automation should only come after you are comfortable with the tools you’re using. It’s not easy to just let AI agents do everything – give them a brief, close your eyes, and wait for results. Just like when you work with humans, you need to make sure the individual steps in the workflow are done correctly.
When your organization is mature enough – meaning people have adopted AI tools, they understand them, and they can validate outputs to ensure they’re not biased or full of hallucinations – then you can gradually shift towards automation. This means taking two manual AI-driven workflows that you used to run manually and putting them together into an automation flow that you give to AI agents to execute more autonomously. Underneath, it connects across systems through 50+ built-in connectors and, critically, open standards – Model Context Protocol and OpenAPI – which extend reach to over a thousand applications. These are just a few examples of the many organizations and customers accelerating innovation and productivity with Amazon Quick:
- 3M sales reps save more than five hours a week.
- Amazon Books cut the time leaders spent on coordination documents by about 80%.
- Propulse Lab reports roughly an 80% reduction in ticket-handling time.
- Amazon’s Last Mile Delivery team compressed a two-week research project into about 30 minutes using the research agent.
Automation and driving more towards agentic AI needs to come with maturity and adoption, to ensure you don’t get outputs that are out of control – high volume with no way to actually verify correctness.
