The AI Shift: Moving Beyond Models Toward Intelligent Agents From government services and enterprise productivity to healthcare, media, and education, the real value will come from AI that is context-aware, locally relevant, and seamlessly embedded into everyday systems.

AI is shaping economies, redefining industries and powering everything from productivity tools to national digital strategies. The AI space is transitioning from an era of competing to build the largest foundational models to one focused on delivering practical, intelligent AI agents that can reason, adapt, and act autonomously. This brings new opportunities for planning and policy, especially for economies like the UAE that have invested heavily in AI infrastructure and national innovation agendas like the UAE's 2031 AI Strategy. Let's explore AI's evolution, from foundational models to intelligent agents and beyond.

The Foundation Model Era

Over the last few years, the centre of AI development has been foundation models, large neural networks trained on massive datasets. There are more than a hundred models in use that are capable of a wide range of tasks: generating content, translating languages, writing code, and even creating music and images.

According to the AI Index Report 2025, model scale continues to grow rapidly, training compute doubles every five months, datasets every eight, and power use annually. The Elo system, used to rank AI models, shows that the skill gap between top AI models is shrinking fast. In just one year, the difference between the best and 10th-best models dropped by more than half, and the top two are now almost identical in skill.

Some say AI development has reached its peak, and smaller, more specialized models are required. Some of the most frequently cited reasons are a lack of available data, computing power, and energy. For example, a single query to ChatGPT consumes about 10 times as much energy as a single Google search.

Foundation Needs Infrastructure

The big established and emerging players behind the foundation models are well-known companies, modern superstars of the AI technology scene. However, their breakthroughs wouldn't be possible without the companies that build and provide the infrastructure for AI development at scale.

Building a foundation model requires a massive investment in infrastructure. According to Epoch AI, the cost of training frontier AI models has grown two to three times per year for the past eight years. Researchers also expect that training the largest models will cost over a billion dollars by 2027.

Training foundation models at scale requires highly advanced and resource-intensive infrastructure. This begins with clusters of high-end GPUs or TPUs, supported by fast networking such as 100 Gbps Infiniband and extensive SSD storage to process and checkpoint massive datasets. Equally critical is a robust data infrastructure that ensures the efficient collection, cleaning, and secure handling of petabyte-scale data volumes. To manage such complexity, distributed machine learning frameworks are employed, alongside orchestration tools for compute job distribution.

Beyond computational systems, maintaining these large-scale operations depends on sophisticated monitoring tools to track system health and performance, reliable backup systems, and substantial physical infrastructure, including uninterrupted power and advanced cooling solutions. The entire process must be overseen by expert teams spanning machine learning engineering and data engineering. These layered infrastructure demands underline the immense cost, complexity, and talent required to develop and deploy state-of-the-art AI models.

The UAE has recognized this and is responding strategically. Through entities like G42, the Advanced Technology Research Council, and Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), the country is building local capabilities in digital infrastructure, chip innovation, and AI research. These initiatives are not solely aimed at fostering innovation, they are fundamental in securing long-term technological self-reliance and ensuring national sovereignty in an increasingly AI-driven global economy.

From Smart Assistants to Smart Agents

With momentum in the AI model race stabilising and infrastructure investments in place, the next leap is toward agentic AI, AI that can reason, plan, and act on behalf of users. AI is already being widely used across the board, whether for entertainment or productivity. From creating AI avatars to AI Companions that handle notetaking during conference calls, AI has become extremely helpful in our daily lives.

Agentic AI is driving this shift toward a more practical, product-driven, and empowering approach, unveiling AI's ability to go beyond just searching, analysing, and generating data to perform real-life actions. It blends context awareness, memory, logic, and goal setting to not only understand but also to act.

This shift is significant for the UAE, where digital services, from government portals to customer care, are becoming more advanced and interconnected. Agentic systems could help residents renew documents, plan travel, manage health routines, and even navigate personalised education. All managed through a single, proactive interface.

Leading the Next AI Frontier

As the AI focus moves from large-scale experimentation to localized, deployable solutions, success in AI will be about integrating smarter AI into real life. With AI to contribute $100 billion to the UAE's GDP by 2030, the nation's forward-thinking leadership, digitally native population and a strong public-private tech ecosystem will enable this growth. From government services and enterprise productivity to healthcare, media, and education, the real value will come from AI that is context-aware, locally relevant, and seamlessly embedded into everyday systems.