From pilots to production: How AI earns trust at national scale
Sergej Loiter, CEO of Search, AI, and AdTech at Yango Group, on why ownership, people, and data matter more than models at Machines Can Think 2026 in Abu Dhabi
27 January, 2026
TT
16
As AI moves from pilots into systems that shape daily life, the real challenge is no longer model performance but trust, ownership, and accountability. Here, Sergej Loiter, CEO of Search, AI, and AdTech at Yango Group, explains why AI only scales at national and city level when people, data, and responsibility are designed in from day one.
What has actually changed in how AI scales in business and daily life?
What has changed isn’t the models, it’s that AI has moved from experimentation into production.
For years, AI supported decision-making. We searched, clicked, and chose. Today, AI systems increasingly act: adjusting campaigns, routing deliveries, optimising operations, and influencing real users in real time. In advertising technology, for example, AI now optimises campaigns live, continuously adjusting creatives, targeting, and spend based on performance signals, rather than analysing results after the fact. Once AI enters production, scaling becomes unavoidable, and so does responsibility.
This is where many initiatives break. Not because the technology fails, but because ownership is unclear. Pilots may succeed technically, but when systems go live without a named internal owner, a trained team, and accountability for outcomes, they quietly decay. AI projects don’t fail at scale because they are weak. They fail because no one truly owns them.
The shift, then, is not intelligence alone, it is agency with accountability. AI becomes transformative not because it thinks better than humans, but because it acts faster within boundaries that humans define and remain responsible for.
What determines whether people adopt AI in everyday use?
As AI becomes more personal, trust stops being optional.
At scale, especially in cities, trust functions like infrastructure. If it is unreliable, everything built on top of it fails, even when the underlying technology is strong. People feel trust immediately when systems behave unpredictably, opaquely, or without recourse.
Trust is not created through slogans or policy documents. It is designed into systems through predictable behaviour, transparency, and clear responsibility when something goes wrong. This is particularly visible with everyday AI assistants: when systems understand local language, cultural norms, and user expectations, adoption happens naturally; when they feel foreign or inconsistent, people disengage quickly. Yasmina, the human-like bilingual AI assistant, is a prime example. Its LLM is trained extensively on Khaleeji content and refined through input from Arabic-speaking experts and even local comedians to ensure the humour, references, and style resonate authentically.
Once AI influences movement, access, or decision-making, people need to know that a human and an institution stand behind it. This is why responsible deployment matters more than speed. It is better to scale slightly slower with trust than faster with abandonment.

How should leaders think about AI moving into city-scale systems?
AI already operates inside cities, shaping logistics, commerce, mobility, and services. The question is no longer how powerful AI becomes, but how responsibly and locally it is designed.
At city scale, AI systems live inside language, culture, institutional workflows, and public expectations. When these factors are ignored, even technically strong systems struggle to gain acceptance and often feel foreign or intrusive.
This is visible in areas like urban logistics and last-mile delivery. Autonomous delivery systems can perform extremely well in controlled environments, but long-term success depends on clear operational ownership: defined safety thresholds, human oversight, and teams empowered to intervene and adapt behaviour over time. Without that, performance degrades regardless of model quality.
As systems begin influencing millions of people, agency inevitably becomes responsibility. Effective public-sector approaches focus on design, dialogue, and accountability — not blanket restriction. Cities that succeed treat AI as a long-term capability, investing in people, data quality, and ownership, knowing models will change but responsibility will not.
Why do so many AI initiatives stall after promising pilots?
The problem is not too many pilots. It is pilots without owners. Many technically successful pilots fail because there is no internal team trained to take responsibility once external partners step away. Without a qualified internal “receiver”, the system has no future. Infrastructure and tools can be outsourced. Ownership of the product cannot.
Ownership means running, maintaining, and improving systems over time and standing behind outcomes long after launch. This is why many AI deployments quietly stop being used 12–24 months later.
Generative AI lowered the barrier to experimentation, but it did not remove the need for deep expertise in production. Chatting with a model is not the same as running a product. Organisations that recognise this early scale successfully. Those that do not often mistake activity for progress.
Looking ahead, what principle should AI leaders keep in mind?
Treat AI as a capability, not a project. Leaders who succeed will stop treating AI as a technology experiment and start treating it as an owned, staffed, data-driven system. Looking back in five years, organisations will be most grateful for the decisions they made around ownership and internal capability — and most regret delaying them.
The capability worth over-investing in now is internal AI literacy and product ownership, so systems can be operated, governed, and adapted over time. The most under-priced risk today is assuming AI runs itself after deployment.
Models will change. Infrastructure will evolve. What lasts is data maturity, human capability, and clear ownership.
In the end, the real difference will not be who had the best models but who built the capability to run AI responsibly at scale.

















