By Alfred Manasseh, Co-Founder and COO of Shaffra

For much of the past two years, AI has been discussed mainly in terms of pilots, productivity, and experimentation. But in moments of uncertainty, the conversation changes. This is when AI needs to move beyond pilots and into execution. When pressure rises, what matters most is speed, consistency, and coordination. The real question is whether institutions have the operational capacity to respond clearly, maintain continuity, and support decision-making under pressure.

In the UAE, that question carries particular weight because resilience, proactiveness, and digital by design have already been established as national priorities. This is no longer a futuristic idea. It is already being implemented across institutions.

This is why the conversation is moving beyond AI as a surface-level capability and closer to the operating core of institutions. In 2024, UAE federal government entities processed 173.7 million digital transactions and delivered 1,419 digital services, with user satisfaction reaching 91%. Once millions of people are interacting with digital systems, resilience depends not only on keeping platforms online, but on making sure information flows remain clear, response times hold steady, and service quality stays consistent under pressure.

Filtering signal from noise

In high-pressure environments, the first challenge is information overload. Fake information, true information, public questions, updates, and warnings all arrive at once, and institutions have to respond without adding confusion. Human teams remain essential because judgment and accountability must stay with people. But people alone cannot process that volume of information at the speed now required.

This is where Autonomous AI Teams become operationally valuable. AI is effective at dealing with large amounts of data, identifying patterns, and helping institutions filter signal from noise. Used properly, that gives leadership a stronger basis for communicating clearly, responding faster, and addressing confusion before it spreads.

Why governed systems hold up

Good governance is what makes AI dependable in sensitive moments. It is not only about speed. It is about consistency in messaging, consistency in how citizens and residents are served, and making sure people are well-informed. In uncertain situations, the public does not only need information. It needs information that is clear, timely, and trusted. Governed AI helps institutions provide that support without losing control or passing ambiguous situations with false confidence.

This is particularly relevant as research has found that six in 10 UAE employees use AI in their daily jobs, while IBM reported that 65% of MENA CEOs are accelerating generative AI adoption, above the global average of 61%.

The UAE can lead this shift because it is building around digital capacity at every layer, from infrastructure to service delivery to workforce readiness. The Digital Economy Strategy aims to raise the digital economy’s contribution significantly by 2031, while broader trade guidance has also framed the ambition as growing from 12% of non-oil GDP to 20% by 2030.

Working model in practice

This is also where Shaffra offers a practical example of how the model is changing. Through its AI Workforce Platform, Shaffra’s Autonomous AI Teams are already saving more than two million manual work hours per month and reducing operational costs by up to 80%. These systems can monitor inbound activity, classify issues, support fraud reviews, prepare draft responses for approval, and help institutions listen at scale to recurring public concerns.

In Shaffra deployments more broadly, this model has also delivered significant time and cost efficiencies across enterprise operations.

That does not replace leadership or human judgment. AI and humans play different roles, and the real value comes when they work together. It gives institutions stronger operational support, with greater speed, consistency, and control when pressure is highest. In the years ahead, the strongest organisations will be the ones that move beyond AI as a productivity tool and build it as a governed resilience layer that stays reliable when uncertainty tests every process around them.

From content creation to media buying, artificial intelligence is quietly reshaping how campaigns are built, delivered, and optimised across the GCC.

In the UAE and across the GCC, artificial intelligence has moved well beyond the stage of experimentation. What was once a buzzword discussed in boardrooms is now deeply embedded in the day-to-day execution of advertising. Brands are no longer testing AI—they are relying on it to run campaigns, generate content, and make increasingly precise decisions about audience targeting and timing.

On the creative front, the shift is particularly visible. AI-powered tools are now capable of producing ad copy, visuals, and even short-form video content at a pace that would have been unthinkable just a few years ago. For marketers operating in a market like the UAE—where campaigns often need to speak to audiences in both English and Arabic, while also resonating across a diverse mix of nationalities, this level of speed and adaptability is more than a convenience. It is becoming a necessity.

Behind the scenes, machine learning has also transformed how media buying is approached. Traditional methods that relied heavily on instinct or retrospective performance reports are steadily being replaced by systems that analyse audience behaviour in real time. These platforms continuously optimise campaign performance, adjusting budgets and placements based on how users interact with content.

In the UAE’s PR ecosystem, brands are already leveraging platforms such as Meltwater, Brandwatch, and Sprout Social to better understand media performance, audience sentiment, and the broader buying landscape.

AI is also influencing emerging advertising formats. Digital billboards, for instance, are becoming more responsive, using live data inputs to tailor content based on factors such as time of day, location, and audience movement. Similarly, augmented reality experiences are beginning to incorporate behavioural insights, offering more contextual and interactive brand engagements.

Looking ahead, the trajectory appears clear. Advertising is moving towards deeper automation, more intelligent recommendations, and tighter integration between creative tools and analytics platforms. The industry is shifting from a model centred on broadcasting messages to one that focuses on responding to audiences in real time, with context and precision.

In this evolving landscape, AI is no longer just an enabler, it is becoming the foundation on which modern advertising is built.

Banks have long been the GCC’s digital pioneers. In the UAE, Saudi Arabia and Qatar, financial institutions were among the first to embrace mobile banking apps, roll out contactless payments at scale and introduce AI-powered chatbots to handle customer queries in Arabic and English. More often than not, banks set the pace and other sectors followed.

Given this decades-long precedent, you would expect the same pattern to be playing out with artificial intelligence. After all, AI is already embedded in the daily lives of Gulf consumers. Ride-hailing, e-commerce, government, and a plethora of other services across the region have increasingly integrated AI into their systems, to effectively personalise experiences and streamline transactions.

And yet, when we look inside banks themselves, the story is more complicated. According to the latest Riverbed Global Survey, only 40% of organizations in the financial sector consider themselves ready to operationalize AI. Just 12% of AI initiatives are fully deployed enterprise-wide, while 62% remain stuck in pilot or development phases. In a sector known for digital ambition, there is a striking gap between intent and execution.

Stuck in Pilot Purgatory

In most industries, pilots fail because the idea simply does not resonate. Testing reveals a weak product-market fit, limited customer appetite, or unclear commercial value.

That is not what we are seeing in banking AI. Regional banks have successfully piloted AI models that detect fraud in real-time, reduce false positives in anti-money laundering checks, predict liquidity requirements, and power conversational assistants capable of resolving complex service requests. Relationship managers have used AI tools to surface next-best-product recommendations based on behavioral data. And operations teams have leveraged machine learning to optimize payment routing and reduce processing delays.

In controlled environments, these pilots often deliver impressive results. And yet, few ever make it past this stage. The initiative remains confined to a sandbox. Expansion is delayed. Integration becomes “phase two.” Eventually, attention shifts to the next promising experiment. So, if the feature works and the value is clear, what is holding banks back?

AI that Fails to Scale

In my experience working with CIOs across the region, two obstacles repeatedly stand in the way of AI moving from proof of concept to production. The first is operational complexity. Most financial institutions operate in highly fragmented environments. Core banking platforms run alongside decades-old legacy systems, with critical workloads split across on-premise data centers, private clouds, and multiple public cloud providers. Third-party fintech integrations also adds further layers of interdependency.

Deploying AI into this landscape is not as simple as plugging in a model. AI workloads are data-hungry and latency-sensitive. They require reliable pipelines, consistent telemetry, and predictable performance across every layer of the stack. In a hybrid, multi-cloud architecture, even minor configuration mismatches can trigger cascading issues.

The second obstacle is limited visibility. Without a unified view of applications, infrastructure, networks, and user experience, AI-driven services can behave unpredictably. A model may be performing perfectly, but a network bottleneck slows response times. An upstream data source may degrade in quality, subtly skewing outputs, and an infrastructure change in one environment may impact inference speeds elsewhere.

When visibility is fragmented, issues take longer to diagnose and resolve, and Mean Time to resolution increases. Operational risk rises, particularly when customer-facing or revenue-critical services are affected. In a heavily regulated market such as the UAE or Saudi Arabia, that risk has compliance implications as well as reputational ones.

Left unaddressed, this kind of live digital environment leaves very little room for innovation. AI cannot become the transformational force many claim it to be if it is constantly constrained by hidden friction.

Conquering Complexity

Moving AI smoothly from pilot to production requires banks to create as frictionless an operating environment as possible. One of the most effective starting points is unified observability. By consolidating telemetry from applications, infrastructure, networks and end-user devices into a single, real-time view, banks can eliminate blind spots, and decision-makers can gain clarity over performance, dependencies and risk across the entire digital estate.

With this foundation in place, AIOps capabilities can correlate signals, reduce alert noise and automate root cause analysis. Instead of firefighting incidents after customers notice them, IT teams can proactively identify performance degradation and resolve issues before they impact revenue or service continuity.

Standardising on frameworks such as OpenTelemetry can further simplify instrumentation across heterogeneous environments, ensuring consistent data collection and analysis. At the same time, investing in data quality, governance and compliance processes ensures that AI models are trained and operated within regulatory boundaries.

In practical terms, this means rethinking infrastructure as an enabler of AI rather than an afterthought. It may involve accelerating data movement between environments, modernising integration layers or rationalising overlapping monitoring tools. The goal is not perfection, but coherence: a shared, real-time understanding of how systems behave and how AI performs under real-world conditions.

From Optimism to Optimisation

The debate about whether AI belongs in banking is effectively over. Across the Middle East, regulators are publishing AI guidelines, governments are investing heavily in digital transformation, and consumers increasingly expect intelligent, seamless services.

Institutions that continue to treat AI as a series of isolated pilots risk remaining in perpetual experimentation. However, those who address operational complexity head-on will move beyond optimism to optimisation.