When Emirates NBD announced in early 2024 that its AI-powered customer service systems had reduced response times by 60% while handling twice the volume of enquiries, the news rippled through boardrooms across the Gulf. Here was concrete evidence that artificial intelligence investments could deliver measurable returns—not in some distant future but in present-day operations. Yet for every success story circulating in the region, dozens of enterprises struggle to demonstrate comparable value from their AI initiatives, caught in pilot programmes that never scale, implementations that underperform expectations, or measurement frameworks that fail to capture AI’s true contribution. The gap between AI’s promise and its documented returns has become the central challenge for executives seeking to justify continued investment.

The return on investment question has grown more urgent as AI spending in the Middle East accelerates. According to IDC’s analysis, Middle East and Africa AI spending reached $3.1 billion in 2023 and is projected to exceed $7 billion by 2027, growing at compound annual rates exceeding 30%. Saudi Arabia alone has committed $40 billion to AI as part of Vision 2030 initiatives, while the UAE’s national AI strategy targets the technology contributing 13.6% of GDP by 2030. These investments reflect genuine conviction about AI’s transformative potential—but conviction does not guarantee returns. A McKinsey global survey found that while 79% of respondents reported AI adoption in at least one business function, only 27% could attribute at least 5% of their EBIT to AI applications. The challenge of demonstrating value is not unique to the MENA region, but the scale of regional investment makes it particularly consequential.

Understanding what MENA businesses are actually seeing from AI investments requires moving beyond vendor marketing claims and conference anecdotes to examine documented outcomes across industries and application types. The evidence that has emerged reveals patterns that can guide investment decisions and implementation approaches. Certain applications deliver reliable returns while others remain speculative despite theoretical promise. Implementation factors—data quality, organisational readiness, change management sophistication—often matter more than the underlying technology in determining outcomes. And the timeline to value varies dramatically across use cases, with some applications generating returns within months while others require years of development before delivering meaningful benefits. Navigating this landscape effectively requires frameworks that connect investment decisions to realistic outcome expectations.

Documented Returns Across MENA Industries

The banking and financial services sector has produced the most extensive documentation of AI returns in the MENA region, reflecting both the sector’s data richness and its relatively advanced digitalisation. Beyond Emirates NBD’s customer service gains, First Abu Dhabi Bank has reported that AI-driven credit decisioning reduced loan processing time from days to minutes while improving approval accuracy and reducing default rates. The bank’s AI systems analyse hundreds of data points beyond traditional credit scores, identifying creditworthy customers who might have been declined under conventional approaches while flagging risks that human analysts might miss. FAB’s disclosures indicate that AI-augmented lending decisions contributed to a 15% reduction in non-performing loans over a three-year period—a return that significantly exceeds implementation costs. Saudi Arabia’s Al Rajhi Bank has implemented AI for fraud detection, reporting a 40% improvement in fraud identification rates while reducing false positives that had previously required expensive manual review. These returns compound over time as systems learn from new data, improving performance without proportional increases in investment.

Retail and e-commerce applications have demonstrated strong returns, particularly in personalisation and demand forecasting. Majid Al Futtaim, the Dubai-based retail conglomerate operating Carrefour stores across the region, has deployed AI systems for inventory optimisation that reduced stockouts by 30% while decreasing overstock positions—a dual improvement that directly impacts profitability. The company’s reported outcomes include multi-million dirham annual savings from reduced waste and improved working capital efficiency. Noon, the regional e-commerce platform, has invested heavily in AI-driven personalisation, with internal metrics reportedly showing double-digit improvements in conversion rates for customers receiving AI-tailored recommendations compared to baseline experiences. The fashion retailer Splash, part of the Landmark Group, implemented AI for trend forecasting and assortment planning, reducing markdown rates by identifying which products would sell through at full price and which required earlier intervention. These retail applications share a common characteristic: they address problems where even modest improvements in prediction accuracy translate into substantial financial impact given the volumes involved.

Industrial and energy sector applications have delivered returns through operational optimisation, though the implementation timelines tend to be longer given the complexity of integrating AI with physical operations. Saudi Aramco’s deployment of AI for predictive maintenance across its production facilities represents one of the region’s most sophisticated industrial AI implementations. The company’s published research indicates that AI-driven predictive maintenance reduced unplanned downtime by 20% at pilot facilities, with each hour of prevented downtime avoiding costs measured in millions of dollars given production volumes. ADNOC, Abu Dhabi’s national oil company, has reported similar outcomes from AI applications in drilling optimisation, where machine learning systems analyse geological and operational data to recommend drilling parameters that reduce costs while improving success rates. The documented improvements include drilling time reductions of 15-20% on optimised wells—savings that multiply across hundreds of wells annually. These industrial applications illustrate AI’s potential in asset-intensive businesses where small efficiency improvements generate substantial absolute returns.

Measurement Frameworks and Attribution Challenges

The challenge of measuring AI ROI extends beyond calculating returns to the more fundamental problem of attribution: determining which outcomes should be credited to AI systems versus other factors that influence business performance. This attribution challenge explains why many organisations struggle to demonstrate AI value even when their AI investments have genuinely contributed to improved outcomes. Business results emerge from complex interactions among market conditions, operational decisions, personnel capabilities, and technology investments; isolating AI’s specific contribution requires analytical sophistication that many organisations lack. Consider a retailer that implements AI-driven demand forecasting and subsequently sees improved inventory performance. The improvement might result from the AI system’s predictions, from concurrent changes in supplier relationships, from seasonal factors, or from heightened management attention accompanying the new system. Without rigorous measurement approaches, the retailer cannot confidently attribute outcomes to AI—leaving executives uncertain whether to expand investment.

Leading organisations have developed measurement frameworks that address attribution through controlled comparison approaches. A/B testing—comparing outcomes for customers or operations exposed to AI systems against matched control groups—provides the most rigorous attribution evidence, though it requires sufficient scale and willingness to withhold potentially beneficial AI capabilities from control populations. McKinsey research indicates that organisations using randomised controlled testing for AI evaluation report significantly higher confidence in their ROI assessments—and correspondingly stronger support for expanded investment. Before-and-after comparisons offer a more accessible alternative, though they require careful adjustment for confounding factors. Time-series analysis that accounts for trends, seasonality, and concurrent changes can isolate AI impact with reasonable confidence when historical data is sufficiently rich. Some organisations have adopted synthetic control methods, constructing counterfactual baselines from comparable business units or time periods that did not receive AI treatments. Whatever the specific methodology, the key principle is designing measurement approaches prospectively—before AI deployment—rather than attempting retrospective attribution that inevitably confronts data limitations.

The intangible benefits of AI investment complicate ROI assessment further. Improvements in customer experience, employee satisfaction, strategic positioning, and organisational capability may generate substantial long-term value while resisting quantification. A bank that implements AI-assisted financial planning might see modest immediate revenue impact but substantial improvement in customer relationship depth that increases lifetime value over years. An industrial company that develops AI capabilities for production optimisation builds organisational knowledge that positions it for future applications beyond the initial use case. These strategic benefits are real but difficult to monetise in ROI calculations, creating tension between analytical rigour that demands quantification and strategic wisdom that recognises broader value creation. Sophisticated measurement frameworks address this tension by distinguishing between different categories of benefit, applying quantitative methods to measurable outcomes while documenting qualitative benefits that inform overall investment assessment. The resulting picture—combining financial returns with strategic value assessment—provides a more complete foundation for investment decisions than narrow financial metrics alone.

Factors Driving Return Variation

The dramatic variation in AI returns across organisations reflects factors that investment decisions should account for but often ignore. Data quality emerges consistently as the strongest predictor of implementation success. AI systems trained on incomplete, inconsistent, or biased data produce unreliable outputs regardless of algorithmic sophistication. A Gartner survey found that poor data quality costs organisations an average of $12.9 million annually through flawed decision-making and operational inefficiency—costs that AI amplifies when systems encode data problems into automated processes at scale. MENA organisations face particular data quality challenges stemming from relatively recent digitalisation, legacy system fragmentation, and organisational silos that impede data integration. The enterprises achieving strong AI returns have typically invested substantially in data infrastructure before or alongside AI implementation, recognising that data quality is a precondition for AI value rather than a problem that AI systems can transcend.

Organisational readiness encompasses the human and process dimensions that determine whether AI systems deliver their potential value. The most sophisticated AI system generates no return if users distrust its outputs, lack training to interpret its recommendations, or operate in processes that cannot accommodate AI-driven insights. Change management proves particularly important for AI implementations because the systems often challenge established ways of working and expertise that employees have developed over careers. A customer service AI that recommends responses may threaten agents who pride themselves on relationship skills; a demand forecasting system may challenge planners whose judgment it supersedes. Organisations that underinvest in change management find that technically successful AI implementations fail to deliver expected value because adoption remains limited or superficial. The Prosci research on change management indicates that projects with excellent change management are six times more likely to meet objectives than those with poor change management—a multiplier that applies to AI initiatives as much as any other transformation effort.

The choice of use cases fundamentally shapes return potential. Not all AI applications are created equal: some address problems where AI offers substantial advantages over existing approaches, while others deploy AI where simpler methods would suffice or where fundamental constraints limit achievable improvement. The highest-return applications typically share several characteristics: they address high-volume decisions where even modest accuracy improvements generate substantial cumulative value; they involve pattern recognition in data too complex for human analysis; they operate in domains where rapid iteration allows systems to learn and improve quickly; and they augment rather than replace human judgment in ways that combine AI and human strengths. Applications that attempt to automate low-frequency, high-stakes decisions often struggle to demonstrate value because the volume of decisions limits learning opportunities while the stakes demand human oversight that constrains AI autonomy. Strategic use case selection—informed by realistic assessment of AI’s advantages in specific contexts—distinguishes organisations achieving strong returns from those pursuing AI for its own sake.