UAE Banks Federation Chief: Agentic AI Is Banking's Next Phase, But Accountability Must Stay Human

Artificial intelligence has already reshaped banking, fraud detection, cybersecurity, customer service, regulatory reporting, risk management and personalized products all run on it today. According to Jamal Saleh, Director General of the UAE Banks Federation (UBF), the next phase will be defined by something more autonomous: agentic AI, which he describes as "capable of reasoning, planning, collaborating and taking action within clearly defined governance frameworks," rather than simply generating outputs for a human to act on.

Saleh laid out this argument in a recent piece for International Banker, describing agentic AI as capable of coordinating entire workflows rather than completing single tasks, arranging specialized AI systems, pulling from multiple data sources, and adapting continuously as circumstances change, all while operating under human supervision. For a bank customer, he described a model where, instead of completing multiple applications, uploading documents separately and speaking to different departments, a customer would interact with a single trusted AI agent that, with their consent, assesses affordability, compares options, retrieves documentation, verifies identity, schedules valuations, coordinates legal paperwork, checks regulatory requirements and prepares a recommendation for human approval.

Saleh frames this as augmentation rather than replacement. "Agentic AI's greatest value lies in augmenting the capabilities of banking professionals," he wrote, arguing that AI should handle pattern recognition and repetitive workflow execution while bankers focus on strategic judgment, relationship management, and the accountability functions automated systems can't own.

Risk management is where Saleh sees the clearest near-term impact. He argues financial institutions currently juggle cyber threats, geopolitical uncertainty, financial crime, climate risk and constantly evolving regulation simultaneously, and that agentic AI systems can monitor transactions, customer behavior, macroeconomic indicators, cybersecurity alerts, sanctions updates and market developments continuously and in parallel, shifting fraud detection and compliance monitoring from reactive to predictive. He's careful to draw a line here: "compliance should not be fully automated," since regulatory decisions require human accountability, and AI's role is to make that process more intelligent and proactive rather than to remove human sign-off entirely.

Saleh cites a $150 billion estimate for the value agentic AI could create across the GCC. That figure warrants context: it traces back to a May 2023 McKinsey and GCC Board Directors Institute report estimating AI's potential value across the GCC broadly, roughly 9 percent of the region's combined GDP, based on a survey conducted before "agentic AI" became the industry's dominant framing. No independently verified agentic-AI-specific estimate at this figure exists; it appears the widely circulated 2023 general-AI number has been carried forward under newer terminology rather than replaced by a current, agent-specific study.

Saleh's broader argument rests on the UAE's existing AI infrastructure and institutional track record. The country appointed the world's first AI minister in 2017, established the first graduate-level AI research university (MBZUAI) in 2019, and in April 2026 the Cabinet announced a framework to transform 50 percent of UAE government sectors and services into agentic AI-driven models within two years, backed by a national upskilling initiative targeting 80,000 employees. A dedicated Artificial Intelligence and Data Authority now consolidates policy, national strategy oversight and digital government compliance standards, part of the broader institutional architecture the UAE has been assembling to govern its AI ecosystem across both public and private sectors, infrastructure that would also support the kind of sovereign, on-premises AI deployment regulated sectors like banking increasingly require for sensitive customer data.

Inside banking specifically, Saleh credits the Central Bank of the UAE with building a regulatory environment that enables innovation "without compromising financial stability, cybersecurity or consumer protection." UBF itself has been running practical exercises toward that end: its fifth annual Cyber Wargaming exercise took place on June 17, 2026, under CBUAE and UAE Cybersecurity Council supervision, drawing more than 350 participants from member banks, Al Etihad Payments, e&, du, Gulf Insurance Group and other financial and technology institutions. Saleh also references a 2025 UBF partnership with the Emirates Institute of Finance's Innovation Hub and KPMG to prototype agentic AI solutions for risk management, compliance, fraud prevention and customer experience with member banks.

Saleh's closing argument centers on governance rather than the technology itself. He maintains that agentic AI in finance must remain "transparent, explainable and continuously monitored," with accountability sitting permanently with people rather than the systems themselves. Whether the UAE's regulatory model, built through CBUAE oversight and industry-wide prototyping exercises, becomes a template other GCC banking sectors follow as they build their own agentic AI strategies remains an open question, one Saleh's own framing raises but doesn't answer.