Dubai-based Mashreq Bank is exploring the potential of artificial intelligence (AI) to enhance its digital banking services. Early tests show that AI agents can significantly cut down integration design and troubleshooting times by 60-70%, demonstrating a promising direction for faster, more efficient software development.

Accelerating Digital Banking Solutions

Mashreq is currently conducting pilot programs that leverage AI assistance throughout different phases of the software lifecycle, including solution planning, development, code review, and production troubleshooting. This initiative aims to expedite the software delivery process while ensuring that the systems linking various applications, payment mechanisms, and digital channels remain reliable and resilient.

On October 19, the bank officially announced this ambitious undertaking, introducing AI into an engineering environment that plays a crucial role in maintaining banking reliability. The integration teams are essential in managing the connections that facilitate information exchange and service operations across diverse systems and markets.

Partnerships for Modernization

The implementation of AI technology at Mashreq aims not just to speed up individual tasks but also to explore broader contributions AI can make to software engineering and operational efficiency. As Mohamed Abdel Razek, the Group Head of Technology, Transformation, and Information at Mashreq, pointed out, the bank is committed to its digital-first strategy. “We are investing in technologies that allow our engineering teams to work more swiftly while ensuring the resilience and quality that our customers expect.”

The pilot project utilizes IBM Bob, an AI-driven software engineering platform. Initial results indicate that tasks requiring two to three hours of manual effort are now being completed in a matter of minutes. However, while these time savings are significant, the announcement does not delve into evaluation metrics for error rates or overall software reliability.

Integration and Contextual Awareness

Mashreq’s integration platform supports various digital banking functions, including instant payments and cross-border transactions, processing an impressive average of 650 transactions per second. This operational context is crucial, as changes in one area can impact others. Therefore, understanding these interdependencies becomes vital amid rapid development cycles.

To modernize this infrastructure, Mashreq is collaborating with IBM and Gulf Business Machines, employing cloud-native technologies for application connectivity and data management. The AI pilot complements these modernization efforts, focusing on enabling AI agents to assist engineers working within the existing architecture.

Moreover, Mashreq is using IBM Context Forge to equip AI agents with relevant technical information, helping them make informed engineering decisions. This platform connects agents with integration knowledge, API specifications, and operational contexts. Approximately 4,000 requests monthly from AI agents support design, code reviews, troubleshooting, and operational analysis.

Assurance and Quality in AI Application

The pilot at Mashreq highlights a significant shift in the role of AI in software development, moving beyond simple assistance to becoming an active participant in the delivery process. This evolution of AI necessitates a reevaluation of the controls that govern engineering activities.

James Hunter, IBM’s Field CTO for Application Development and AI Software Engineering, emphasizes the importance of establishing confidence in software changes at an accelerated pace. Organizations must now not only create software faster but also ensure meticulous oversight regarding changes, testing, and compliance with governance standards.

Mashreq is also contemplating on-premises deployment for mission-critical applications that may need local models to comply with data residency regulations. While the bank has reported promising efficiency gains, the challenge remains: can these improvements be replicated across all engineering tasks while ensuring high-quality outcomes and operational reliability?