Mashreq is piloting IBM Bob to help engineers develop and maintain the software connections supporting its banking services. The trial applies agentic AI to integration engineering, with IBM reporting early reductions of up to 60–70% in time spent on design and troubleshooting.
Mashreq’s integration platform handles an average of 650 transactions per second. The bank is also working with IBM and GBM on modernization and using IBM Context Forge to provide agents with enterprise context. A gateway handles about 4,000 agent requests monthly, a separate measure from transaction throughput.
Muralidharan Rajagopalan, interim general manager of GBM Dubai, said: "As UAE banks accelerate their digital transformation and adoption of AI, combining global technology expertise with deep local understanding is increasingly important. At GBM, we bring that perspective to help organizations translate emerging technologies into practical outcomes. We’re proud to work with Mashreq and IBM to explore how agentic AI can help engineering teams move faster, simplify complex integration work and create more capacity for innovation.”
Mohamed Abdel Razek, Mashreq’s group head of technology, transformation and information, said: “As we continue to advance our digital-first strategy, we are investing in technologies that help our engineering teams move faster while maintaining the resilience and quality our customers expect. Through our pilot with IBM Bob, we're seeing how agentic AI can help simplify complex integration engineering, accelerate delivery and improve developer productivity. As we continue our AI journey, we're also evaluating how these capabilities can transform software engineering and operations across the enterprise.”
IBM announced Bob’s global availability in April, describing a platform that works across planning, programming, testing, deployment and application modernization.
The launch documentation describes specialized agents, reusable workflows and configurable approval checkpoints. IBM says developers can choose how actions are authorized, ranging from manual review to automatic approval for particular task types.
Bob also supports agents that handle separate tasks in parallel, natural-language instructions for generating code, and a command-line tool called Bob Shell. These capabilities extend its intended role beyond writing code to supporting other tasks within a codebase.
Shukri Eid, IBM’s vice president and general manager for the Gulf, Levant and Pakistan, said: "The next chapter of AI will require empowering developers and engineers with intelligent agents that can collaborate across complex software engineering workflows. Mashreq is taking a strategic approach by combining a modern, cloud-native integration platform with agentic AI and IBM Bob to transform how software is designed, developed and maintained. Together, we're helping build an intelligent digital foundation that accelerates innovation, improves operational efficiency and prepares the bank for the future of AI-driven banking."
IBM offers analytics intended to track agents’ contributions to software delivery. Such measurement could help teams assess whether faster completion of individual tasks translates into better delivery across a project.
The supporting gateway addresses another part of the engineering workflow: access to tools and information. IBM’s ContextForge project documentation describes an open-source gateway, registry and proxy that brings different services behind a common interface, with centralized discovery and management. Its supported interfaces include the Model Context Protocol, commonly known as MCP.
MCP provides a way for AI applications to connect with tools and contextual resources. ContextForge adds a management layer around those connections. In practical terms, an organization can make approved services discoverable to agents through a shared gateway rather than maintain a separate connection process for each service.
The distinction between a development assistant and a gateway is useful when assessing the project. Bob is designed to perform software engineering work; ContextForge manages access to connected resources.
Santosh Vaidya, head of payments and digital services at Mashreq, said: "Our vision is to create an intelligent engineering environment where AI agents can work alongside our teams to support design, code reviews and troubleshooting. IBM Context Forge provides the foundation to securely explore this using the enterprise knowledge our engineers rely on every day."
The project’s documentation describes authentication, access controls and observability among its features. These capabilities give engineering teams mechanisms for controlling and monitoring agent interactions.
IBM’s self-hosted deployment announcement describes options for running Bob in customer-managed environments. Published September 30, it explains that source code, application context and build artifacts can remain inside that environment when a supported self-hosted model configuration is used.
The company also describes hybrid arrangements that connect Bob to supported external model services, alongside on-premises and air-gapped options. Where information is processed therefore depends on the selected configuration.
IBM presents financial services as one potential use case, with restricted banking applications using local model inference while less sensitive work connects to approved external services.
