Mashreq is advancing its adoption of agentic artificial intelligence through a pilot of IBM Bob, IBM’s AI-powered software engineering platform, aimed at improving the development and maintenance of the complex integration systems supporting its digital banking operations. The initiative has already demonstrated reductions of up to 60% to 70% in the time required for selected integration design and troubleshooting activities, with tasks previously taking several hours completed within minutes. For a bank whose integration infrastructure processes an average of 650 transactions per second across digital channels, payments and enterprise applications, the potential benefits extend beyond developer productivity. The pilot reflects a growing shift in how financial institutions are applying AI, moving beyond customer-facing applications and employee assistants towards intelligent engineering environments that can support software development, system integration and the operational infrastructure underpinning modern banking services.
Bringing Agentic AI Into Banking Integration Engineering
As banks expand their digital services, the technology infrastructure connecting applications, payment systems, customer channels and external partners has become increasingly important to their operations. Integration platforms now support everything from instant payments and cross-border transactions to API-based partnerships and real-time information exchange, making their reliability and adaptability essential to the delivery of banking services.
Mashreq has invested in a modern integration environment designed to support these requirements across multiple markets and business functions. The platform processes an average of 650 transactions per second, connecting digital banking channels, payment infrastructure and enterprise systems. Maintaining and expanding such an environment requires engineering teams to manage complex application dependencies, integration designs, technical specifications and operational issues while continuing to deliver new functionality.
The bank is now piloting IBM Bob to examine how agentic AI can assist engineers throughout the software development lifecycle. The platform is being evaluated for activities including solution planning, integration design, software development, code review and production troubleshooting, with the objective of reducing repetitive technical work and improving the speed at which engineering teams can deliver and maintain services.
Unlike conventional coding assistants that primarily generate or explain software code, agentic engineering platforms are intended to support broader workflows involving multiple stages of analysis, development and problem resolution. In a banking environment, this could include interpreting existing architecture, identifying integration requirements, reviewing code against established standards and assisting engineers in investigating technical issues.
Early Pilot Results Show Significant Time Savings
According to IBM and Mashreq, early testing has demonstrated reductions of up to 60% to 70% in the time spent on selected integration design and troubleshooting activities. Tasks that previously required two to three hours of manual work were completed within minutes using AI-assisted processes.
These results suggest that AI agents could help reduce some of the time-consuming activities involved in managing enterprise integration environments, particularly where engineers must work across multiple applications, technical documents and system configurations.
However, the reported improvements relate to the pilot’s selected activities rather than an independently established reduction in overall software delivery time. The broader impact will depend on how the technology performs across different engineering workloads, how effectively AI-generated outputs are validated and whether the productivity improvements can be sustained in production environments.
For Mashreq, the objective is to enable engineering teams to spend less time on routine analysis and troubleshooting while increasing their capacity to support new digital banking services.
Mohamed Abdel Razek, Group Head of Technology, Transformation and Information at Mashreq, said the bank is investing in technologies that allow engineering teams to work more efficiently while maintaining the resilience and quality expected by customers. The institution is also evaluating whether similar AI capabilities could eventually support software engineering and technology operations across the wider organisation.
Connecting AI Agents With Enterprise Engineering Knowledge
A significant component of Mashreq’s approach involves IBM Context Forge, which uses the Model Context Protocol (MCP) to support interactions between AI agents and enterprise systems.
One of the challenges in applying AI to complex banking software environments is providing models with the technical context needed to generate useful and reliable outputs. Banking integration systems depend on detailed API specifications, architecture standards, operational procedures and established engineering practices that are not necessarily available to general-purpose AI models.
IBM Context Forge is intended to address this limitation by giving AI agents controlled access to relevant enterprise information, allowing them to operate with greater awareness of the systems and requirements they are supporting.
Mashreq is using the technology to establish a secure gateway through which AI agents can access integration knowledge, API specifications, architecture documentation and operational context. The bank currently manages approximately 4,000 requests per month through this gateway from AI agents interacting with connected digital banking, payments and enterprise systems.
This architecture is intended to support AI-assisted design decisions, code reviews, troubleshooting and operational analysis while maintaining controls over how agents access enterprise resources.
The distinction is important because deploying agentic AI in banking involves more than introducing intelligent software tools. Banks must also establish which information agents can access, what actions they are permitted to perform and how their activities can be monitored within existing security and governance frameworks.
Santosh Vaidya, Head of Payments and Digital Services at Mashreq, described the bank’s objective as developing an engineering environment in which AI agents work alongside employees, using the enterprise knowledge and technical context that engineering teams rely on to make decisions.
Building on a Cloud-Native Integration Foundation
The AI pilot forms part of a broader technology modernisation programme involving Mashreq, IBM and Gulf Business Machines (GBM).
The bank is developing its integration infrastructure using a combination of IBM Cloud Pak for Integration, IBM MQ, IBM App Connect, IBM API Connect, Red Hat OpenShift and Confluent. These technologies support application connectivity, messaging, API management, containerised deployment and real-time data movement across the banking environment.
The modernisation programme is intended to provide a scalable and resilient foundation for digital banking services while supporting enterprise governance and future AI capabilities.
For financial institutions, the relationship between integration modernisation and AI adoption is becoming increasingly relevant. AI agents can assist with engineering tasks, but their effectiveness depends heavily on the accessibility, consistency and reliability of the underlying systems and technical information.
A fragmented integration environment can make it difficult for AI tools to understand application dependencies or generate recommendations that align with established architecture standards. Conversely, a well-structured integration platform can provide a more consistent foundation for introducing AI-assisted engineering workflows.
Mashreq’s approach therefore combines the development of its underlying integration infrastructure with the introduction of AI capabilities intended to improve how that infrastructure is designed, operated and maintained.
Exploring On-Premises AI for Regulated Banking Environments
Mashreq is also evaluating the possibility of deploying IBM Bob on-premises for mission-critical engineering activities that may require locally hosted AI models.
This reflects an important consideration for banks introducing AI into sensitive technology environments. Engineering systems may contain proprietary application logic, infrastructure configurations, security information and operational data that institutions need to manage within specific regulatory or organisational boundaries.
Locally hosted AI models could provide greater control over where information is processed and how AI capabilities interact with internal systems. However, deployment location alone does not guarantee security or regulatory compliance, and banks must still establish appropriate access controls, governance arrangements and oversight mechanisms.
The proposed on-premises deployment remains under consideration rather than being part of the confirmed production environment described in the announcement.
For banking technology providers, the ability to support different AI deployment models may become increasingly important as financial institutions assess how to introduce agentic capabilities into critical technology functions without compromising operational controls.
AI Moves Beyond Banking Applications Into Software Delivery
Mashreq’s pilot illustrates a broader development in the banking technology industry. While much of the attention surrounding generative AI has focused on customer interactions, fraud detection, employee productivity and automated decision-making, software engineering is emerging as another area where financial institutions can explore measurable operational improvements.
Banking technology environments are particularly complex because changes to one application can affect multiple connected systems. Introducing new payment capabilities, modifying APIs or updating digital banking services often requires extensive coordination between engineering, testing, security and operations teams.
Agentic AI has the potential to support these activities by helping engineers analyse technical requirements, understand existing integrations and identify possible issues earlier in the development process. The opportunity is not necessarily to replace software engineers, but to reduce the amount of manual investigation and repetitive technical work required to deliver reliable banking systems.
For technology providers such as IBM, this also represents an expansion of the enterprise AI market. Rather than supplying AI capabilities solely for business applications, vendors are increasingly positioning intelligent agents as part of the software development and technology operations infrastructure used by large organisations.
The longer-term value will depend on whether these tools can deliver consistent improvements without introducing new risks into critical banking environments. AI-generated code, architectural recommendations and troubleshooting outputs still require appropriate testing, validation and human accountability, particularly where failures could affect payments, customer access or other essential services.
Mashreq’s pilot provides an early example of how banks are beginning to evaluate agentic AI within these engineering processes, combining AI-assisted development with enterprise integration infrastructure and controlled access to technical knowledge.
What this means for the industry
- Agentic AI is expanding into banking technology operations. Financial institutions are beginning to evaluate AI agents for integration design, code reviews and troubleshooting, extending AI adoption beyond customer-facing applications.
- Integration engineering could become a significant productivity opportunity. AI-assisted workflows may reduce time spent analysing technical specifications, investigating issues and preparing integration designs.
- Enterprise context is becoming essential to effective AI deployment. Technologies such as MCP-based gateways can help AI agents access relevant institutional knowledge and technical documentation within defined controls.
- Modern integration infrastructure can support broader AI adoption. Banks with well-structured APIs, application connectivity and operational information may find it easier to introduce AI-assisted engineering capabilities.
- Governance will determine how far agentic engineering can progress. Banks will need clear controls over agent permissions, generated code, technical recommendations and human approval before expanding automation into critical systems.
- Banking software vendors are finding new opportunities in AI-powered engineering. Platforms that combine development assistance with enterprise integration knowledge and operational context could become increasingly important to financial institutions modernising their technology environments.
