By PYMNTS | October 12, 2026
Highlights
In an early pilot of IBM Bob, Mashreq cut integration design and troubleshooting time by up to 70%, turning two- to three-hour tasks into minutes.
A secure gateway built on Model Context Protocol (MCP) handles about 4,000 AI-agent requests a month across Mashreq’s banking, payments and enterprise systems.
Fifty-three percent of financial institutions call manual-intensive processes their biggest obstacle to real-time payments modernization, PYMNTS Intelligence found.
Every new payment rail, app feature or partner a bank adds has to be wired into the systems already running it. Most of that wiring is still done by hand, and so is the cleanup when a connection breaks.
Mashreq is handing more of that work to artificial intelligence (AI) agents. In an early pilot of IBM Bob, IBM’s AI software engineering platform, the Dubai-based bank cut the time spent on integration design and troubleshooting by up to 70%, according to a Monday (Oct. 12) announcement from IBM. The press release added that tasks that took two to three hours of manual work now take minutes.
The bank can’t afford to slow down while it makes that shift. Mashreq’s integration platform averages 650 transactions per second across digital channels, payments and enterprise systems. “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,” said Mohamed Abdel Razek, Mashreq’s group head of technology, transformation and information.
AI Agents Take on Mashreq’s Integration Work
Mashreq is using IBM Bob across the engineering cycle, from solution design and planning to development, code review and production troubleshooting. To do that work, the agents need to know how the bank’s systems fit together. Mashreq uses IBM Context Forge, a tool built on Model Context Protocol (MCP), to give agents secure access to integration knowledge, application programming interface (API) specifications, architecture standards and operational context. The bank handles about 4,000 requests a month from AI agents to connect to its digital banking, payments and enterprise systems.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
Santosh Vaidya, Mashreq’s head of payments and digital services, said the bank wants an engineering setup “where AI agents can work alongside our teams.”
The AI pilot builds on earlier upgrades. In 2024, Mashreq connected its corporate online banking channels to Aani, the UAE central bank’s instant payments platform, to offer round-the-clock transfers with no cut-off time, PYMNTS reported.
Manual Work Slows Bank Modernization
Mashreq’s bottleneck is common across the industry. Fifty-three percent of financial institutions named manual-intensive internal processes as their biggest obstacle to real-time payments modernization, according to the PYMNTS Intelligence report “From Adoption to Execution: How FIs Are Turning Real-Time Payments Into Competitive Advantage,” produced with The Clearing House. Another 52% cited outdated technology, and 47% pointed to siloed payment systems.
Data is a related hurdle. Three in 10 financial services leaders at U.S. companies with at least $1 billion in revenue named data quality and fragmentation as their main constraint on AI, according to the PYMNTS Intelligence report “Financial Services Pulls Ahead in the Enterprise AI Race.”
AI agents give banks another way to work with old systems. Agents can reach a bank’s core through the same APIs its apps use, or through model context protocol (MCP) connections that expose approved functions as tools, PYMNTS reported. Coding agents may also automate parts of mapping, coding, testing and documentation. That could change the economics of projects run by systems integrators such as Accenture and Capgemini.
The pilot is part of a broader rebuild of Mashreq’s integration platform with IBM and Gulf Business Machines. The platform runs on IBM’s integration software, Red Hat OpenShift and Confluent’s data streaming technology. Mashreq is also exploring an on-premises deployment of IBM Bob for critical work that requires locally hosted models to meet data residency and regulatory requirements.
For all PYMNTS AI and digital transformation coverage, subscribe to the daily AI and Digital Transformation Newsletters.