AI-powered ESB Transformation
Reply has developed a methodology to support a leading Italian financial institution in converting existing enterprise integration setups built with legacy ESB technologies into cloud architectures.
#Integration
#Modernisation
#Generative AI
,allowExpansion)
STREAMLINING INTEGRATION MODERNISATION WITH GEN AI
Reply has developed an approach that uses AI to modernise integration layers by collecting data and managing feedback, continuously improving performance.
Cost reduction & insourcing
Migrating from a product-based solution to a Java-based open technology reduces licensing and operational costs, avoids vendor lock-in, and promotes the insourcing of resources, which alleviate the need for expensive and scarce product experts.
Move to cloud
The adoption of new open technologies, which are more cloud-friendly and enable modular deployment through containers, facilitates smoother integration orchestration and horizontal scaling of frequently used services without the need to replicate all other services.
Less legacy, more agility
Refactoring and migrating from a monolithic technology to a microservices architecture enabled the financial institution to completely reconsider their integrations, eliminate constraints, and reduce infrastructure overhead.
)
Adopting LLMs without overhead for developers
Reply and the financial institution confronted the challenge of lacking ready-to-use datasets due to the specificity of the use case, involving proprietary XML source code generated by traditional ESB low-code tools. Starting from scratch, they needed to gather data concurrently from ongoing projects without burdening the development teams. Reply devised a process and built a supporting tool for automated dataset lifecycle management and LLM model training. The process involved a blend of deterministic automation (scripts) and LLM, encompassing guided dataset collection with developer-friendly versioning, data preparation for cleaning and segmentation, and multi-platform/multi-model/multi-approach LLM training support.
Reply experts and Human-in-the-Loop
The financial institution's integration flows slated for migration were in closed, proprietary formats. Reply’s integration experts offered their knowledge and experience in integration tools, becoming specialised trainers for the model. Specialist feedback after tool-generated content review and refinement was crucial for continuous model learning.
Upon release into production, DevOps pipelines loaded the final translated component into the tool's knowledge base, which served as additional data for subsequent training, refining the AI's performance over time. The tool learned to implement flows according to the development team's coding style, ensuring seamless integration with previously implemented material.
New opportunities to extend the solution’s adoption
The project involved migrating legacy integration artifacts utilising technologies such as Webmethods and IBM Business Automation Workflow to an Apache Camel stack. Due to the results achieved, the financial institution is now working with Reply experts to extend the ESB Transformation project to modernise other integrations, such as those based on TIBCO and Oracle stacks.
Blue Reply specialises in consulting, design and implementation of innovative differentiation solutions based on open source technologies and market-leading vendors such as IBM, Red Hat, Automation Anywhere, Axway, CoreMedia, HCL, Pega, Quadient, Software AG and VTEX. With a consolidated expertise in Low Code, Big Data & Analytics, Microservices & Devops, Artificial Intelligence, Business Consulting, Commerce data Platform, Industry 4.0 & IOT, Composable Commerce, Blue Reply assists and supports its clients, mainly belonging to the Banking, Credit, Insurance, Manufacturing, Telco & Media and Retail sectors, in their digital transformation process.
,allowExpansion;Resize,width=660)
;Resize,width=660)
,allowExpansion;Resize,width=660)
;Resize,width=660)
,allowExpansion;Resize,width=660)