Case Study

Assisted coding for financial institutions

Boosting software developer efficiency through AI-powered agents and contextual knowledge.

#SDLC
#Generative AI
#Coding

Picture

Reduce up to 20% effort in coding, improving quality

Reply collaborated with one of the major European financial groups to establish a new factory supported by coding accelerators based on Generative AI technologies.

Various activities were tested and adopted, including code generation, unit test case generation, code documentation, code review/refactoring, and version control. During the experimentation on a Java enterprise technology stack, the new framework helped reducing “boilerplate” coding tasks effort by up to 20%, while in some cases, improving the quality of outputs.

The features of the code assistant

Picture
Conversational interface

A chat allows the developer to interact with the assistant in natural language.

Picture
Multi-agent architecture

A specialised ecosystem of agents collaborates to assist the developer

Picture
Expertise on finance context

Integration with the knowledge base enables strong technological and functional contextualisation

Picture
Plug & Play

A flexible environment that allows integration of customised agents for specific contextual needs

Many specialised agents in a uniform ecosystem

The project covered Code Generation for rapid development by completing and implementing methods/classes as per instructions, Unit Testing for thorough code analysis and test generation to ensure high coverage and enhanced quality, and Review & Refactor for identifying areas of improvement in best practices, maintainability, and testability, proposing enhancements. Moreover, instant inline code documentation was used to improve maintainability.

New challenges in the roadmap

The innovation project is progressing across various domains. Extending the array of technologies supported by the tool. A "Concierge" agent will assist developers in constructing requests to streamline plugin usage. A “Developer Journal" will facilitate the resumption of historical interactions for picking up interrupted tasks. Self Correction will involve automatic execution and validation of unit tests, utilising console output for self-correction of generated code.

Picture

Iriscube Reply is the Reply Group company focused on innovation in the banking and financial fields with a vertical specialisation in digital banking solutions, digital sales platforms, digital onboarding, advanced CRM, digital architectures, mobile technologies, biometrics and advanced multichannel solutions. All managed solutions are based on innovative technologies such as cloud-native architectures to microservices with event communication, AI systems, proximity communication and financial machine learning. Iriscube Reply supports its customers belonging to the banking and financial sector in all phases of the implementation of digital solutions: from the initial study, to the design and definition of the architectures to their development and implementation.