Generative AI

AWS Blu Insights was designed and released with the goal of lowering the barriers to modernization, automating repetitive tasks, establishing standards, streamlining processes, and expediting onboarding. Working backwards, we built and iterated over dozens of features to address those requirements (e.g. code inventory, dependencies analysis, project management, versions management and code transformation). We tirelessly continue this investment by mixing innovation initiatives and incorporating users' feedback on ongoing projects. The rise of Generative AI (GenAI) and services like Amazon Bedrock allow us to easily deliver and rapidly imagine more features. We are excited to present the available features across various areas of mainframe codebase assessment and code transformation support.

Code Summary

Code Summary offers Generative AI code summary of each mainframe source file which can then be used as metadata for AWS Blu Insights Query Language (BQL) filtering and code decomposition during mainframe codebase assessment. In Blu Insights, each source code file has a list of meta properties. Certain metrics (e.g. lines of code) are automatically calculated, while others (e.g. business domain) can be added by users. The assessment phase relies on those meta properties to decompose the modernization project into different iterations, considering various technical and functional parameters. By generating summaries that can be used as metadata, this capability can enable faster and better legacy application decomposition and refactoring to Java.

Code Explanation

Code explanation capability provides rapid Generative AI based creation of high-level explanation of what a specific code snippet does just by selecting it. This can be useful while testing functional equivalence of modernized mainframe applications.

BMS Preview (User Screen Design Preview)

BMS Preview capability creates AI-generated previews of mainframe Customer Information Control Systems (CICS) application screens, using Basic Mapping Support (BMS) files that define the structure and appearance of user screens. A BMS file defines the attributes of the screens, including fields, labels, and navigation elements, presented to users to facilitate interaction with the application through a terminal interface. Developers can request a preview of the screen as displayed to end-users of the legacy application. This preview helps developers rapidly check the expected layout in the modernized application, and perform iterative design of user experience screens.

BQL Generation

BQL Generation uses Generative AI for natural language driven creation of AWS Blu Insights Query Language (BQL) query filters used during mainframe codebase assessment. AWS Blu Insights Query Language (BQL) is a flexible custom language, similar to SQL, that allows filtering artifacts throughout various Blu Insights views. It is one of the most used features, especially during the assessment phase where users decompose the application into work packages. The query filter generation capability is useful to look for specific information in the codebase and automatically generate, display, and run BQL filters after optional human reviews and adjustments.

Comments Translation

Comments translation capability employs AI for real-time translation and visualization of foreign-language comments in a codebase from any language to English, without altering the source files. It can be useful to globally distributed multi-lingual developer teams collaborating on code development.

Classification

Codebase classification offers Generative AI based file type detection. In legacy environments, such as mainframes, it is common to have source code files without extensions. Recognizing the programming languages (usually dozens) that compose each file can be challenging. The goal of codebase classification capability is to detect those languages by analyzing the files' content. Blu Insights already handles dozens of programming languages. However, large codebases (50,000+ files) typically include files with rare languages and frameworks. With the enhanced codebase classification capability, Blu Insights can detect the types of those files. Such files classified using Generative AI will be marked to stand out from the rest. Users can always leverage the Manage Types and Manage Extensions features to customize the results. Such intelligent classification can speed up preparation of large codebases for modernization.

Code Transformation

Code transformation capability offers Generative AI enabled conversion of mainframe-based source files written in languages like REXX, C, or Easytrieve to Java. AWS Blu Insights automates the code transformation process using the Transformation Center which handles the critical programming languages that contain the business logic of the legacy application. Some codebases contain dozens of source code files written in other languages (e.g. ASM, C, Easytrieve) for reporting, scripting, operations, etc. Such files are usually short (a few dozen lines) and don't contain business logic, making them easy to transform. Leveraging Generative AI, the Transformation Center translates those files into Java code that can be manually verified and integrated in the final solution.

Notes-to-actions “Do it for me!”

Notes-to-actions “Do it for me!” capability uses Generative AI to augment AWS Blu Insights project management features with textual inputs, e.g. meeting or personal notes, to trigger automated actions like scheduled events, To-Dos card creation, and invitations to team members. AWS Blu Insights already offers a set of features, such as to-dos, schedules, teams, timelines, documents, etc., that allow managing and tracking modernization projects. With the notes-to-action capability coupled with the existing features, users will be able to create events in schedules, cards in to-dos, and invite team members automatically using textual inputs (e.g. meeting or personal notes). By turning notes into automated actions, this capability helps users save time, reduce low-value project management overhead, and enhance team collaboration.

Activities Summary

Activities Summary capability automatically generates a high-level summary of a project's last week activities and displays it in the Dashboard to enable faster reviews on project changes. A typical ongoing project has dozens of daily activities (file update, to-do card creation, team member invitation, etc.). Activities summary is useful to quickly identify the main information of the latest activities performed in a given project.

As impressive as LLMs are, the current level of generative AI technology is not perfect and LLMs are not infallible. Bias and incorrect answers, although rare, should be expected. Please exercise reasonable caution when relying on AI outputs. For more information refer to What are Large Language Models (LLM)? and our AWS Responsible AI Policy.