Reimagining Modernization with Agentic AI: Commonwealth Bank of Australia's Journey
Overview
Commonwealth Bank of Australia (CBA) embarked on an ambitious modernization program to migrate applications from on-premises to the cloud, leveraging agentic AI to transform a slow, manual process into a scalable, high-velocity engine. This presentation showcases how CBA and AWS Pro collaborated to reimagine the modernization journey, overcoming challenges of scale, technical debt, and lack of documentation.
Modernization Challenges
- 70% of workloads still on-premises, with legacy code written over 20 years ago
- Average enterprise transformation time of 1-2 years
- Slow, labor-intensive process due to limited resources and expertise
- Limited scalability due to monolithic application architectures
- Lack of documentation and knowledge about legacy applications
AWS Migration Framework
CBA customized the AWS migration framework to fit their internal DevOps hosting platform (DHP), which provided deployment automation, continuous delivery, and immutable environments. However, even with this platform, the team faced significant challenges:
- Extensive documentation archaeology required to understand legacy applications
- Significant technical debt, such as outdated packages and missing binaries
- Baseline of only 10 applications migrated per year
Agentic AI Approach
To address these challenges, CBA placed agentic AI at the center of their modernization efforts, leveraging various AI agent patterns:
- Basic Reasoning Agents: Understand policy and licensing documents to determine software upgrade requirements.
- Tool-based Agents: Generate code that aligns with enterprise standards by calling APIs for compliance and formatting.
- Memory-augmented Agents: Leverage short-term memory to minimize peer review comments and long-term memory to apply best practices across sessions.
- Multi-agent Workflows: Orchestrate complex tasks, such as code discovery, refactoring, testing, and compliance documentation generation.
The Lumos Modernization Portal
CBA developed an in-house modernization portal called Lumos, which leverages this agentic AI approach across the following key stages:
Analyze and Design
- Capture application details through interviews with owners
- Analyze network connectivity and code structure
- Generate solution documents and cybersecurity assessments using multi-agent workflows
Transform and Test
- Automate code modernization using open-source tools (OpenRewrite) and custom agents (Q Developer)
- Generate UI test cases using AI-powered Selenium script generation
Deploy and Operate
- Leverage CBA's DevOps hosting platform (DHP) for deployment automation, continuous delivery, and immutable environments
- Integrate agentic AI into the deployment workflow to ensure compliance and reliability
Results and Future Improvements
- Increased modernization velocity from 10 applications per year to 20-30 per quarter
- Improved reliability, safety, and speed of modernization efforts
- Ongoing work to further automate the process, including:
- Agents that continuously monitor and update documentation
- Cross-repository analysis to understand dependencies
- Expansion to support more programming languages
Key Takeaways
- Placing agentic AI at the center of the modernization process can significantly improve speed, scale, and reliability
- Combining deterministic engines (e.g., static code analysis) with AI agents provides a powerful, auditable approach
- Developing a customized, end-to-end modernization platform (like Lumos) can streamline the journey and unlock new capabilities
- Continuous improvement and automation are crucial to sustaining the benefits of an agentic AI-powered modernization program
