AWS re:Invent 2025 - Driving modernization using Mphasis’ Agentic AI framework (MAM219)

Driving Modernization with Mphasis' Agentic AI Framework (MAM219)

Legacy Modernization Challenges

• CIOs struggle to innovate fast enough due to the risk of touching legacy platforms
• Enterprises are "anchored" on deeply monolithic legacy systems with embedded business logic
• Lack of specialized engineers to maintain legacy systems leads to technical debt and costly changes

Mphasis' Agentic AI Approach

• Focuses on extracting intelligence from legacy systems and converting it to data, rather than just modernizing code
• Utilizes a suite of autonomous and semi-autonomous AI agents to:
  1. Neozeta: Reads and converts legacy code/documents into human-understandable knowledge
  2. Neosaba: Generates user stories, governance, compliance, and business process insights from the extracted knowledge
  3. Neorena: Defines a customizable target state architecture aligned to enterprise standards
  4. Neorux: Prompts existing coding agents to automatically generate new code from the defined architecture
• Builds an "Ontosphere" - an enterprise knowledge graph that captures the meaning and context of the extracted intelligence

Technical Details and Results

• Neozeta can reverse-engineer legacy COBOL code, generating a data dictionary, business rules, and confidence scores using large language models
• The extracted knowledge is mapped to domain ontologies and ingested into the Ontosphere knowledge graph
• Neosaba allows business analysts to reimagine the application by defining epics, features, user stories, and acceptance criteria
• Neorena generates logical, physical, and observability models based on defined standards and a provided playbook
• Neorux can generate Java Flink code from the models, with performance comparisons between CPU and GPU execution
• A typical 50 million line of code modernization project can be completed in 7 years using this approach, compared to the industry average of 7 years

Business Impact and Applications

• Eliminates the need to maintain legacy systems by extracting the embedded intelligence and converting it to a living, evolving knowledge graph
• Enables faster innovation and agility by automating the modernization process and aligning it to enterprise standards
• Provides a future-proof solution that avoids the cycle of legacy system re-emergence
• Applicable to various industries with deeply embedded legacy systems, such as finance, insurance, and engineering

Key Takeaways

• Mphasis' Agentic AI framework takes a fundamentally different approach to legacy modernization by focusing on extracting and preserving the intelligence, rather than just rewriting the code
• The suite of AI agents automates the entire modernization lifecycle, from reverse-engineering legacy systems to generating new, standards-aligned architectures and code
• The Ontosphere knowledge graph serves as a central, living repository of enterprise intelligence that can continuously evolve and power future innovations
• This approach can significantly accelerate modernization projects and prevent the re-emergence of legacy systems, providing a future-proof solution for enterprises