Before AI coding assistants became mainstream and really good, we partnered with a global financial behemoth to modernize a mission-critical Fund Management and Administration platform built on legacy technologies - using AI.

This program stands out as an early enterprise adoption of AI-assisted software delivery, where LLMs were used not as experimental tools, but as practical accelerators in a large-scale modernization effort.

The client’s platform was critical to investor reporting and financial operations but faced severe limitations:

• Built on Classic ASP and VBScript (near end-of-life)
• ~500,000 lines of tightly coupled code
• Business logic embedded within UI layers
• Single-server deployment with no scalability
• File system–dependent storage with no resilient backup
• High infrastructure and maintenance costs
• Limited internal understanding of system behaviour

At the time, late 2024, traditional migration approaches would have been slow, costly, and risky.

This program was initiated before coding agents became widely adopted.

We leveraged early LLM-based tools (such as Gemini) and initial versions of Copilot to:

• Accelerate code translation from Classic ASP to ASP.NET MVC
• Assist in identifying reusable patterns
• Support developers in navigating a large, poorly documented codebase

Unlike today’s mature AI workflows, these tools required heavy human validation, making this a hybrid model of AI-assisted + engineering-led delivery.

1. AI-Accelerated Code Migration:
   Used LLMs to assist in translating legacy code while engineers validated and refactored outputs.‍
2. Architectural Modernization‍
   
   • Migrated from coupled 2-tier to 3-tier MVC architecture‍
   • Introduced modular design and reusable components
   • Containerized applications using Docker
   • Enabled deployment on Kubernetes‍
3. Cloud-Native Transformation‍
   
   • Migrated file storage from local systems to AWS S3‍
   • Enabled scalable backup and recovery‍
4. Phased Delivery‍
   
   • Phase 1–2: Investment facing Portal
   • Phase 3–6: Admin Portal (internal operations)‍
5. Custom Validation Framework:
   Built an automated framework to compare outputs between legacy and new systems—critical due to AI-assisted code generation.

‍

Because this was done before modern AI tooling matured, the team navigated:

• LLM hallucinations in generated code
• Inconsistent translation of business logic
• Poorly structured legacy HTML incompatible with modern frameworks
• Limited access to real production data
• Lack of mature AI debugging workflows

✅ Investments Portal successfully deployed (November)

✅ Admin Portal deployed (January)

✅ No major production incidents post-release

✅ Cloud-native, scalable architecture achieved

✅ Significant improvements in maintainability

• Reduced technical and security risk
• Established a modern, extensible platform
• Enabled future transformation (API-first, React-based UI)
• Demonstrated real-world viability of AI-assisted engineering