Accelerated Business Requirements Documentation Generation for Regional Electric Utility Provider

460,000

Lines of Code Analyzed

9,900%

Accelerated BRD generation

100%

Program Coverage

The Customer

One of the largest electric utility providers in the United States, serving millions of customers.

The Challenge

The utility customer needed to modernize a large IBM Z mainframe application consisting of over 460,000 lines of COBOL code. The customer required comprehensive Business Requirements Documentation (BRD) to support their modernization initiative and enable an application rewrite strategy.

Working with their systems integration partner, PwC, the customer faced the challenge of extracting and documenting complex business logic from hundreds of legacy COBOL programs – while lacking clear initial visibility into the target-state architecture. The project required significant modernization tooling knowledge transfer and training, while managing the technical complexity of processing over 250 COBOL programs efficiently and accurately.

Additionally, the customer needed to accelerate BRD generation dramatically to meet project timelines, requiring optimization of both processing infrastructure and AI model selection to achieve production-scale throughput.

The Approach

The AveriSource Platform enabled the customer to rapidly generate comprehensive business requirements documentation from their extensive IBM Z COBOL portfolio, supporting an accelerated rewrite strategy through AI-powered business rule extraction and requirements generation. The AveriSource Platform was deployed on Amazon Web Services (AWS) leveraging Amazon EC2 (Elastic Compute Cloud) for horizontally scaled compute resources and Amazon Bedrock for AI-driven services.

Inventory and Discover

Using AveriSource Inventory and Discover, the customer and PwC conducted a comprehensive analysis of the COBOL code – mapping data flows, program structures, and interdependencies, providing a complete understanding of the legacy application estate.

Product Reports and Features Utilized:

Calibrated parsing
Program Analysis
Business Logic Visualization
Dependencies
Data Lineage

Analyze

Using AveriSource Analyze with integrated Amazon Bedrock Generative AI (GenAI) capabilities, the team identified, analyzed, and extracted core business rules from the COBOL programs. The AveriSource Platform leveraged advanced AI models to automatically generate business requirements documentation.

Product Reports and Features Utilized:

Business Rules Extraction
AI-Powered BRD Generation
Code Analysis
Rule Analysis
Amazon Bedrock Integration

Platform Calibration and Optimization

The AveriSource customer success team provided hands-on coaching and knowledge transfer via Microsoft Teams, ensuring smooth tooling adoption and effective onboarding.

The AveriSource delivery team optimized document generation by testing multiple Anthropic large language models and ultimately achieving optimal results with a throughput-optimized Meta Llama model.

Through iterative optimization, the infrastructure was scaled to seven GenAI service instances, each running 20 containers, enabling 140 parallel processing threads. This architecture, combined with targeted AI model optimization, accelerated BRD generation from 1 to 100 per hour—a dramatic 100x improvement in productivity that enabled the project to meet aggressive timeline requirements.

The Results

The AveriSource Platform successfully delivered comprehensive business requirements documentation for the customer’s mainframe modernization initiative, achieving the following outcomes:

460,000+ lines of IBM Z mainframe COBOL code analyzed across 253 programs
280 Business Requirements Documents (BRDs) generated, including 243 concise BRDs using Meta Llama 3.1 70b and 37 using Anthropic Claude 3.0
9,900% acceleration in BRD generation—from 1 per hour to 100 per hour through infrastructure scaling and AI optimization
100% program coverage achieved, with successful BRD generation for all programs including initial failures
Successful knowledge transfer and tool adoption through remote coaching via Microsoft Teams
Customer satisfaction with BRD quality and completeness, jumpstarting their application rewrite initiative