This research introduces DRAFT (Decisions Recorded As Formalized Text), a novel framework that uses LLMs to generate Architecture Decision Records (ADRs) with minimal human effort.

• Addresses the challenge of documenting architectural decisions, which is critical but often neglected
• Proposes a structured process combining domain-specific prompting and iterative refinement
• Demonstrates significant reduction in manual documentation effort while maintaining quality
• Addresses privacy concerns by implementing local model deployment options

For Engineering teams, this approach offers a practical solution to capture architectural knowledge systematically, improving knowledge transfer and reducing technical debt while respecting organizational security requirements.

DRAFT-ing Architectural Design Decisions using LLMs