Revolutionizing ADRs with GenAI
summary
Describes using custom LLMs to semi-automate Architecture Decision Record (ADR) creation by ingesting meeting transcripts and Google Drive documents with a Python app that chunks long transcripts for model context and runs a multi-agent draft–reviewer–writer pipeline. Produces outputs like draft ADRs, PlantUML/mermaid diagrams, and a polished Google Docs file; demonstrated generating 55 pages in 4 hours, used Teams Enterprise sandboxing for data privacy, and reports the workflow won a project and met generated timelines.
AI-written summary · about
chapters (12)
- 0:00GenAI experiment intro
- 5:32Architect's role in ADRs
- 10:08ADR documentation problems
- 12:38Optimizing architect work
- 15:29Custom ADR bot demo
- 20:13Multi-agent ADR generation
- 22:25ADR diagram limits
- 26:29GenAI architecture use cases
- 27:30Text generation risks
- 30:19New model capabilities
- 33:16Context window limits
- 34:43GenAI data security
related talks
21mRevolutionizing software creation: Harnessing the power of GenAI
32mRevolutionizing Automation: Retrieval-Augmentation with LLMs
27mAI-First Software Maintenance & AI-First Software Delivery
18mLLMs and GenAI: A Guide to Building Your Own RAG Model for Document Interaction
39mLeveraging AI and Emerging Technologies for Competitive Advantage
24mNatural Language Coding Revolutionising Programming with GenAI
30mUsing AI to Create Better Requirements and Designs
43mDesign Thinking: Generative AI Style