The Agentic
Design Loop.
A high-velocity framework built for designers who ship.
Bridging strategy and code with targeted AI intervention.
Discoverability
Aggregating raw qualitative data and SME insights into a structured, queryable knowledge base.
Key Actions
- Discovery sessions with subject matter experts
- Consolidation of discovery logs
- Feed logs and educational artifacts into NotebookLM for RAG-based context
Active Tool
NotebookLM
Definition & Specs
Transforming high-level requirements into technical blueprints and persona-aligned success criteria.
Key Actions
- Business requirements meet persona-based JTBD
- Specific KPI & Success Criteria definition
- Feed data into Gemini Canvas for structural synthesis
- Architecting the master prompt for Claude Code
Active Tool
Gemini Canvas
AI-Assisted Design
Rapidly scaffolding components and design patterns directly within the codebase using agentic coding tools.
Key Actions
- Branching strategy: Git checkout -b ai-design-sprint
- Execution of structural prompt via Claude Code
- Iterative design cycles & output assessment
- Initial solution prototyping within the codebase
Active Tool
Claude Code + VS Code + GitHub
Socialization & Feedback
Bridging the gap between code and stakeholders with asynchronous walkthroughs and engineering alignment.
Key Actions
- Loom video walkthrough of the AI-generated solution
- SME interaction and qualitative feedback loops
- Engineering sync: Define backend dependencies in parallel
Active Tool
Loom + Expert Review
Finalization
Applying final refinements, updating documentation, and shipping to staging for peer review.
Key Actions
- Alignment with design principles & pattern consistency
- Modifying project documentation (.MD files)
- Final Pull Request for dev review & backend integration
Active Tool
Claude Code (PR)