C3: AI First
Organizations retrofit AI into existing processes, treating it as an afterthought. This creates friction, inefficiency, and underutilizes AI capabilities. Human-centric-only designs often require significant rework for AI integration.
AI First means asking "How would AI approach this?" before "How have we always done this?" It shifts the default from human-only to human-AI collaboration, enabling multiplicative productivity gains.
Principle
- All technical and process decisions MUST consider AI-native approaches as the default.
Design systems, workflows, and documentation for optimal AI collaboration from the outset. Prefer Markdown and OpenAPI over Word and PDF. Git-based knowledge management over wiki pages and email threads. Decisions made in meetings without documentation are lost. AI cannot learn from or act on undocumented decisions.
- Product decisions MUST occur within agent-manageable contexts rather than extended human-only meetings without actionable outcomes.
Context is agent-manageable when it is accessible (stored in files AI can read), structured (organized predictably), current (reflects latest decisions), traceable (history visible via version control), and actionable (clear enough to implement). Security-sensitive operations, regulatory compliance, and user experience critical paths may require human-first approaches.
Related Principles
- C1: Context Engineering Competency
- E1: Design-Centric Work
- G1: Single Source of Truth
- G3: Centralized Requirements
Related Proposals
- AI-DLC Mob Elaboration
- AI-First Decision Making
- Agent-Friendly Knowledge Base
- Design System
- Frontmatter Spec Coordination