AirSDLC Overview¶

Introduction¶

The AI-Responsible Software Development Lifecycle (AirSDLC) is an actionable framework for AI-driven software development. It provides a structured methodology for transforming ambiguous business requirements into production-ready code through a series of validated, traceable artifacts.

The Problem AirSDLC Solves¶

Traditional software development faces several persistent challenges:

Requirements Gap: Business intent often gets lost in translation to technical implementation
Premature Technical Decisions: Teams jump to "how" before fully understanding "what"
Documentation Debt: Design decisions are poorly documented or quickly become stale
Context Loss: When bugs occur, engineers lack the full context from inception to deployment
Inconsistent Processes: Different teams solve the same architectural problems differently

AirSDLC addresses these challenges by formalizing a structured, AI-augmented process with built-in traceability and validation.

Core Concept: AI-Driven with Human Oversight¶

AirSDLC inverts the traditional human-AI relationship in software development:

Traditional Model: Human writes code → AI assists with completion
AirSDLC Model: AI generates artifacts → Human validates and directs

In this model: - AI's Role: Generate plans, model domains, propose solutions, identify risks - Human's Role: Validate outputs, make strategic decisions, provide domain expertise, give final approval

This shift moves the human expert from creator to curator—a higher-leverage position that focuses cognitive energy on evaluation rather than generation.

The Three Sequential Phases¶

AirSDLC is built on a sequential knowledge handoff where each phase enriches context for the next:

Business Intent → Inception → Design → Construction & Operations
                     ↓           ↓              ↓
                   AI-DAA     AI-ADR      Production Code

Phase 1: Inception (The "WHAT")¶

Objective: Translate business goals into a pure, technology-agnostic domain model

Input: Business Intent / Product Requirements Document (PRD)
Process: AI analyzes requirements and applies Domain-Driven Design (DDD) patterns
Output: AI-DAA (Domain Architecture Analysis) - a vendor-neutral domain model
Validation: Human expert reviews and validates the domain understanding

Phase 2: Collaborative Design (The "HOW")¶

Objective: Synthesize domain understanding with technical implementation decisions

Input: Validated AI-DAA + Engineer's RFC (Request for Comments)
Process: Human-AI collaborative "Socratic Sparring" session using Architectural Playbook
Output: AI-ADR (Architectural Decision Record) - finalized technical decisions
Validation: Human architect approves architectural choices and trade-offs

Phase 3: Construction & Operations (The "BUILD & RUN")¶

Objective: Generate code, deploy, and provide context-aware monitoring

Input: Finalized AI-ADR
Process: Break work into "Bolts" (focused implementation units), AI generates boilerplate
Output: Production-ready code, tests, and deployment artifacts
Operations: Context-aware incident response using full traceability chain

Key Principles¶

1. Technology Agnosticism in Domain Modeling¶

The AI-DAA is intentionally 100% technology-neutral. It describes the business domain using DDD patterns without specifying databases, frameworks, or languages. This ensures domain understanding is not polluted by technical constraints.

2. Sequential Knowledge Handoff¶

Each phase's validated output becomes the non-negotiable input for the next. This creates an unbroken chain of traceability:

Business Goal → PRD Feature → DAA Aggregate → ADR Decision → Code Implementation

3. Validation as First-Class Activity¶

Validation is never skipped. Every AI-generated artifact passes through human review before proceeding. This includes: - PRD-to-DAA Coverage: Does the domain model address all requirements? - DAA-to-ADR Alignment: Do technical decisions respect domain invariants? - ADR-to-Code Compliance: Does implementation match the design?

4. Pragmatism & Adaptability¶

AirSDLC provides two workflow paths: - Full Workflow: PRD → AI-DAA → RFC → AI-ADR (for complex/risky features) - Lightweight Workflow: PRD → TIP → RFC → AI-ADR (for simple/well-understood features)

Teams choose based on feature complexity, risk, and domain familiarity.

5. Living Knowledge Repository¶

All artifacts are stored in a "Knowledge Repository" that serves as the single source of truth. This repository enables: - End-to-end traceability - Context-aware operations - AI-powered analysis and reporting - Post-mortem correlation

Core Artifacts¶

AirSDLC defines several key artifact types:

Artifact	Purpose	Phase	Technology-Specific?
PRD	Business requirements	Inception	No
AI-DAA	Domain model (DDD-based)	Inception	No
TIP	Technical proposal (lightweight)	Inception	Yes
RFC	Design discussion forum	Design	Yes
AI-ADR	Finalized architecture decisions	Design	Yes
Bolt	Discrete implementation unit	Construction	Yes
Playbook	Library of approved patterns	Design (reference)	Yes

See Artifacts for detailed specifications.

The Workflow at a Glance¶

graph TD
    A[Business Intent/PRD] --> B{Complex Feature?}
    B -->|Yes| C[Generate AI-DAA]
    B -->|No| D[Write TIP]
    C --> E[Create RFC]
    D --> E
    E --> F[Collaborative Design Session]
    F --> G[Consult Playbook]
    G --> H[AI as Socratic Partner]
    H --> I[Generate AI-ADR]
    I --> J[Break into Bolts]
    J --> K[Generate Code]
    K --> L[Deploy & Monitor]

See Workflow for step-by-step details.

Benefits¶

For Engineering Teams¶

Faster Design Cycles: AI generates first drafts, humans focus on validation
Better Documentation: Design decisions are captured by default, not as an afterthought
Consistent Patterns: Architectural Playbook ensures team-wide consistency
Reduced Context Switching: Full traceability from business goal to code

For Architects¶

Strategic Focus: Spend time on high-level decisions, not boilerplate
Knowledge Capture: Architectural wisdom is codified in reusable patterns
AI-Augmented Review: AI challenges designs using organizational knowledge

For Operations¶

Context-Rich Debugging: Correlate production errors to DAA/ADR/PRD
Faster MTTR: AI proposes fixes using full design context
Post-Mortem Integration: Learnings feed back into Knowledge Repository

For Product Teams¶

Requirements Validation: AI-generated DAA surfaces misunderstandings early
Traceability: Track which features map to which technical components
Predictable Estimates: Complexity metrics based on domain model

Framework vs. Implementation¶

Important Distinction: - AirSDLC (this repository): The framework specification, manifesto, and guidelines - Implementation Tools: Concrete software that implements the framework

This separation ensures: - Multiple implementations can exist - The framework evolves independently from tools - Teams can adapt implementations to their needs while maintaining conformance

Next Steps¶

Understand the Foundation: Read Philosophy for core AI-DLC principles
Learn the Process: Study Workflow for practical guidance
Explore Artifacts: Review Artifacts for detailed specifications
See Examples: Browse Examples for sample artifacts

Extensibility¶

AirSDLC is designed to be extended. Teams can: - Add custom artifact types - Define additional validation steps - Integrate domain-specific patterns into the Playbook - Customize templates and workflows

See Extensibility for guidelines on adapting the framework.

Next: Philosophy - Core principles and AI-DLC foundation