AIDLC: AI-Driven Development Lifecycle
4 min read
AIDLC (AI-Driven Development Lifecycle) organizes AI-assisted requirements analysis, design, implementation, and testing around Intent → Unit → Bolt. An Intent states a business goal, a Unit divides that goal into a bounded piece of work, and a Bolt is a short iteration for carrying out the work. AI proposes plans and artifacts; people review them against the project's requirements and constraints.
4 Tracks
The AIDLC guide is organized into 4 tracks based on the reader's role and interests.
Figure 1. The four tracks of the AIDLC guide and their relationships
Expand the diagram to read labels and follow connections.
Diagram
Learning Path by Role
Table 1. Recommended learning path by role
| Role | Recommended Path |
|---|---|
| Executives · PM | Enterprise Adoption → Cost Effectiveness → Case Studies |
| Architects | Methodology → Ontology → Harness → MSA Complexity |
| Developers | 10 Principles → DDD Integration → AI Coding Agents |
| Operations · SRE | AgenticOps → Observability → Autonomous Response |
| Security · Compliance | Governance → Harness Engineering → Open Weight Models |
Core Concepts
Dual Axes of Reliability: Ontology × Harness
This playbook’s optional reliability framework separates domain definitions from the mechanisms that check and enforce them:
- Ontology (WHAT + WHEN): Defines entities, relationships, and domain constraints in a typed schema. Inner/Middle/Outer feedback loops provide input for revising those definitions.
- Harness Engineering (HOW): Implements checks and execution limits around agent work, including circuit breakers, retry budgets, and output gates. Reliability depends on which constraints are enforced and how the controls are validated.