AWS AI-DLC Workflows 2.0 experimental

Core approach

Encodes the AI-Driven Development Life Cycle as one harness-neutral core/ rendered into native distributions for several coding agents. A conductor selects or composes an appropriate scope, coordinates 14 specialist and review agents across 5 phases and up to 32 stages, persists every intent's artifacts and audit trail under aidlc/spaces/, and pauses at approval and phase-verification gates before downstream work proceeds.

Workflow

1. Initialization — scaffold an intent record, detect greenfield or brownfield context, and initialize deterministic state and audit records
2. Ideation — capture intent, assess feasibility, define scope, form the team, explore mockups, and approve the initiative
3. Inception — reverse-engineer existing code when needed, discover team practices, analyze requirements, design the application, generate units, and plan delivery
4. Construction — implement reviewable Bolt slices through functional, non-functional, and infrastructure design, code generation, build and test, and CI
5. Operation — prepare and execute deployment, provision environments, establish observability and incident response, validate performance, and feed findings into the next intent
6. Verification and control — run traceability checks between phases, approve material decisions, and resume, redo, or jump between persisted stages with /aidlc ($aidlc in Codex)

Supported tools

• Claude Code
• Kiro
• Codex
• OpenCode

Notes

AWS AI-DLC Workflows 2.0 is the executable implementation of AWS’s broader AI-Driven Development Life Cycle, not a new name for AI Unified Process. Its distinctive boundary is the whole delivery system: the version 2 documentation defines requirements and design artifacts, but also a deterministic state machine, specialist reviewers, construction slices, deployment stages, operational feedback, and durable team knowledge. The methodology is unusually explicit about its ceremony: a feature or enterprise scope includes all 32 stages, while smaller stock scopes and the adaptive composer remove stages only after showing the proposed plan for approval. That breadth and its fast-moving v2 implementation support an experimental rating rather than treating the project’s GA label as evidence of broad adoption or a settled methodology. This assessment observes the v2 branch at 6b26408.

Added Aug 1, 2026 · Assessment last reviewed Aug 1, 2026 · How we track