AI-DLC: how AWS is trying to turn SDD into an operating model (Category AI4SDLC)

Dealing with the document.AI-Driven Development Lifecycle (AI-DLC) Method DefinitionRaja SP is the Principal Solutions Architect of Amazon Web Services. The document appeared about a year ago and Official AWS DevOps blog post With reference to this whitepaper came 31 July 2025 years. Interestingly, AWS presented Kiro 14 July 2025 Agentic IDE with spec-driven development requirements.md, design.md, tasks.md, steering files, hooks and so on. AI-DLC appears immediately after that and looks not like a standalone tool, but as an attempt to give Kiro/Amazon Q a more mature methodological framework for enterprise development.

The paper formulates a gap between the two extremes and proposes a third AI-DLC mode.

1. AI-assisted development Helps in small tasks: code, tests, documentation
2. AI-autonomous developmenT promises to build an app almost without a person, but according to the guys at AWS this approach quickly comes down to quality and control.
3. AI-DLC assumes that AI leads the process, but the person remains the owner of the intention, risk and final decision.

The basic concept is built in the reverse direction of the conversation – now it is not a person who constantly asks AI to “write me this”, but AI itself lays out intention on plans, questions, trade-offs, units and tasks, and people validate decisions at critical points. In order to work on this AI-DLC process

• Introduces its artifacts: Intent, Unit, Bolt, Domain Design, Logical Design, Deployment Unit
• Phase Inception, Construction, Operations
• Introduces rituals like Mob Elaboration and Mob Construction
• Special attention is paid to DDD (domain driven design)AI should not just write code, but help to highlight bounded contexts, user stories, ADRs, tests, infrastructure, and deployment units.

Compared to Kiro and SDD from AWS, the difference is similar.

1. Kiro - it's a product interface. Its spec-flow turns a prompt into three clear files: requirements, design, tasks. The new versions include Quick Plan, bugfix specs and parallel running of independent tasks. This is a practical form of SDD for feature or bugfix inside a specific repository. Kiro's SDD is the discipline of "first fix durable spec, then let the agent execute." Marc Brooker describes the specification well as the big picture and human-readable super prompt: it holds intent, makes changes versionable, and reduces the chaos of prompt-by-prompt development.
2. AI-DLC It offers not only to “describe the features for the agent”, but “to fix how the entire delivery system should work if AI became a member of the SDLC”. So there's operations, risks, audit trails, different depths of the process, gates for people to confirm, and the idea that the workflow doesn't have to be hard-wired. For a small feature, a full AI-DLC will be heavy, but for the modernization of legacy, several commands, a regulated domain, it can suit well.

The original document also had a continuation. 29 November 2025 AWS published two posts this year: open-source adaptive workflows for AI-DLC and Walkthrough for Amazon Q Developer. There, AI-DLC is already turning from PDF to rules and steering files for agents such as Amazon Q Rules and Kiro Steering. In the repository. awslabs/aidlc-workflows stable v1 releases.0.0/v1.0.1 came out 19 and 30 June 2026 years.

The movement went even more interesting: in README Announced AI-DLC Workflows 2.0 Preview. Branch v2 describes the approach "one core, many harnesses": Claude Code, Kiro IDE, Kiro CLI, Codex CLI. There is no longer just a set of markdown rules, but an attempt to assemble a workflow engine: 5 phase 32 stages, 11 domain-expert agents, adaptive scopes, depth levels, test strategy levels, approval gates, two-tier knowledge system, learning loop and structured audit trail.

That is, the direction of the continuation is clear: from the methodological manifesto to the executable development infrastructure. At first, the whitepaper explained why the old SDLC needed to be rethought. AWS then gave rules/steering to try out in Kiro and Amazon Q. Now v2 is trying to make AI-DLC more verifiable, portable between agent harnesses and less dependent on manual prompt engineering.

#AI #AI4SDLC #Engineering #Architecture #DevTools #Management #Agents