Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification.

Agentic Software Development Lifecycle

The agentic software development lifecycle structures AI assistance across requirements, architecture, implementation, testing, and delivery rather than treating coding agents as isolated autocomplete tools. BMAD packages this as specialised agents and guided workflows for AI-driven agile development ^[src-184].

Related

• AI Coding Agents
• Agent Workflows
• Agile AI-Driven Development

Historical recovered snapshot — 2026-08-23

Preserved from ai-watch-src393-ai-native-sdlc-playbook-20260823. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

The agentic software development lifecycle structures AI assistance across requirements, architecture, implementation, testing, and delivery rather than treating coding agents as isolated autocomplete tools. BMAD packages this as specialised agents and guided workflows for AI-driven agile development ^[src-253].

• For legacy systems, the lifecycle must expose hidden assumptions before implementation: inventory dependencies, establish behavioural tests, review a bounded plan, execute incrementally and verify acceptance outcomes ^[src-385].
• Model capability does not replace release controls. Tests, code review, security review, observability and rollback remain independent evidence layers ^[src-385].

• Anthropic's playbook makes intent, specification, plan, code/test evidence, review and incident records the durable state transitions of an AI-assisted lifecycle ^[src-393].
• Agents accelerate work between gates; product owners, code owners and operators retain accountable decisions. This is governed delegation, not autonomous end-to-end delivery ^[src-393].
• Git history can connect those artifacts into an audit trail, but the source reports no controlled improvement in DORA, defect rate or production safety ^[src-393].

• ^[src-253] bmad-code-org — BMAD method (2026-06-22)
• ^[src-385] Google Cloud Tech — legacy-code modernisation tutorial (2026-08-20)
• ^[src-393] Louis Claxton / Claude by Anthropic — “The AI-Native SDLC playbook” (2026-08-21)

September 18 evidence update

Verified on 2026-09-18; scope is the new sources below, not a refresh of all historical claims.

The new OpenAI and GitHub accounts reinforce explicit review, release and compatibility gates around agent implementation. Their reported adoption and code-volume figures remain case-study evidence, not universal productivity estimates. [src-484, src-485]

See Agent Delivery Evidence for attribution, availability and source summaries.

BMAD implementation demonstration

The Benjamin Code walkthrough shows the lifecycle as a chain of human-readable state transitions: discovery and brainstorming, PRD, UX and architecture, epics and stories, implementation, tests and optional QA. Story-sized context narrows the agent's working set, while an architecture index lets it retrieve only the relevant design sections. This is practical evidence of the workflow shape, not a controlled comparison of delivery speed, defects or maintainability. ^[src-522]