The Inversion of the SDLC: From Software Development Lifecycle to Agentic Development Lifecycle (ADLC)
The Software Development Lifecycle (SDLC), as it has been taught and audited for half a century, has taken many forms, from phase-gated waterfall to spiral, agile, and DevOps. Every one of them presumes that humans manually produce the artifacts (requirements, designs, code, tests, and documentation), whether those artifacts move through sign-offs and review boards or through sprints and deployment pipelines. That presumption is now obsolete in practice, even where it remains intact on org charts. A new operating model has emerged that this paper calls the Agentic Development Lifecycle (ADLC). Under ADLC, humans no longer produce artifacts directly; they orchestrate systems (language models, agents, design-token engines, infrastructure-as-code platforms, and policy engines) that generate artifacts dynamically. Requirements migrate into executable specifications and constraints. Design becomes a machine-readable governance layer. Quality assurance becomes continuous synthetic validation. Documentation is backfilled from working systems rather than written before them. Developers shift from authoring to orchestrating and reviewing. Approval gates compress from fixed review checkpoints to risk-tiered human-in-the-loop interventions.
Research library