Evolution of the Software Development Lifecycle (SDLC)

| Essay

Into the Agentic Era

Cover slide over a black-and-white photograph of Ford’s Highland Park assembly line, circa 1913. Photograph from the Library of Congress.

SDLC in the Agentic Era

In 2026, software engineering is undergoing its most dramatic change yet.

So how does this change the disciplined software engineering process?

Let’s look at how the lifecycle evolved to understand it.

SDLC v1 (pre-2000). Waterfall with handovers. Planning, requirements and spec were seen as upfront design. Development, testing and deployment were treated as manual manufacturing.

SDLC v2 (2002–2025). Iterative development with CI/CD. Development and testing are part of iterative engineering design. Repeatable automation handles deployment. Use what you learn to revise the design.

SDLC v3 (2026+). AI-assisted iterative development. The same six stages appear in two lanes. AI automates repeatable engineering work in the pipeline and collaborates in the iterative design process.

What can we learn from this for 2027?

  1. Automation: Move from CI/CD to CE

    CE is Continuous Engineering.

    Use AI to automate repeatable work throughout the lifecycle that CI/CD alone could not automate before.

  2. Design: Use the extra room to iterate

    AI makes it easier to try, test and revise a design. As in earlier versions of the SDLC, a focus on automation and handovers can obscure the design loop.

  3. Tooling: Build end-to-end pipelines

    Today’s tooling is split between agentic development environments (ADEs) and traditional CI/CD tools. I’d like to see them joined in one pipeline that carries a change from specification through delivery in a disciplined way.

Download the PowerPoint slides