The Agentic SDLC: Building Apps via Structured AI Stages

Description and use

The author demonstrates how AI agents can follow the Software Development Lifecycle (SDLC) by breaking down application creation into five stages: requirements, design, implementation, testing, and deployment. Each stage produces an artifact for the next, resulting in a working word-frequency web app.

Problem

AI coding agents are good at writing code from single prompts but struggle to follow a structured software development process, leading to potential lack of rigor in requirements, design, and testing.

Outcome

A functional word-frequency web application was successfully built by guiding an AI agent through a structured five-stage SDLC process, demonstrating that AI can follow engineering processes when prompted sequentially.

Workflow

• Define Requirements: Generate a requirements document for the application.
• Design: Create a design specification based on the requirements.
• Implementation: Write the code based on the design specification.
• Testing: Generate and run tests based on the implementation.
• Deployment: Prepare the application for deployment.

Evidence

Here is a simple word-frequency web application built using five separate stages: Requirements → Design → Implementation → Testing → Deployment

Source

Everything described in the video is on this GitHub repo

Source