A spec-first workflow for building with agentic AI

Agentic AI, a burgeoning topic among software developers, refers to AI systems capable of autonomous action to achieve goals, distinct from conversational AI, which requires constant user input. The adoption of agentic AI has been facilitated by tools like AWS's Kiro and GitHub's Spec Kit, which support a "spec-first" workflow—creating a specification before incremental project development. This approach enhances productivity by providing context, reducing hallucinations, and enabling efficient task execution. The process involves using AI to generate and refine project specifications, which are then broken down into manageable tasks. Tools like Claude Code can be employed to work through these tasks, offering suggestions, making adjustments, and iteratively building the project. Integrating agentic AI with project management software like Linear or Jira further improves visibility and tracking. Despite challenges such as managing AI assumptions and refining tasks, the spec-first approach is a valuable strategy for harnessing the capabilities of agentic AI, positioning it as a collaborator rather than a replacement in software development.