I recently read Anthropic’s AI-native SDLC playbook. One thought stayed with me: as implementation speeds up, the weight of the process shifts elsewhere.
Writing the code itself is becoming less of a bottleneck. What matters more is identifying the right problem, gathering requirements, and expressing them in a form that other people and AI agents can work with.
Anthropic proposes starting with a conversation with AI and creating an intent.md file. The person raising the request describes their need in their own words, AI helps clarify it, and the Product Owner reviews the result.
I wonder, though, how to make this process accessible to people who don’t use Claude Code, Codex, or repositories.
That’s why I’m experimenting with a simple interface for gathering requirements. The user describes a problem, and AI guides them through a series of questions to produce a structured intent. Product context, templates, versioning, and repository integration stay in the background.
The user shouldn’t need to know the whole procedure. The system should know it and guide them through the process.