Spec-driven AI Software Development: Kiro or BMAD?

AI coding assistants are great at typing. They’re less great at knowing what to type.

AI coding assistants are great at typing. They’re less great at knowing what to type.

That’s why spec-driven development is having a moment: define requirements and design first, then let AI implement against a clear source of truth. Two names keep showing up in that conversation - Amazon Kiro and the open-source BMAD Method - and they’re often compared as if they were the same thing.

They’re not.

  • Kiro is an AI IDE with Spec-Driven Development built into the product.

  • BMAD is a portable agile methodology that runs on the AI tools you already use.

Same idea (intent before code). Different packaging (productized IDE vs open workflow).

In this week’s post I break down:

  • What each actually is (and isn’t)

  • A side-by-side comparison (lock-in, cost, AWS affinity, team fit)

  • When to choose Kiro, when to choose BMAD, and when to use neither

  • A simple decision tree for teams

  • Why “methodology cost,” “model tokens,” and “cloud runtime” are three different bills

The punchline: the winning move isn’t collecting frameworks. It’s making requirements and design legible before the model starts typing.

If you’re building on AWS with a single-IDE strategy, you’ll lean one way.
If you want portable, role-based planning across tools, you’ll lean the other.
If the change is small and clear — keep the process light.