Teaching Amazon’s ‘Working Backwards’ to ML product teams

What it does

ML Enablement Workshop is a structured curriculum—delivered as markdown guides and Jupyter Notebook examples—for running a facilitated workshop. It is designed to break down silos between business planners and developers by having them jointly practice Amazon’s “Working Backwards” methodology. Teams draft press releases and build lightweight functional mocks with generative AI, then refine their plans based on feedback.

The interesting bit

The workshop treats generative AI not as a final product but as a facilitation tool: participants use it to parallelize hypothesis testing, draft specs in Kiro, and spin up interactive mocks in roughly ten to twenty minutes. The goal is to compress the loop from press release to measurable prototype so teams can outline a concrete 3–6 month release trajectory by the end of the second session.

Key highlights

• Three-phase format: a one-hour Day 0 for alignment, a 3.5-hour hands-on “practice” session, and a three-hour “improvement” session for self-directed refinement.
• Heavy emphasis on cross-functional collaboration; case studies from MUFG, JINS, BASE, and others cite it as a way to align sales, product, and engineering on concrete milestones.
• Includes a fully AI-generated example of a Working Backwards press release and a mock EC-site feature with dummy-response backends and built-in metrics collection.
• Materials are in Japanese and assume an organizational context where business and engineering are separate units that rarely co-design.

Caveats

• This is a methodology repository, not a software library; the “code” is largely example notebooks and mock templates that you will need to adapt heavily.
• The six-month release target and cited customer metrics reflect workshop goals and reported outcomes, not guaranteed results.
• All documentation and case studies are Japanese-only, so non-Japanese speakers will need translation support.

Verdict

Product managers, engineering leads, and ML specialists in Japanese-speaking organizations should look here if they need a battle-tested facilitation guide for aligning siloed teams around generative-AI features. Purely technical developers looking for reusable model code or infrastructure templates will find only reference mocks and process documentation.