Library skill · Thinkers' principles

Bezos-style one-way or two-way door check

v1.0.1 · Updated 4 November 2026

Sorts pending decisions into one-way and two-way doors, matches each to a careful or a quick process, and checks the information level, disagree and commit, and what needs escalating. Use it when A decision has been stuck in discussion for weeks.

Not affiliated with or endorsed by Jeff Bezos or their companies. About these principles: Jeff Bezos's ideas, explained.

Add to my team's pod

Works in Claude, ChatGPT, and coding agents. On Team and Business, this library skill in your pod updates itself when we improve it.

When to use it

A decision has been stuck in discussion for weeks
Every change, small or large, goes through the same approval
Two teams disagree and nobody decides

What you give it

The pending decisions, one sentence each
What it would take to undo each one
What is known and what is missing
Who decides now, and any disagreement

What you get

A decision table with undo cost, door and the process each needs, the mismatches, an information check, disagreements sorted into disagree-and-commit or escalation, and next actions with deciders.

How it works

List the decisions One line each. Split a bundled decision (for example, launching in a new country with a new pricing model) into its parts; the parts may be different doors.
Test reversibility (principles 2, 3) For each: if we do this and don't like the result, can we go back? What would it cost, and how long would it take? Who outside the team would notice?
Test consequence (principle 2) How big is the downside if it goes wrong and we can't fully undo it? A reversible decision with a large, lasting side effect (customer trust, a public commitment) moves toward one-way.
Assign the door and the process (principles 1 to 4) One-way: name the deliberation it needs (a written memo, the people to consult, a date). Two-way: name one person or a small group who decides, and a short deadline. Flag any two-way door currently going through a one-way process, and any one-way door being decided casually.
Check the information level (principle 5) What do we know, what would we wish to know, and would waiting get it? For a two-way door, say how we would notice a wrong call and how fast we could correct it.
Check for disagreement (principles 6, 7) Is it a genuine disagreement about a shared objective? Suggest disagree and commit, with the person who disagrees saying so openly and then committing. Are the objectives themselves different? Then it is misalignment: escalate now, and name to whom.
Write the next action for each decision Who decides, by when, and what would make us revisit it.

Share it with your team

Add it to an AI Skill Pod and everyone on the team, contractors included, uses the same version from their own AI. Read the best ways to share AI skills with a team