Hire an AI team decision review planner

Everything it can hand back

Your task sets the actual outputs. The specialist picks the methods that cover them.

• The choice sorted as a one-way or two-way door
• A verdict on whether math or judgment settles it
• A weighted evidence bar, a decide-now or wait call and a date to stop collecting facts
• A note with each dissent on record, the commit lines and the review date

How it does the job well

• Pin down the exact question, scope and the standard of evidence needed before searching.
• Prefer primary sources, and say how strong each source is.
• Separate what is established, debated and unknown, with confidence levels.
• Cite specifically: author, work, date or page where possible.

Mistakes it avoids

• Presenting one source or a popular claim as settled.
• Inventing citations, quotes or page numbers.
• Over-generalising across places, periods or groups.

What a good result looks like

An example, not your task: A decision has been argued in several meetings with no end.

• Step 1: Door. The choice sorted as a one-way or two-way door, with the reason and the process that fits.
• Step 2: Math or judgment. Whether your own numbers can settle it, or a written judgment with the math against it.
• Step 3: Evidence bar. Your share of the facts you wish you had, the bar, a decide-now or wait call, and the one fact to get first.
• Step 4: Disagree-and-commit note. The decision, owner and review date, each dissent on record, and the commit line for each person.
• Hand-back summary. Each step's file, any step not run, and every MISSING line in one list.

How the work is checked

• Each step passed its own checks, or the run stopped at the step that failed and says why.
• The steps ran in order, and each one used the result of the one before.
• Nothing was invented to fill a gap. Every gap is a MISSING: line.
• Nothing was sent, published or paid for without a clear yes.
• States which checks passed, failed or could not run, and names missing inputs.

These checks have not been run against a model yet. Your AI uses them to check its own work.

A second opinion

Before it hands back, your AI reviews the draft the way Jeff Bezos would. This is a way of reviewing the work. It does not mean they wrote or endorse it.

Capabilities

• Sort the decision by reversibility. Asks what would be hard to undo. Sorts the choice as a one-way or two-way door and picks the process that fits.
• Decide whether math or judgment settles it. Tests if your own history can settle it. If not, writes the judgment with the short-term math against it and what the math cannot see.
• Set the evidence bar. Weights the facts you wish you had and works out your share. Sets the bar, compares wrong with slow and gives a call date.
• Write the disagree-and-commit note. Takes the decision and each person's view, which you give it, plus the plan from the steps before. Records each dissent in one fair line and sorts it as view or aims. Escalates a clash of aims and writes the commit lines and team note.

How the role works

1. Checks your task, context and available tools.
2. Picks the methods needed and prepares a work plan.
3. Carries out supported tasks, with each method's limits and review gates.
4. Checks the real deliverables and corrects supported faults.
5. Hands back files, evidence and unresolved gaps.

Your specialist's working setup

Download the base specialist without a task form or account. README.md and SETUP.md explain how to use the folder. SOUL.md defines its scope. INTAKE.md asks missing questions inside your AI. PLAYBOOK.md, SKILLS.md, BEST_PRACTICES.md, CHECKS.md and HANDOFF.md guide the work and its checked deliverables.

Complete associated skills, supporting files and licences are included. Choose only the methods needed for the actual task. Keep context, working files and accepted feedback in your AI project.

Use this specialist through MCP or the API

Call prepare_specialist with role_id fast-decision-evidence only. No task is required to load it. Your AI asks what job you need done, what you want back and which relevant context it can use.

For task-led hiring, call find_specialist, explain the suggested specialists and let the user accept or change the team.

Complete ZIP: Download base specialist. API: GET /api/v1/specialists/fast-decision-evidence/pack. The host AI supplies its own model, tools and execution; a pack does not connect accounts or run a task.

Inputs and helpful context

The specialist starts from the task brief. Only active methods require their inputs.

For the primary method, when used

• The choice, and who it touches
• The numbers and history you have
• What each person thinks and why, in a line or two
• Who owns the decision, and any deadline

Helpful context (optional)

• Meeting notes where the views were argued
• Cost of delay per week
• Results of similar past decisions

Tools and access

• Any good chat AI: Holds the sort, the bar and the views across the steps.
• A shared document: The team can read the note and add their commit line.
• Any good chat AI: Sorting a choice and writing the process is text work. Claude, ChatGPT and Gemini all handle it with this skill.
• A shared document: The written call and dissent stay where the whole team can see them.
• Any good chat AI: Weighing tests and writing a judgment is text work any strong chat AI does well with this skill.
• Any good chat AI: Weighting facts and doing the share sum is text work any strong chat AI does well with this skill.
• Any good chat AI: Writing fair summaries of views and a short team note is text work any strong chat AI does well with this skill.

Functions and deliverables

• Team decision review planner: Check how hard a decision is to undo, set the evidence needed and draft a team note that records disagreement.
  • The choice sorted as a one-way or two-way door
  • A verdict on whether math or judgment settles it
  • A weighted evidence bar, a decide-now or wait call and a date to stop collecting facts
  • A note with each dissent on record, the commit lines and the review date
• Reversible decision sorter: Sort a stuck decision as a one-way or two-way door. Get the review it needs, plus an owner, a small trial and a review date.
  • The choice sorted as a one-way or two-way door, with the reason
  • The cost of reversing it, in time, money and trust
  • The review that fits: wide and slow, or small and fast
  • For a two-way door, an owner, a small trial and a review date
  • For a one-way door, what to learn first, and whether a smaller step would make it two-way
  • A written call with any dissent on record
• Decision method planner: Sort a stuck decision into math or judgment call. Paste your options and numbers to see if your own data can settle it, and what to do next.
  • The decision sorted as math or judgment, with the three-test result
  • For a math call: the model set-up and the input that would flip the answer
  • For a judgment call: a written judgment, the math against it and what the math cannot see
  • A review date and a signal
• Decision evidence planner: Find out how much evidence is enough before you decide. Paste your decision and what you know. Get a decide now or wait call and a date.
  • A weighted list of facts, marked known, partly known or unknown
  • Your information share as a percent
  • The bar for this decision
  • A decide now or wait verdict with the reasoning
  • The one fact to get first and a call date
• Team decision and dissent note writer: Share a decision and each person’s reasons. Record doubts fairly, name the decision owner and draft actions, signals and a review date.
  • One fair line of dissent per person, including the boss if they disagree
  • Each dissent sorted as a difference of view or a clash of aims
  • A named decider, a time limit and what each side brings, if the aims clash
  • A commit line each person says back, with an action in it
  • A signal that would show the decision is wrong, and a review date
  • A short note for the team
• Research paper and code note writer: Share a paper, repo or messy notes. Get claims, methods and evidence sorted into a clear research folder.
  • A research folder with claims, methods and results
  • Every claim linked to its supporting table or log
  • Results tables copied with exact numbers
  • A record of dead ends and decisions
  • A short report on what was filled in and what was missing
  Source: Orchestra Research.
• Jeff Bezos: Review a product or stuck decision through Jeff Bezos’s recorded approach. Start with the customer, test what can be undone and write the reasoning.
  • Your problem worked on the way Jeff Bezos would: their questions, their method, their push-back
  • The answer or the work, in Jeff Bezos's voice, using their ideas by name
  • Anything that goes beyond their public record, marked
  • One suggested next step

Tasks you can use it for

• A decision has been argued in several meetings with no end.
• The team keeps asking for more data before it will act.
• Someone keeps reopening a call after it was made, or you disagree with a call as the boss.

Limits to check before starting

• The choice is already set by policy or law: follow the policy, or ask the person who owns it.
• The choice is purely technical, like picking a code library: use a normal project plan with a short test.
• You want to weigh whose opinion counts most in a group: use the Group decision reviewer.

Best way to use this role

Give it one clear outcome, the actual files and a definition of done. Keep the work in one AI project so it can reuse checked context and earlier files.

• Download the handoff pack. It's a zip with HANDOFF.md plus one skill file for each step.
• Give the folder to Claude (in Claude Code, or as files in a Claude Project) and say "Follow HANDOFF.md".
• Paste the choice, your numbers and what each person thinks. It works step by step.
• Send the team note. Act on the call and look again on the review date.

Read it as data

Read the complete role setup as JSON. Connect your AI to the AI Blocks MCP library and read aiblocks://roles/fast-decision-evidence for its role instructions and pinned sources.

This is a prepared AI role. No task has run. Your AI carries out authorized work using its available tools and reports the actual deliverables and checks.