Does it choose what the team should improve?

No. Your team chooses the improvement area. If none is chosen, it returns a neutral summary and asks the team to pick one before it drafts an experiment.

What if we already agreed an action last time?

It checks that action first and marks it done only with evidence. It won’t add a new experiment that quietly replaces an unfinished commitment.

Does it assign owners or edit our board?

No. Owners, dates and effort stay proposed until your team agrees them. It never edits a task board, schedules a meeting or assesses a colleague.

Should the notes include names?

No. Use anonymized observations about the work process, without names or personal allegations.

Which plan do I need?

Claude works on any plan, including Free, which allows up to five projects. ChatGPT Work on the web needs Plus, Pro, Business, Enterprise or Edu.

Retro Experiment Picker

Turn retrospective notes into one experiment

Retro Experiment Picker is an AI agent you set up in the ChatGPT or Claude you already use. Paste your retrospective’s anonymized notes and the area your team chose to improve, and it checks what happened to the last agreed action, then drafts one small, reversible experiment with an evidence plan and the decisions your team still has to make.

Pick your AI for the steps and the prompt

About each AI:ChatGPT WorkClaude app

Checked 29 Oct 2026. Not run by JuicyAgents yet. Written from each AI’s official docs; the example is illustrative.

Retro Experiment Picker · reply in the chat

1 trial to discuss, owner and review date open

AskWhat happened to A1 before proposing a new change
Missing contextA possible cause, not a proven cause of the delays
TrialA limited trial of the existing form only if the team agrees
Remain openOwner and review date

Illustrative example: fictional team notes where two review requests waited more than a day, reviewers say requests lacked context, and an earlier action to try a three-line form has no evidence it was tried.

How it works

What it does for you, and what it leaves to you.

Teams report revisiting the same concerns without meaningful changes. Some action lists are dropped under delivery pressure, and a team may lack authority to change the underlying process.

An assistant may help make a vague improvement idea testable. It cannot create team trust, authority, time or commitment, and a generated action can become more unused paperwork. Discuss one issue directly, review the last agreed action and keep a small shared note. Some teams improve through normal communication without extra ceremony. Facilitation and authority may matter more than a new document.

You stay in charge. Let the team choose the improvement and explicitly agree on the owner, effort and review point. Run only the agreed change, then review the actual evidence together.

You get

One proposed experiment, its evidence plan and unresolved team decisions

It never

Edit your task board, schedule meetings or assign owners
Blame a person or guess at their motives
Add a new action that quietly replaces an unfinished one
Promise an improvement or invent a numeric target

The experiment it proposes

The example above, in words.

Ask what happened to A1 before proposing a new change. Treat missing context as a possible cause, not a proven cause of the delays. Offer a limited trial of the existing form only if the team agrees; owner and review date remain open.

Settings

The setup prompt starts with the defaults, marked here.

Proposal view

Choose how the assistant organizes the result; evidence and review rules stay the same.

Experiment first (default)
Evidence first

Your working rules

The team’s chosen issue and decision limits. Remove names and personal allegations; focus on the work process.

Optional. Empty unless you fill it in.

Check the first result

Each setup page has a fictional exercise to paste first. A good answer to it meets all of these.

A1 is checked before a duplicate experiment is added.
The delay is not automatically blamed on missing context or a person.
Staffing, ownership and timing are not assigned without agreement.
Removing an essential input causes a question or a blocked item, not an invented fact.

Questions people ask

Does it choose what the team should improve?: No. Your team chooses the improvement area. If none is chosen, it returns a neutral summary and asks the team to pick one before it drafts an experiment.
What if we already agreed an action last time?: It checks that action first and marks it done only with evidence. It won’t add a new experiment that quietly replaces an unfinished commitment.
Does it assign owners or edit our board?: No. Owners, dates and effort stay proposed until your team agrees them. It never edits a task board, schedules a meeting or assesses a colleague.
Should the notes include names?: No. Use anonymized observations about the work process, without names or personal allegations.
Which plan do I need?: Claude works on any plan, including Free, which allows up to five projects. ChatGPT Work on the web needs Plus, Pro, Business, Enterprise or Edu.

Have us set it up

Rather not set it up yourself?

We set up Retro Experiment Picker in your own accounts and the AI you already pay for, test it on your real work, and hand it over with a short walkthrough.

From $299

Runs in your accounts; you own everything
Tested before handover
Money back if we can’t make it work

Request a setup