Difficult Conversation Rehearsal — by an AI council
Rehearse a hard conversation so you go in calm and clear. Four rival AIs debate it and hand back: A plan for the conversation: your opener, the hard part, and how to respond.
A plan for the conversation: your opener, the hard part, and how to respond.
Signed off by an independent Final QA audit — with a “verify before you rely on this” list, never a black box.
Why this team
Hard conversations go wrong in the first thirty seconds and the last five minutes — the opening that puts the other person on trial, and the ending that trades the message away for relief. The Negotiation Expert choreographs both, the Steelman argues the other person’s side so well that your position improves before you speak, and the Brutally Honest Friend checks that what you plan to say is actually the thing you need said. The Executive Coach handles the part scripts cannot: what you do when the emotion arrives anyway.
Each seat runs on a different frontier model — GPT, Claude, Gemini, Grok — so it’s a genuine cross-model review, not one AI in costumes. How a run works →
What the debate sounds like
An illustrative excerpt — A co-founder needs to tell his business partner and oldest friend that the partner’s spending through the company account has to stop — without ending the friendship or the company.
His side first, at full strength: the “spending” is mostly client dinners and travel he genuinely believes builds the business; nobody has ever defined a policy; and the company was founded on not treating each other like employees. From inside his head, a sudden confrontation about receipts is a change in the relationship, not a correction of behaviour. Any opening that ignores this history will be heard as an accusation of theft — because without the context, that is what it sounds like.
And the message you are avoiding, stated plainly: it is not really about the dinners — you no longer feel you can see the money clearly, and that is corroding trust weekly. If you script a conversation about expense categories, you will win the category argument and keep the corrosion. Say the real thing: “I need us to both see every number, because I can feel myself starting to audit you in my head, and I hate it.”
Structure it as a system fix, not a verdict on character: open with the friendship as the stake, not the accusation — “this partnership matters too much for me to let something small grow teeth”. Then one concrete proposal: shared visibility and a joint sign-off threshold, applying to both of you symmetrically. Symmetry is the whole trick; a rule for him is an indictment, a rule for both is governance.
Questions
Does the council write me an actual script?
It gives you the opener verbatim, the core message in your own reviewed words, and the exact responses to the two or three most likely hard moments — plus the one thing not to say. It stops short of a full script deliberately: over-scripted conversations sound like depositions, and the structure is what holds when the words wobble.
What makes rehearsing with a council better than with a friend?
Your friends share your side by default — the Steelman seat argues the other person’s position at strength your friends will not, which is where the preparation actually happens. Discovering the strongest version of their reply before the conversation is the difference between responding and reacting.
What if the conversation is overdue and the resentment has built up?
Say so in the brief — accumulated resentment changes the design: the council will usually split one giant conversation into a first conversation with a narrow, winnable goal, because ten grievances delivered at once guarantees defence and settles nothing. The backlog gets sequenced, not dumped.
Your material is used only to run your review — never to train public models. Encrypted in transit and at rest. Security & privacy →
Want full control — pick your own minds, set the depth? Open the full council →