Five rival AI minds. One pressure-tested answer.

Executive Council puts your hardest decision in front of five adversarial AI seats, makes them argue, then has an independent auditor check the result before you see it.

See how it works

The protocol

Built to disagree with itself

You brief the council

One hard question, with your context. The higher the stakes, the better this works.

Five seats argue it out

Opportunity, Skeptic, Systems, Operator and First Principles each take a position, then attack each other's reasoning in focused duels and revise.

The Chairman synthesises

One clear recommendation, built from the surviving arguments rather than a polite average.

An independent auditor checks it

A separate evaluator ground-checks the claims and flags anything the council cannot back up. Disagreements are preserved in a Dissent Log, not smoothed over.

Why a council

One model cannot disagree with itself

Ask one AI to play five personas and you get five voices with one set of blind spots. Executive Council seats genuinely different frontier models from rival labs, so the disagreements are real and the weak arguments get found.

No single AI lab will ever build this. None of them will route your question to a competitor and report back that the competitor was right.

What you keep

A decision record, not a chat log

  • The recommendation, with its reasoning shown
  • The auditor's notes on what held up and what did not
  • The Dissent Log: who disagreed and why
  • A DOCX export you can put in front of your board

Coming: Confidential Mode. Deliberations that never touch our database.

$9 a decision

No subscription. Pay for the decisions you actually run.

See pricing

Questions

What does it cost?

A full deliberation is $9, or $29 for a pack of four. No subscription. See the pricing page for what a deliberation includes.

Where do the answers come from?

Each seat runs on a different frontier AI model, chosen so the seats genuinely think differently. One model arguing with itself shares one set of blind spots; rivals do not.

Is my data used to train models?

We never train models on your data. What upstream providers do is governed by their policies; the privacy page spells out exactly who touches what.

When can I buy?

Payments open shortly. Leave your email and we will tell you the day they do.