Executive Council

Pressure-test the decision before you commit.

Five AI advisory seats take one real decision of yours, argue it out with each other, and hand back a single recommendation with the disagreement still attached.

✓Three free, no signup: one question, one follow-up, one challenge ✓After that, your own key and no markup on usage ✓The disagreement is shown, not smoothed
Try it free, no signup See how it works
Deciding: raise prices 20 percent for new customers only
Deliberation
five seats, one decision, positions after revision
Opportunity BACKS IT
Skeptic OBJECTS
Systems Thinker CONDITIONAL
Operator OBJECTS
First Principles REFRAMES
CHAIRMAN

Raise for new customers only, and hold the increase until you have a clean read on churn.

2 SEATS STILL OBJECT the dissent log carries both, word for word
Skeptic, aimed at Opportunity
Your upside case only holds if churn stays flat after the increase. Nothing in your position supports that.
5seats at
the table
1recommendation
you can act on
3free questions: ask,
follow up, challenge
Zeromarkup on what
the models cost
The problem

One model agrees with you.
That is the failure mode.

Ask one model about a decision you have already framed and it will mostly work inside your framing. The weak part of the idea survives, and you find it after you commit.

One model

AGREEABLE
We are raising prices 20 percent for new customers. Good idea?
Yes, that is a sensible move. Here is a rollout plan and the messaging to go with it. agreed with the framing
You get a plan. Nobody checked the assumption holding it up.

A council

ADVERSARIAL BY DESIGN
We are raising prices 20 percent for new customers. Good idea?
Opportunity Skeptic Systems Thinker Operator First Principles
Objection lodgedTwo seats pushed back. The upside case only works if churn stays flat, and nothing in the room supports that yet.
You get the answer, and the argument behind it.
How it works

Three steps, then a real argument.

1

Bring one real decision

Not a search query. The kind of call you would otherwise sit on for a week, with context about your situation.

What counts as a good question
2

The seats argue it out

Five mandates write their own position, attack an assigned opponent, then revise. A seat can call for evidence when it needs it.

See the protocol
3

Read the answer and the argument

One recommendation, the dissent behind it, and an independent audit of that recommendation against the evidence.

What you get back
The table

Five seats, five mandates.

Every seat reads the same decision and is told to want something different from it. They are roles, not personalities.

Opportunity

Argues for the upside. Optimistic, but it has to hold up.

Looks forLeverage, advantage, room to scale.

Skeptic

Attacks the assumptions. Adversarial on purpose.

Looks forWeak reasoning, contradictions, downside.

Systems Thinker

Follows the second-order effects out past the first move.

Looks forIncentives, knock-ons, long-term dynamics.

Operator

Grounds it in execution. Who actually does this on Monday.

Looks forFriction, cost, feasibility, edge cases.

First Principles

Ignores convention and challenges the framing itself.

Looks forFalse premises, and the one constraint that decides it.

A Chairman and an independent auditor act on what the seats produce. They are not seats, and they never get a vote of their own.

What you get back

An answer, the argument,
and a check on both.

One recommendation

The Chairman reads every revised position and every critique, resolves the contradictions, and writes a single call you can act on. It is told not to smooth over real disagreement.

Claims the evidence does not support
Overconfidence in the recommendation
Reasoning that does not hold up

An independent audit

A separate auditor checks the recommendation against the evidence from the run, not against its own opinions. It writes what it found in plain words.

Dissent log · taken straight from each seat
Dissent log
Built from the written critique each seat produced. It does not pass through the step that writes the summary, so the disagreement reaches you the way the seat put it.
Skeptic OBJECTION Weakest assumptionThat churn holds flat through a 20 percent increase, with nothing in the position to back it. Missing variableWhat the last two price moves did to renewals.
Operator OBJECTION Weakest assumptionThat sales can hold the new number without a discount policy written down first. Missing variableHow many deals in the pipeline were quoted at the old price.
Systems Thinker CAVEAT Weakest assumptionThat existing customers never see the new price. They talk to each other. Missing variableThe support load if they ask to be moved onto it.
First Principles REFRAME Weakest assumptionThat the question is the size of the increase. The binding constraint is what you are worth to a new buyer. Missing variableThe one number nobody in the room has: what a new buyer compares you against.

Opportunity had nothing to push back on here, so it is not in the log. A seat with no objection is left out rather than padded in.

Honest limits

Worth the argument, or not.

A council is slower than one model, and it costs more per question. Here is when that trade is worth making, and when it is not.

Where a council earns its keep

  • A pricing change with several stakeholders pulling different ways.
  • A counter-offer you only get one shot at.
  • A hiring call where the case for and against are both real.
  • Anything where being wrong is expensive and slow to undo.

Where one model is the better tool

  • A simple factual question with one right answer.
  • Anything you need back in ten seconds.
  • Drafting and rewriting, where speed matters more than scrutiny.
  • For those, one model is faster and cheaper, and we would rather say so than take the run.
The audit is a diagnostic, not a gate.

It has missed things before. It is one more pair of eyes on the recommendation, and we would rather tell you its limits than sell it as a guarantee.

The dissent log is raw on purpose.

Because it is lifted straight from each seat's critique, a line occasionally comes through formatted oddly. We would rather show you the unsmoothed version than clean it up in the summary.

Your account, your bill

You pay the model provider directly. We add no markup on what the models cost.

You open your own account with the model provider and put your own credit into it. That account is billed for exactly what a deliberation uses. We never handle the payment, and we add no fee on top of what the provider charges.

Your key is encrypted on our server the moment you save it.
It is never shown back to your browser, not even to you.
You can delete it in one click, any time.
You set the spending limit on your own account, and nothing can go past it.
Your own accountyour credit, your key
billed for what a run uses
The model providercharges your account
Executive Councilis never in the payment path, and we add no fee on top of your bill.
Questions

The things people ask first.

What does Executive Council actually do?

It gives one decision of yours to five AI advisory seats that deliberate on it, argue with each other, and then hand you a single recommendation along with the disagreement and an independent audit of that recommendation. You bring a real decision, not a search query.

How is this different from just asking one AI?

One model tends to agree with the way you framed the question, and that agreement is the failure mode this is built to remove. Five seats with different mandates attack the reasoning in the other positions, so the weak parts of an idea surface before you commit rather than after.

Why show me the disagreement instead of just the answer?

Because the disagreement is usually the most useful part of the whole exercise. The dissent log is built straight from the written critique each seat produced, so it is not quietly smoothed over by the step that writes the final summary.

What does it cost?

We add no markup on model usage. You bring your own model access key and pay the model provider directly for exactly what a deliberation consumes, which in practice is typically under a dollar for a full run.

Can I try it before signing up?

Yes, you can ask three real questions with no account and no key of your own, funded by us. Nothing about it is simulated: it is the same five seats and the same protocol you would get as a signed-up user.

What kind of decisions is it good for?

It earns its keep on high-stakes decisions with several stakeholders and real trade-offs, like a pricing change, a counter-offer, or a difficult hiring call. For a simple factual question a single model is faster and cheaper, and we would rather tell you that than take the run.

Are the answers made by AI?

Yes, every seat and the final synthesis are AI generated. Read the deliberation and the dissent, then apply your own judgement before you act on it.

Bring one real decision

Three free questions. No account, no key of your own.

The three are one opening question, one follow-up, and one challenge to the answer you get. Funded by us, and nothing about it is simulated. It is the same five seats and the same protocol you would get as a signed-up user, and the deliberation comes with you if you make an account afterwards.

Try it free, no signup

This is AI, and we say so. Every seat, the synthesis, and the audit are AI generated. Read the deliberation and the dissent, then apply your own judgement before you act on it. How Executive Council handles your data