The Casting Standard. Five Voices Every Big Decision Needs.

When a decision matters now, I take it to a panel. Five world-class minds, cast fresh for that one question, each writing an independent brief on my situation, and at the end a synthesis that commits to a recommendation.

I built the thing in April. It is a saved routine in my AI workspace (a skill, in Claude terms) that I call the Think Tank, and it does one job. I bring it a question I need to get right. It proposes five minds fit for that specific question, has each of them write separately, then pulls the threads together and tells me what it would do. Since then it has sat with me on pricing, on positioning, on the go-to-market plan for a cohort launch, and on a piece of copy that read fine and felt wrong.

The first time I ran it, I was half testing the plumbing. I handed it a launch schedule I had already committed to and asked how to fill the first event. Two of the minds, writing separately with no sight of each other's briefs, came back with the same finding in different words. The problem was structural. The first event sat three days after the session that was supposed to sell it, with no runway, and the whole plan had been built to fill the second event. I could not see that from the inside, because I had asked a promotion question about what turned out to be a scheduling problem.

One voice answers the question you asked. A well-cast panel catches the question you should have asked. That difference is the whole tool.

Who gets a seat

The instinct, when you first try something like this, is to pick five people you admire and ask what they would do. That gives you an echo with famous names on it. You chose those people because they already think the way you think, or the way you wish you thought, and their counsel lands soft.

The panel works when the seats are structural. Every cast has to cover five angles. I hold that rule hard enough that I gave it a name, the Casting Standard.

• Critical. A voice that pokes holes and finds the weak seam. Without it, the panel just applauds.
• Domain. A voice that is world-class in the field the question sits in. Skip it and the counsel floats free of how that field works.
• Lateral. A voice from an adjacent field who comes at the question sideways. Otherwise you get your own industry's consensus back, and you already had that.
• Pressure. A voice that stress tests what breaks under load, at scale, on a bad week. Leave it out and the plan holds right up until it meets reality.
• Aligned. A voice that shares your values and protects the thing you are building. Lose this seat and you can win the metric while losing the point.

One mind can hold two angles, and the cast changes with every question. A pricing question calls operators to the table. A question about a book chapter calls writers. The minds are real, named people, never archetypes, because "a marketing expert" flattens into mush, while a specific person carries a worldview and a set of scars the AI can hold onto.

Steal the standard even if you never use it with AI. Any group you gather around a live decision, a mastermind, a peer circle, a kitchen-table council of two, should cover the same five angles, and most cover two.

It has to end in a recommendation

Ask AI for advice and you usually get a menu. Three options, pros and cons for each, "it depends on your goals." It reads as balance, and it hands the hardest part of the work straight back to you while appearing to help.

The Think Tank is built to commit. The synthesis names where the minds converged, where they split and what the split means, the sharpest single insight of the sitting, and then a decision in plain words with the reasons attached.

I stay free to overrule it, and I have. Sovereignty stays intact because I remain the one who decides, and a committed recommendation gives me something solid to decide against. A panel that hedges gives me nothing to push on, so my own thinking never sharpens.

The panel is only as sharp as what it can see

There is a catch, and it is the same one I wrote about in The Dark Parts of the Map. Hand any version of this three lines about your business and you will get generic counsel with famous names attached. The panel sharpens in proportion to what it can see of your situation.

The version I run reads my written business before anyone speaks. Who I serve, what I charge, what is live this week, what I decided recently and why. That written base is the Human Stack, and it is the reason the same five minds who would hand a stranger general principles hand me counsel aimed at this week's decision. This is what I mean by a Wisdom Amplifier, the third of the Three Levels of AI Partnership, the level where AI works inside your world and extends your own thinking instead of replacing it with someone else's.

You can start with far less. One honest page is enough, and if the panel earns its keep and you want to build the written base it deserves, the setup takes about thirty minutes.

Convene your own

The version below runs in any capable AI chat, with no setup and no files, in one sitting. The work is in the situation brief, so give it the numbers, the constraints, and the thing you have been half avoiding. The panel can only work with what you put on the table.

You are the chair of my Think Tank. One question, one sitting, one
clear recommendation at the end.

MY SITUATION
[Three short paragraphs. What you do and who you serve. The decision
on the table, with real numbers, dates, and constraints. What you have
already tried, and what you are assuming.]

MY QUESTION
[One sentence. The sharper the question, the sharper the counsel.]

STEP 1. CAST THE PANEL
Choose five real, named, world-class minds for this specific question.
Between them they must cover five angles:
1. Critical: pokes holes, finds the weak seam, disagrees when warranted.
2. Domain: world-class in the field my question sits in.
3. Lateral: works in an adjacent field and comes at this sideways.
4. Pressure: stress tests what breaks under load, at scale, on a bad week.
5. Aligned: shares my values and protects the thing I am building.
One mind can cover two angles. No generic archetypes. Show me the cast
first, one line per mind naming their angle.

STEP 2. HEAR THE BRIEFS
Each mind writes an independent brief of roughly 200 words, in their
own voice, grounded in my situation rather than a generic version of
it, ending with a one-sentence recommendation. No mind sees another
mind's brief.

STEP 3. SYNTHESIZE
Show me where the minds converge, where they split and what the split
means, the single sharpest insight of the sitting with a name on it,
and what this panel cannot see from where it sits, the part I have to
weigh myself.

STEP 4. THE RECOMMENDATION
Commit to one clear recommendation in two or three sentences. Do not
hedge. Do not hand me a menu. I am free to overrule you. That is my
job, not yours.

Read the briefs slowly when they come back. The value is rarely inside any single one. It sits in the tension between two minds who want different things for you, both of them right.

The seat that stays yours

In four months the Think Tank has never made a decision for me, and it has changed several. A launch got re-sequenced. Two positioning decisions landed somewhere I would not have found alone, and a campaign plan went out the door stronger than it came in.

That is the shape I want AI to take in my work and in yours. Your ultimate panel of world-class minds, on call, cast fresh for every question that deserves one, willing to commit to a recommendation, and structurally incapable of taking the one seat that matters. The deciding seat stays yours.

The next time something keeps you turning at night, take it to a properly cast panel. Cover the five angles, ask for a recommendation with a spine, and then decide.