A virtual board of directors: getting six opinions on one decision
The head of a small company has no board where the owner looks at money, HR at people and the lawyer at risk — so decisions get made from one point of view. Below is a technique that replaces it, and an explanation of why the value lies in the disagreements.
The head of a small company has no board where the owner looks at money, HR at people and the lawyer at risk — so decisions get made from one point of view. Below is a technique that replaces it, and an explanation of why the value lies in the disagreements.
A decision with nobody to discuss it with
The head of a small company, or a freelancer working alone, has a problem with nothing to close it: nobody to discuss a decision with.
Not because there are no acquaintances. Because a proper review needs people with different points of view — and they either see the situation exactly as you do or know nothing about it.
In a large company a board plays that role: the owner looks at money, the operations director at manageability, HR at people, the lawyer at risk. Each sees their own part, and the decision comes out in the round.
You do not have that. So you make decisions from one point of view — yours — and discover what was missed after the fact.
The technique that replaces it is below. Nobody in this niche writes about it at all.
How a virtual board of directors works
The same question is put to several assistants with different roles.
Not "give me different opinions" to one assistant — that produces a polite list. But a separate run with each role, because each has its own frame, its own criteria and its own set of questions for you.
The difference is fundamental. A legal assistant will ask about the contract and liability. An analyst about figures and hypotheses. An HR one about people and resistance. None will ask the others' questions.
Six positions worth covering:
The owner — what about the money, will it pay back, where is the risk of loss.
The operational view — how manageable is this, who will do it, what will break.
People — how the team will react, who will sabotage it and why.
Legal — what obligations, where liability sits, what goes in the contract.
Market and client — how this looks from outside, what a buyer will say.
Analytical — which figures support this and which refute it.
The sixth position is particularly useful last: it checks what the first five said.
The order of the runs
To get an analysis rather than six similar texts.
Step 1. State the decision specifically
Not "should I expand" but "should I take on a second contractor for the current volume of work".
A vague question gives vague answers from all six roles — and the review turns into general reflection.
Step 2. Give identical inputs
The same context to each role. Otherwise you are comparing not positions but different understandings of the situation.
Step 3. Run them one at a time
A separate chat for each role. Not in one conversation — otherwise subsequent answers will adjust to the previous ones.
Step 4. Collect the disagreements
This is where the value appears. Not where everybody agrees but where the positions contradict each other.
The owner says "it pays back in two months", HR says "there are no people and there will not be". That is the fork you would not have seen alone.
Step 5. Decide yourself
The assistants do not vote and do not make the decision. They show you the field. The choice is yours, and that matters.
What this technique finds
Three kinds of finding that do not appear in a single review.
A blind spot by role. You think as a contractor — and do not think about the legal side. Or as an entrepreneur — and do not think about who will do the work.
Contradictions in your own plan. A plan looks coherent while viewed from one side. Two different frames quickly find the places where one thing rules out another.
Questions you had not asked yourself. Each role starts with its own clarifying questions. Half the value is in those, before any answers.
A related technique: analysing the failure in advance
Worth mentioning separately, because it works as a pair with this.
Instead of "what are our risks", a different question is asked: six months have passed and the project has failed spectacularly. Why?
The difference is psychological and decisive. The person is no longer speculating — he is explaining something that has already happened. That technique surfaces roughly a third more real threats than the direct question.
And a second effect: doubt is legitimised. Naming a cause of a failure that "has already occurred" is not criticism but completing an assignment.
How to combine it with the board: first six positions give you the field, then the failure analysis tests the chosen decision for strength.
Where a virtual board does not work
Honestly, so as not to overrate it.
It does not replace data. Six opinions on empty inputs give six high-quality reflections about nothing. If you have no figures, get the figures first.
It does not remove responsibility. The decision stays yours, and "the assistant said so" will not do as an explanation.
It is not suitable for small decisions. A run takes time. For "which colour should the button be" it is excessive.
And it requires honest inputs. If you describe the situation as you would like it to be, all six roles will confirm what you had already decided.
How to use this with a client
Here is a non-obvious application, and it has money in it.
A decision review is a service in its own right. Not a consultation but a structured document: six positions, the disagreements between them, the forks, what to check before starting.
How to approach it. You go to a client not with an offer of services but with a review of his own planned move. Three precise observations about his situation work better than any commercial proposal.
What is delivered: positions by role, a map of disagreements, a list of questions with no answers, and a priority — what to check first.
The order of prices. In our bot's feed of 5 986 jobs over four months: a system or a pipeline runs $490–1 460 against a median one-off job of around $60. Analytical work producing a document belongs in the second category.
What the platform offers for a virtual board
Over 200 AI assistants across different categories, and the board's roles are taken straight from there: entrepreneurs, strategic analytics, legal, HR, sales, marketing.
Strategic analytics — twelve assistants, most of them decision-making methods: analysing a failure in advance, devil's advocacy, six thinking hats, design thinking, Cynefin, the Pareto method, blue ocean strategy.
That is not a "clever chat" but a run through a method — you get the outcome of a procedure rather than an opinion.
Refinement modes on most assistants: harsh assessment, focus group, expert review, criteria-based refinement. For reviewing decisions the harsh assessment is the most useful — it finds the weak points in your own reasoning.
If a review like this is needed regularly, set up one request template: six roles, one question, each role's result going into a shared document. You can merge the answers into a final memo in Cerebrum: describe the task in plain words, the price is visible before launch.
Chat access from the Basic plan. Registration is free and opens three days of full Basic.
Where to start today
One action. Take a decision you are currently postponing — the one where it is unclear what is right.
State it in one specific sentence and put it to two assistants with different roles: the one that looks at money and the one that looks at people.
Not six. Two.
Usually the disagreement between those two positions is what explains why the decision is being postponed.
Where the quality of each position comes from
A board of six roles works exactly as well as each role can reason rather than produce the first thing that comes.
The “Advanced prompting” course covers this as sixteen methods in four groups: fundamentals and context · logic and reasoning · multi-step processes · autonomous methods.
The second group is directly about this: how to make a model show its working rather than only the conclusion. In reviewing a decision that is critical: you need to see why a role reached that conclusion, or six opinions become six unverifiable assertions.
And a technique from the Bot architect — a seven-step reasoning chain: Understand → Basics → Break Down → Analyze → Build → Edge Cases → Final Answer.
The penultimate step, edge cases, is usually what the board is convened for: it finds what nobody thought about in the main scenario.
The “Advanced prompting” course — Basic.
A ready board of twelve methods
The "Strategic analytics" category in the assistant catalogue, Basic plan, is exactly a virtual board, except that each member is built on a specific method rather than an invented role:
- Brainstorm AI: Pre-mortem — preventing failures before they appear
- Devil's Advocacy — finding the weak points in an idea in order to strengthen it
- Six Thinking Hats and Six Value Medals — Edward de Bono's methods
- The Pareto method — which 20% of factors give 80% of the result
- Cynefin Framework — classifying the type of situation: it determines which method applies at all
- Blue Ocean — creating market space outside competition
- Design Thinking — solving through the user's needs
- Emergent Thinking — hidden connections in uncertain problems
- Growth Hacking — strategies for rapid growth
- AI strategist — 50+ frameworks in one place
- Universal AI marketer — the market side of the question
The advantage over roles you invent yourself: a method asks questions you would not have reached. A "contrarian" in your prompt will argue to the extent of your imagination — Devil's Advocacy works to a procedure.
Who to add to the board by subject: Business consultant and AI startup adviser for business decisions, Sales analyst for commercial ones, the 13 legal ones for legal questions.
Skills where the board is already assembled
The article describes the mechanics. In the catalogue it is implemented in several ways — and the difference between them matters more than the choice of any one. All nine skills below are advanced: from the Creator plan, or $9 one-off, or 1000 XP.
- A multi-perspective council — four to six archetypes out of twelve: the Architect, the Contrarian, the Empiricist, the Ethicist, the Pragmatist, the Outsider. Each with its own style and its own blind spots. Three phases, and disagreement in the debate round is obligatory
- A council of high intelligence — up to eighteen participants with modes: the full board, a subject-matter trio, a dialogue between two polar figures, an accelerated two-round review
- An expert council — a simpler version: several profiles speak in turn, then a weighted conclusion is issued
- A virtual chat room with experts — each expert is run separately and does not see the others' answers, then a judge reconciles the differences. The room stays active: the next question goes to the same group
- A committee of two agents — deliberately contrasting reasoning styles, and if they diverge the arguments are passed between them until they agree
- A meeting simulation — five stages from opening statements to minutes with risks and a plan
- A war council before a decision — participants answer without seeing each other's answers, and that matters: otherwise the first opinion sets the tone for the rest
- An independent adviser — if a council is excessive: one second opinion, preferably from a model of a different family
- A cross-check by a second agent — the checker sees only what is needed for the check and does not rewrite the result
The main difference between them. Whether the participants see each other's answers. Where they do, you get agreement around the first opinion expressed. Where they do not, you get the genuine spread the council is convened for.
What to take on the platform for this task
- Brainstorm AI: Pre-mortem — an assistant: analysing a failure in advance; from the Basic plan
- Brainstorm AI: Devil's Advocacy — an assistant: finding weak spots in an idea; from the Basic plan
- Brainstorm AI: 6 thinking hats — an assistant: Edward de Bono's method; from the Basic plan
- Business consultant — an assistant: scaling strategy and crisis management; from the Basic plan
- AI strategist — an assistant: 50+ frameworks for marketing and product analytics; from the Basic plan
- Advanced prompting — a training course: sixteen methods, and how to make a role show its reasoning; Basic plan
What to read next
A pre-mortem: analysing failure in advance — the paired technique for testing a decision.
One assistant against three — the mechanics of chains in general.
AI for business owners — where decisions like this get made.
Six brief fields — what to give each role as input.
AI agents: what they are and what they pay — the full guide to agents, with market rates.