A free, open method for turning an AI chatbot into a virtual board of advisors for your business — different perspectives, real disagreement, and no single right answer handed to you.
Most uses of AI race to an answer. This method is built to improve your thinking instead: it helps you develop new insights, challenges your assumptions rather than just confirming them, and keeps your "aperture" open and real disagreement in view rather than smoothing it over. You may not end up with an answer. You may discover you were asking the wrong question.
Download it to your hard drive. Then upload it into the prompt window of your favorite chatbot and say "Launch this, please."
Download the current AI Virtual Board Method
In 15 minutes, whether you use a real question or simply want to try it with a hypothetical, you should know if this method will be valuable to you.
You're the One Steering
Control of the conversation passes back to you quickly — sometimes as early as Round 1 for experienced users, by Round 4 for most everyone else. The board's early responses only hint at where this can go.
Some people push hard right away — challenging the board in detail, reshaping its composition, taking it well past a first pass. That's welcome; it means the Method is generative, not a rigid tool. The further you take it in your own direction, the further you'll naturally drift from the original design intent. You're meant to be the one steering.
If you're not getting what you need, the fix is usually in your hands. Push the conversation where you want it to go.
Challenge the board's makeup — ask a member for a different angle, swap someone out, or add a seat you think is missing.
Feed it new data or context it doesn't have yet.
Say plainly what you want more or less of, and in what form.
See "Prompts & Techniques That Have Worked Well," further down this page, for more ideas.
Origin Story
The MIT Venture Mentoring Service has been providing advisory services to startup founder members of the MIT community for 25 years. One of the mentors prepared a working paper on the importance of an ongoing advisory board for his mentees. He remembered reading How I Built a Personal Board of Directors With GenAI by Vipin Gupta in the Sloan Management Review. He and other mentors developed this Method along the lines of that article and other research as a complementary, always-available, no-cost support tool. This first version initializes for startup founders, with suggestions for a board drawn from research on characteristics of advisors particularly well-suited to startups. But you can use it anywhere: simply direct your chatbot to ignore that startup default when forming your initial board.
What's Here Right Now
This site is still very early. Over time it will grow into a curated knowledge base — not a manual for how the Method works. It will draw on what real users have discovered: questions that turned out to matter, board setups that worked well, and things people didn't think to ask until they did.
This knowledge base is a public good, free for anyone to use. We hope it's genuinely useful to you.
Please Help
If you use this and something works, doesn't work, or surprises you, we'd like to hear about it. Your feedback is one of the few ways the Method can improve. Write to us at AIVirtualBoard@gmail.com. You can opt out of future tips-and-updates email at any time by replying and asking, and we won't sell or share your address. Anything you send may help shape anonymized patterns shared on this site, so please leave out details that identify you or your venture.
Privacy & Security
This method is designed to run on your chatbot with the settings you set. AI Virtual Board does not and cannot collect any information unless you submit it. If you are using this within an institution, remember to comply with its policies.
Know Its Limits
The board's Members all draw on the same underlying AI, so their differences are real but not fully independent. And the board can challenge your logic but cannot verify your facts.
Remember: virtual advisors enhance human ones, they don't replace them. Use the board to surface diverse insight and sharpen your own thinking, then bring what matters most to real advisors, real evidence, and real-world judgment for the final call.
Insights from Using the Method
A running collection of what we've learned — strengths, limits, and techniques that work. This will grow as more examples come in.
(A stable core of timeless findings or insights at the top of each category— things that persist and are worth a first-time reader always seeing, plus a rotating “what's new this cycle” note appended or refreshed as new feedback comes in.)
Weaknesses, Cautions & Limitations
(What is underperforming, what should be changed to make the Method better.)
All seats share one brain. Every Member draws on the same underlying AI model — the same training data, the same systematic tendencies. Five confident, differentiated-sounding opinions aren't the same as five independent humans triangulating. Real disagreement among Members reduces this risk, but doesn't eliminate correlated blind spots. Six voices adopted the same wrong premise at once.
The board can't verify your facts. It only reasons as well as what you report — including things you might be blind to. A human advisor sometimes catches something that contradicts your framing on their own; this board can challenge your logic, not your facts, unless you bring in outside verification. If a Member's advice leans on a specific fact, check it — with another AI if nothing else.
The board can spend time answering the wrong question efficiently if the initial framing is off.
Sycophancy / convergence, in both directions — drifted toward agreeing with the user's direction. One case names a variant: the board converging on disagreeing with the user. Remember, AI is biased towards making the user feel well served.
The board is supposed to challenge reasoning, not invent facts and treat them as established, but it can do the latter anyway.
It can miss process check-ins.
Open threads and disagreements can get quietly dropped — good questions can be raised and never resurfaced.
Seats going stale — a member contributes little after an early error and becomes largely silent or redundant.
Personal/identity-level coaching, explicitly outside the board's remit — one user argues the board format is often the wrong tool there, which is itself useful boundary-setting for who should and shouldn't reach for it
Strengths
(What the Method does surprisingly well.)
You can submit the same scenario to a different AI and get a genuinely different board, different questions, different emphasis, sometimes different conclusions. Even the same AI produces a different board on relaunch, which makes re-running a way to stress-test a recommendation rather than just confirm it.
Genuine reframing — vague presenting problems get turned into specific, testable findings.
Real-time self-correction — in one case, the board caught itself solving the wrong problem without the user having to say so first.
Disciplined synthesis timing — It waits until real distinctions are explicit before converging, rather than converging early.
Genuine (not staged) disagreement among members is seen directly in transcripts.
Complements a real board rather than duplicating it — one user's revised composition was built specifically to cover blind spots their actual board and advisors didn’t.
Prompts & Techniques That Have Worked Well
(What people are trying that is pleasing them, worth sharing so it pleases others too.)
Tell the AI what's new or interesting about your situation, and be specific — if you want fresh thinking, ask for it directly.
Feed it data or context even if the Facilitator didn't ask for it.
Re-feed the original setup plus a summary of the board's first recommendation, and ask it to find every flaw and propose a better alternative.
Challenge a particular board member — push on their reasoning, or ask them to fine-tune how they think or what background they bring.
Challenge the tone of the conversation — ask for something more collaborative, more adversarial, more exploratory, whatever fits what you need.
Challenge the kind of responses you're getting — push back directly, ask for sharper disagreement, or ask a member to defend their position harder.
The one-on-one mode of using a chatbot as a thinking partner has strengths and advantages: good at personal/identity-level material, better at helping someone discover what they think before it's articulable.
With your AI outside of the Method begin one-on-one and use it to discover and deepen the real question. Then consider which kind of thinking would add the most value. If expansion or challenge would benefit from sustained, differentiated perspectives, then launch the Method. Otherwise, continue one-on-one.
Move back and forth between one-on-one and your board. Refine your board’s implementation plan then resubmit it for pressure testing – perhaps to a different chatbot.
Things to possibly add or remove
(Nothing logged yet)
Troubleshooting?
(If..., then here's what you might try)
Steering hard with direct corrections. Ask what is your explicit disappointment? What can be done about it?
Ask the board to evaluate itself: "Please evaluate our conversation using all of the history since launch."
Iteratively re-scoping seats against the user's real-world structure.
Other Observations?
(nothing logged yet)
Other Applications & Audiences
(Who is using this for what, beyond the original startup use case. This bucket serves two purposes: it's useful data on how the tool is actually being used, and reading real examples prompts visitors to imagine new uses for themselves or think of others who might use it that way.)
Beyond its original purpose, people (could use) (have used) this for:
Preparing real board members or management teams for a discussion: commercial enterprise, nonprofit boards, community boards and committees
Consultants coaching executives or teams
OD consultants preparing for — or running — a strategy session
Supporting a mentor working with a mentee
Solo founder/consultant personal-operations questions — "how do I run myself and my small team," distinct from classic startup strategy
Mature, non-startup firms far from the tool's original startup framing