VEDANSHI. ← Back to portfolio
AI product Working prototype AI behaviour design

Designing a room that argues with your idea before you build it.

Good product ideas rarely die because nobody had them. They stall because the right perspectives were not in the room while the idea was still flexible. Getting a designer, a PM, an engineer and a customer voice together takes weeks, and most ideas lose momentum before that happens. Boardroom AI is a working prototype that puts an idea in front of six conflicting perspectives in a single sitting. They question it, argue with each other and end with a verdict on what to do first. I designed how that room behaves, not only how it looks.

Type
Working AI prototype
Role
Product Design
AI behaviour design
Status
Self initiated exploration
Working prototype
Inside the prototype
6
Perspectives
3
Debate stages
8
Behaviour rules
5
Design decisions
01The Moment

A good idea should not disappear because the right people were not in the room.

Three of us were in a discussion, a designer, a PM and another product person. One of us shared a genuinely good idea. It got some feedback, and everyone agreed it needed more discussion. The PM said stakeholders had to weigh in on feasibility and direction, but getting them together was hard. A month later it still had not moved, and it eventually went nowhere.

Four panel comic. 1, a product person says it is a good idea but more perspectives are needed before building it, and getting everyone in the room is hard. 2, what if the missing perspectives did not have to wait for the meeting, shown as Boardroom AI with six lenses: customer, product, engineering, business, design and risk. 3, the six perspectives challenge each other around the product idea. 4, a verdict with a proceed signal, biggest opportunity, biggest risk and next step.
The insight

The problem was not a missing process. The right perspectives were hard to bring together while the idea was still flexible. This was one observed moment, not a study, and it gave me a hypothesis worth prototyping.

02The Problem

Why do some ideas move forward before their weakest assumptions have been challenged?

So I traced the path an idea usually takes. Pick a problem to see where it bites.

03Why AI

What if the missing perspectives did not have to wait for the meeting?

That left one question, how to get more perspectives in front of an idea early. I compared six ways. The first five each solve part of it.

I was not trying to replace the boardroom with AI. I was trying to give the boardroom more perspectives before it made a decision.

04The Board

Six perspectives. Six different questions.

Next, who sits at the table.

Decision

Six lenses, not six job titles. More roles started to overlap, so I kept only the six that ask fundamentally different questions.

05The Behaviour

Six personas did not create a boardroom.

Then the first version failed. It felt like six separate ChatGPT responses. Flip between the two.

Idea in. Six answers out. Nobody reads anybody else.

How the room moved

What I decided

I had to define how the AI behaves, not just who it is. Persona design became behaviour design.

Eight rules written into every seat

    A real exchange from the recorded demo

    06The Prototype

    Enough explanation. Watch the room work.

    So I built it and recorded it running. The verdict, the scores and the follow up questions are all in here.

    07What I Learned

    The challenge was not getting AI to generate opinions.

    It was designing constraints that make those opinions useful to a human decision maker.

    The early impact was not changing the decision. It was changing the conversation before the decision.

    Early validation

    I tested the concept with a small group. People did not want more AI generated content. They wanted AI to structure their thinking and surface what they had not considered. This was directional feedback, not a usability study.

    What it demonstrates

    • Product concept
    • AI behaviour design
    • Interaction design
    • Prompt architecture

    Not proven yet

    • Product market fit
    • Real business impact
    • Large scale adoption
    • Decision accuracy

    My role

    Product
    Problem framing, product concept, user flow and interaction model.
    AI behaviour
    Six perspective definitions, behaviour rules and prompts, disagreement logic.
    Design
    UI and visual system, interaction states and motion, information hierarchy.

    With AI products, the interface is only part of the design. The behaviour you design underneath it is the product.