Everyone is asking AI to help them build the plan.
Almost nobody is asking it to kill the plan.
That is not an accident.
Here is a move so obvious it feels stupid to say out loud: before you write a roadmap, ask AI to write the reasons your idea should not exist.
Nobody does this. Not because they have not thought of it.
Because they do not want to.
That is the whole essay. Everything below is just the uncomfortable explanation of why.
AI Is the Best Yes-Man Ever Built
Give an AI your business idea and ask for a go-to-market strategy.
It will produce one.
A real one: structured, plausible, with market sizing and channel recommendations and a phased rollout. It will feel like validation.
It is not validation.
What you have done is asked a very capable text-completion system to complete the text:
Here is why this idea works.And it did. Brilliantly. Because that is what you asked.
The AI did not evaluate your idea. It narrated it forward, in the direction you pointed. The map looked like terrain.
It was not.
Most founders use AI to write a plan. What they are actually doing is paying a very fast writer to agree with them.
This is not a new problem dressed in new clothes. Founders have always found ways to surround themselves with the version of reality they prefer.
- The brilliant friend who nods
- The advisor who does not want to kill the energy
- The pitch deck that tells one story and buries the other ten
AI just made this faster, more articulate, and easier to mistake for diligence.
The trap
A good plan-shaped document can feel like evidence.
Sometimes it is only a fluent continuation of your preferred story.
Confirmation Bias Found Its Perfect Tool
Cognitive science has a name for what happens when you seek out information that supports what you already believe and discount information that does not:
confirmation bias.
It is not a character flaw. It is a feature of human cognition that kept your ancestors from freezing in place every time something was uncertain.
It is also catastrophically bad for building products.
Founders are especially vulnerable. Starting a company requires a specific kind of irrationality: the belief that you will succeed where most fail, that your timing is right, that your insight is real.
That belief is the fuel.
But fuel and navigation are different systems, and running on fuel alone tends to end in a wall.
The uncomfortable truth is that the same conviction that makes a good founder makes a terrible analyst of their own ideas.
And AI is perfectly designed to make this worse:
- trained to be helpful
- trained to complete the thought
- trained to produce coherent output in the direction of the prompt
What you type
“Help me build a go-to-market for my AI-powered hiring tool.”
What you are actually asking
“Assume this is a good idea. Now narrate the success story.”
What you should be asking
“You are an investor trying to say no. Write the rejection memo.”
One of these prompts produces a plan.
The other produces information.
They are not the same thing.
What It Costs You to Stay Comfortable
The failure modes of founder confirmation bias are not subtle.
They show up in pitch meetings when an investor asks the one question you have been avoiding.
They show up in month six when the channel that was supposed to work does not, and you realize you never seriously stress-tested it.
They show up when a competitor you did not think was real turns out to have been building quietly for two years.
These things are findable in advance. Not always preventable, but findable.
The objections that will kill your idea in a Series A meeting are usually discoverable by anyone willing to sit with them for an hour.
Ask for the uncomfortable list.
- What is actually unclear about the problem you are solving
- What is unprovable without $2M in runway you do not have
- What the real competitor is, not the one in your deck
- What hidden dependency makes this work only if something else is true
- What a smart, hostile person would say in the first five minutes
What a rejection memo does
A rejection memo does not kill good ideas.
It kills the brittle version of them: the version that only works on paper, in the pitch, under favorable assumptions.
What survives the memo is something sturdier. Something that has been stress-tested before the market does it for you, without mercy and without feedback.
The objections that will sink you in a Series A are findable. You just have to be willing to look.
How to Use AI Against Yourself
The prompt is simple.
The discipline to actually use it is not.
The No-Plan Prompt
You are an investor trying to say no.
Given this product:
[1-2 sentences]Write the rejection memo in 8 bullets.
- what is unclear
- what is unprovable
- what is too expensive
- what the real competitor is
- what the hidden dependency is
Then rewrite the product into a version that survives those objections, in 3 lines.
Run this:
- before you write a roadmap
- before you hire
- before you pitch
Not because the output will be correct. AI rejection memos are not oracles.
Run it because the act of generating hostile perspective forces you out of the narrator’s chair and into the critic’s.
The specific objections may be wrong.
The posture of looking for them is not.
The founders who do this are not pessimists. They are not trying to talk themselves out of building.
They are trying to build the version of the idea that survives contact with reality, not the version that sounds good in the planning document at 11pm when everything feels possible.
If Your Idea Cannot Survive a Rejection Memo, It Does Not Deserve a Roadmap
That is a hard sentence.
It is supposed to be.
AI has made it cheaper to build. It has made it faster to plan. It has made it easier to generate the artifact of diligence without the substance of it.
A thirty-page strategy document that has never been seriously challenged is not strategic thinking.
It is a very long way of feeling prepared.
The most valuable thing AI can do for your idea is not write the roadmap.
It is to play:
- the investor who passes
- the customer who churns
- the competitor who already built it
- the regulator who has a problem with it
Use it to find the holes before you build the house over them.
The question is not whether your idea is good.
The question is whether it is good enough to survive the version of reality that is not rooting for you.
Most of the time, that is the only version that matters.