Stop asking AI what problems exist in your industry
The question is the problem. Ask an AI assistant what problems exist in your industry and you will get a well-organised summary of what everyone already says, because it has no idea what you know. Three reframes fix that, and all of them work by making your experience the context rather than the question.
Download the AI Prompts accompanying worksheet.
Do deep research first
Before any of the reframes, run deep research on the problem you are trying to solve. Most assistants have it, and most of them have it on the free plan.
It gives you a foundation to have the conversation from, rather than starting on a blank page. The rest of this only works properly once that is done.
Reframe 1: from “what problems exist” to “what am I assuming”
Instead of: what problems exist in my industry?
Ask: here is what I have seen over the past 20 years. What am I assuming?
The first version asks the assistant for information it does not have. The second gives it something to work with and asks it to do the one thing it is actually good at, which is spotting the structure in what you already know.
The assumptions list is the part most people skip, and it is usually where the interesting thing is.
Reframe 2: from “who is my ICP” to “argue against my ICP”
Instead of: who is my ideal customer?
Ask: I think this is my ICP. What would you say to argue against that?
Asking who your customer is invites a persona. Asking it to argue invites a stress test. The second one is useful and the first one is decoration.
Reframe 3: from “write my value proposition” to “what have they already tried”
Instead of: write me a value proposition.
Ask: what would someone in my ICP have already tried as an existing solution or workaround?
This is the most underrated of the three. Your competition is almost never another product. It is a spreadsheet, a manual process, and somebody’s patience. Knowing what the workaround is tells you what you are actually replacing.
Two more worth running
Ask it to argue with you. Go back and forth like a tennis match. Ask for the perspectives you have not considered.
Ask why this would fail. The anti-argument. What would make this not work, and what has to be true for it to be a bad idea.
What AI does not have
It has access to most of the generic information online. It does not have the expertise you built over 10, 15 or 20 years in your industry, it does not have your context, and it does not have the insight that came from watching the same problem repeat for a decade.
So it builds structure around your idea. It does not supply the idea.
And the output is not a set of conclusions. It is a better set of questions. Questions for real people in your market, because that is the only place the need actually gets proved or disproved.
Running deep research
All three of the major assistants have a deep research mode, and most are available on the free plan.
- Claude: enable research before sending your prompt
- ChatGPT: select deep research from the tools menu
- Gemini: choose the deep research model
The walkthroughs are in the video from 5:11.
Find out which of the five is holding you back. The Founder Blindspot Check is 11 questions, about four minutes, weighted across five areas. No email gate, and your score appears on the page.
FAQ
Can AI validate my business idea?
No. It has no access to your customers and no way of knowing whether anyone will pay. What it can do is structure what you already know, surface assumptions you have not noticed, and help you build a better set of questions to take to real people. Validation happens in conversations, not in a chat window.
What is deep research in ChatGPT, Claude and Gemini?
A mode that runs a longer, multi-step research process across sources rather than answering immediately. It produces a more comprehensive document and takes several minutes. Most of the major assistants offer it, and versions are available on free plans.
What should I ask AI about my business idea?
Give it your own experience as context first, then ask what you are assuming, ask it to argue against your ideal customer profile, and ask what your customers have already tried as a workaround. Also ask why the idea would fail. Avoid asking it what problems exist in your industry, because it does not know.
Why does AI give generic answers about my industry?
Because you have not given it anything it does not already have. It is drawing on general information, so a general question returns a general answer. The specific thing you bring is twenty years of context, and that has to be in the prompt.
How do I find my ideal customer profile with AI?
Do not ask it to generate one. Propose one from your own experience and ask it to argue against it. A generated ICP is confident fiction, and you will build on it without noticing.
No spam. Unsubscribe whenever. I would rather you did that than sat on a list you do not read.
Curious first? Read the back issues →