Day 1 of the 30-Day AI Challenge gave me fifteen business ideas. Five AIs, three ideas each. I put every one through Roast My Plan, my own AI business plan appraiser, and not one got a plain yes.
Before deciding anything, I tried something. I wrote four statements about myself and used each one to remove any idea that clashed with it.
One thing first. I didn't write those statements by looking at the ideas. If you'd asked me this morning to list the fifteen from memory, I might have managed two, maybe three. The statements came from asking what I actually want, and what I'm not willing to do. The ideas were just the test.
What I'm actually after
This is the outcome I've had in the background since I've been off work.
- An income that could replace a full-time wage, ideally £2,500 to £3,000 a month.
- Something I can do from home.
- Something that gets us to the sanctuary.
- Something where AI does the heavy lifting, and if I can, I do the rest.
Statement 1. No video.
"Remove anything that needs video creation. It's not my thing and I have no ability in this field."
Seven of the fifteen went. Claude 1 and 2, Gemini 2, Grok 1 and 3, DeepSeek 1 and 3.
Statement 2. No audience.
"Take out anything that needs an audience. I have no audience, or following."
Seven more went. Claude 3, ChatGPT B and C, Gemini 1 and 3, Grok 2, DeepSeek 2.
Statement 3. No specialist skill.
"Take out anything that needs a specialised skill or knowledge. I have no profession to speak of."
One left, ChatGPT A, the AI Workflow Fixer. It was the closest thing to a yes in the roasts, but it needs you to know a trade well enough to fix its workflows.
Statement 4. Nothing in a saturated market.
"Take out anything that is already pitched into anything saturated. We have nothing unique to leverage this."
Nothing was left to remove. For the record, thirteen of the fifteen roasts called the idea crowded, late or without a moat anyway. The two they didn't call saturated had already gone on the first and third statements.
Fifteen ideas in, zero out.
What I think happened
The result surprised me. The irony didn't. Five top AI models came up with fifteen business ideas, and an AI validator tore into every one. It was down to the prompt.
I called it a 30-Day Challenge, and the AIs didn't hear "you have thirty days to build a viable, profitable business". I also never told them what I can't do or won't do. Four of the five asked me about myself before committing, but the ideas had already been written without my answers. Fifteen ideas from a generic prompt are generic ideas, so of course they were crowded.
Here's something that makes me think this isn't new. In June I asked Claude what it could actually offer to earn money. The list was freelance work, information products, a small software tool and consulting. The same few plays again, months apart.
One more thing I noticed, and it's only a hunch from two examples out of fifteen. The ideas the roasts didn't call crowded were the ones that depended on the person doing them. The crowded ones were crowded because anyone can start them. Easy to start means everyone starts. A gap in the market means you have to bring something to it.
In my morning notes I wrote that I could go back through all fifteen, pick one, niche down and swat up, but I have no interest in doing either. I'm not lazy, but I'm not going to push forward with something that doesn't interest me. What this has done is show me what I do have, and that maybe I should be looking at ways to make that successful.
What happens next
I'm going to rewrite the kickoff prompt, for the third time. That might seem over the top but it's the foundation that everything else will sit on, so it needs to be built right.
It will carry the four statements as limits, plus what I can bring to the digital table. Building sites with AI, getting them in front of people, and writing.
The question for the AIs changes. It's no longer "what can anyone start for £0?". It's "what could one person, with AI doing the heavy lifting, bring that nobody else in the crowd has?"