I Asked AI to Roast My Business Idea. Here's What Happened.

I had been sitting on a business idea for about three months. I had told a few friends about it. They were enthusiastic. I had done some research. The market seemed large. I had sketched out a rough business plan. Everything looked promising. But something was nagging at me — a suspicion that I was seeing what I wanted to see rather than what was actually there.

So I did something uncomfortable. I opened Brutally.ai, typed out my idea in as much detail as I could, and asked for the most honest assessment it could give. Not encouragement. Not a list of ways to make it work. The honest truth about whether the idea was actually good.

What followed was one of the most useful — and most uncomfortable — 20 minutes I have spent thinking about a business idea.

The idea

The idea was a B2B SaaS tool for small professional services firms — accountants, solicitors, consultants — to manage client onboarding. The problem was real: I had spoken to several small firms who described their onboarding process as a mess of emails, PDFs, and manual follow-ups. The solution seemed obvious: a simple, automated onboarding workflow tool.

I had convinced myself this was a good idea. The market was large (there are hundreds of thousands of small professional services firms in the UK alone). The problem was real. The solution was technically straightforward. I had even identified a price point that seemed reasonable.

The honest assessment

The AI's response was structured and specific. It gave the idea a Brutal Score of 5.5 out of 10 — not terrible, but not good. And then it explained why.

The first problem: the market was crowded. Not with direct competitors, but with adjacent solutions that professional services firms were already using — practice management software, CRM systems, document management tools. The question was not whether the problem existed, but whether firms would pay for a standalone solution when their existing tools could be configured to handle it.

The second problem: the sales cycle. Selling to professional services firms is notoriously slow and relationship-driven. A small SaaS company with no existing relationships in the sector would face a customer acquisition cost that would make the unit economics very difficult to make work at a price point that small firms would actually pay.

The third problem: the switching cost argument. The AI pointed out that I had assumed firms would switch to a new tool because it was better. But switching tools in a professional services firm involves training, data migration, and process change — all of which are costly. The bar for 'better' is much higher than I had assumed.

What I did with the feedback

My first reaction was defensive. I wanted to argue back. The AI was being too pessimistic. It did not understand the specific niche I was targeting. The competition was not as strong as it was suggesting.

But I sat with the feedback for a day and then went back to it with fresh eyes. And I realised that all three criticisms were valid. I had not seriously investigated the adjacent solutions. I had not modelled the customer acquisition cost. I had assumed switching costs were lower than they probably were.

The feedback did not kill the idea — but it changed it significantly. Instead of a standalone onboarding tool, I started thinking about an integration layer that worked with the tools firms already used. Instead of targeting all professional services firms, I narrowed to a specific sub-sector where I had existing relationships. Instead of a pure SaaS model, I started thinking about a hybrid model with a services component that would reduce the switching cost barrier.

Why honest AI feedback is different from human feedback

The most striking thing about the experience was the contrast with the human feedback I had received. My friends had been enthusiastic. My colleagues had been encouraging. Even the people I had spoken to who had the problem I was solving had been positive about the idea of a solution. None of them had identified the three core problems the AI identified.

This is not because the AI is smarter than my friends. It is because the AI has no social incentive to be kind. It does not care whether I feel good about the conversation. It is not worried about damaging our relationship. It just tells you what it actually thinks — which is exactly what you need when you are evaluating a business idea.

The uncomfortable truth about business idea feedback

Most people who have a business idea are not looking for honest feedback. They are looking for permission. They want someone to confirm that the idea is good so they can move forward with confidence. The problem is that permission from people who are not qualified to give it — or who are too kind to give it honestly — is worse than useless. It is actively harmful.

The founders who succeed are not the ones who receive the most encouragement. They are the ones who receive the most honest feedback and use it to make their ideas better. The discomfort of honest feedback is not a bug — it is the feature.