Tell Me What's Wrong

I asked for a second opinion on my own offer, with one instruction: tell me what's wrong. The useful part came after the answers.

Last week I asked two reviewers with no reason to be nice to me to take my business apart. Both of them were AI models, which made the no-reason-to-be-nice part easy.

The offer, the steps, the prices, the promise on the page. All of it, in one document, with one instruction on top: you have no stake in this plan, find what's wrong with it, and don't try to cheer me up.

I do this for clients all the time. I sit with a founder, look at how their business actually earns, and find the one thing that's holding the rest back. It's a lot easier to do that for someone else than for yourself.

So I built myself a second opinion.

The Brief

If you ask a model "what do you think of my offer?", you get a mirror that nods. It tells you the positioning is strong, suggests three tweaks to the headline, and wishes you luck.

That's useless. So I spent more time on the question than on anything else.

First, I took every name out. No company, no clients, no people. A reviewer who knows who you are starts being polite, and that goes for software too, because it picks up on how you describe yourself.

Second, I wrote down what was not up for debate. I work solo. I don't give strategy away on a free call. I don't start anyone on a monthly retainer. Those are decisions I've made with my eyes open, and I didn't want to spend the review defending them. You can challenge these, I wrote, but only if they make the whole model impossible.

Third, I described what success looks like. By early next year: a repeatable way for people to become paying clients, deliverable by one person in thirty to forty hours a week, without sliding back into open-ended work where I end up as an unpaid part-time employee.

That last bit is my known pattern. I put it in on purpose.

Then I added the material. The offer as I designed it in June, and next to it, how the paid work of the last few months actually came in.

That comparison turned out to be the whole review.

What Came Back

Both models answered separately, and both landed on the same point within the first few paragraphs.

My offer asks people to start with a diagnosis. A few weeks, a written report, one clear answer about what's holding them back, and a plan for what comes next.

The work that actually came in looked different. People came with a problem they could already name. The process nobody owned. The tool that half worked. The step between two people where work kept getting lost. They didn't want it described. They wanted it fixed, and they wanted to see the first piece running.

One of the models put it more bluntly than I would have: the front door sells a report, and your clients are buying the fix.

It stung a bit, mostly because it was obvious once someone said it.

The Part They Got Half Right

Here's where it gets interesting, because I didn't just take the answers and run.

I went through every point and held it against what I actually know: emails, call notes, who asked for what.

Most of it held up. One point didn't. One model went after my guarantee, the one I was proud of in the Start here piece a month ago: I name your biggest constraint or you don't pay. It called that false precision, an argument waiting to happen about whether I found the right thing. I don't buy that. I can always name something, and nobody has ever disputed it.

But the other model had spotted the real problem hiding underneath. My promise covered naming the problem. What a client wants is that it's solved and working in their team. I was guaranteeing the part I'm good at, and staying quiet about the part they care about.

And then there was the point that was half right.

Both models assumed people had looked at my front door and walked around it. That they'd seen the diagnosis and said "no thanks, I want something else."

When I checked, that wasn't what happened. Most of the people I work with never saw the front door at all. When a conversation got going, I improvised something that fit. A smaller piece of work here, a different shape there. I built those side doors myself, one conversation at a time, and then wondered why nobody came through the front. The smallest yes I wrote about two weeks ago was one of them.

That's the bit I'm keeping.

It also wasn't new. In August I wrote a note to myself after a project that had grown in every direction at once: make it smaller, solve one problem, do it very well, and let it roll from there. Then I carried on with a front door that asked for weeks of commitment before anything got fixed.

I had written down the lesson and then designed around it.

What Changed

In that same Start here piece, I put it in writing: everyone starts with the Blueprint. I meant it. It's no longer true, and I'd rather say that here than let it drift quietly. I've changed that line there and on my site.

The change itself is small, which feels right.

When someone has a problem they can name in one sentence, we start there. One problem, a couple of weeks, a fixed price agreed up front, and an end date. At the end there's a decision on what to do and what to drop, and the first working piece is already running in their team. One condition: someone on their side builds it with me, so it stays when I leave.

If you've read The Loudest Problem, you might spot a tension here. That whole piece was about how the problem people bring is often not the one holding them back. That's still true. What changed is where I look for it. I used to ask for weeks of diagnosis before anything got touched. Now I start with the problem you name, and if something else sits underneath, it shows up while we build the first piece, usually faster than it would in a report.

The Blueprint still exists. It's the right tool when someone can't name the problem yet, when everything feels stuck and nothing is clearly the cause. It just isn't the only way in anymore.

I posted a short version of this on LinkedIn on Tuesday. What I didn't say there is that the idea came from a document I'd asked to be torn apart.

One Layer Down

The easy lesson here is "use AI to review your business." That's true and it's also the least interesting part.

The review itself took an afternoon. The value came from two things on either side of it.

Before: writing a question that couldn't be answered with a nod. Taking the names out. Saying what was off the table. Describing what good looks like, including the mistake I tend to make.

After: checking every answer against what I actually know, and noticing where the reviewers had to guess. They guessed that buyers had rejected my offer. The truth was closer to home, and no model could have known it, because it lived in how I behave in a conversation.

If you run something, you can do this on Monday. Pick the one thing you're least sure about, your offer, your pricing, the way your team hands work over. Write it down without names. Write down what isn't up for debate and what success looks like. Then ask for what's wrong, and nothing else.

Most of what comes back you'll already half know.

That's usually the part worth hearing.

What's one thing in your business you'd never ask anyone to review?

The Real Account is the honest version of this work. No hype, no playbook. If this was useful, I'm on LinkedIn, and I help founder-led brands find the one constraint holding their marketing back at heerdenco.com.