Confessions of an AI Consultant: 3 Projects I Would Not Take Again

I run an AI consultancy. Most days I tell business owners to automate, delegate, and eliminate. This is the post where I admit I did not always follow my own advice.

Every consultant has a private list. Projects that drained the calendar, paid late or never, and taught more than they earned. Here is mine. Names removed. Lessons kept.

The Free Audit That Became Free Consulting

Early on, I used deep audits as a door opener. A prospect would get a real teardown of their website or their process. Serious work, hours of it, delivered free to prove I knew my trade.

Some of those audits turned into clients. Some turned into silence. One turned into the worst outcome. The prospect took the audit, implemented half of it with someone cheaper, and I never heard from them again.

The problem was not generosity. The problem was that I gave away the diagnosis and kept nothing for the prescription. A free audit answers the question "what is wrong." It should never answer "how to fix it."

What I ask now

I still do diagnostics. But the deep work is paid, and the free version ends with findings, not a roadmap. If a prospect will not pay for the diagnosis, they will not pay for the cure.

The Client Who Wanted AI But Wouldn't Change Anything

A business owner called me because his team was drowning in manual work. He wanted AI. What he did not want was to change a single step of how his team worked.

I should have walked away in the first meeting. Instead I told myself I could build around it. I automated the edges. The core stayed manual, because the core was how they had always done it.

The system worked exactly as designed. The project still failed. You cannot automate a process the owner refuses to rethink. AI makes a good process faster. It makes a bad process faster at being bad.

What I ask now

Before I quote anything, I ask what they are willing to stop doing. If the answer is nothing, I pass. I am not expensive enough to be a very clever way of avoiding change.

The Project Where I Was the Product

This one stung the most. A client did not buy a system. They bought me. Every decision needed my eyes. Every exception needed my judgment. The automation worked, but the business around it still ran on Simon.

I had built exactly what I warn clients about. A process with a single point of failure, and the single point of failure was me. It paid well and scaled terribly. I could not delegate it, productize it, or leave it alone for a week.

What I ask now

If I cannot describe how the client runs it without me in the third month, I do not take the project. Automatiser, déléguer, éliminer. It applies to my own business first.

The Pattern

Three different projects, one mistake repeated three ways. I sold my time instead of my systems. Every time, the work felt good in the first month and bad by the sixth.

So here is my rule now, and I give it to you free. Before you hire anyone, consultant or employee, ask what happens when they leave. If the answer is "everything stops," you did not buy a solution. You rented a person.

That is what I build now. Systems that survive without me. If you want one, you know where to find me.