That sentence is the shortest honest description of what I do. It also disqualifies me from a lot of sales conversations, which is exactly why I lead with it.
Nobody selling AI ever says "you don't need it"
Think about who a business owner can ask about AI right now.
The software vendor? Their answer is yes, you need it, and conveniently they sell it. The agency that just added "AI" to their service list? Yes, and they have a package. The consultant who bills by the build? Yes, and the bigger the build the better.
Everyone in the room gets paid when the answer is yes. Nobody gets paid to say no. So the owner never actually hears a diagnosis. They hear a pitch wearing a diagnosis costume.
I've spent 25 years in online advertising, business development, marketing, and analytics. I've watched this same movie with every tool wave. The businesses that got burned weren't dumb. They just never had anyone in the room whose answer was allowed to be no.
What I actually look at
When I sit down with a business, I'm not looking for places to install AI. I'm looking for drag. Where the hours go. The questions are boring on purpose:
- What does somebody here do three or more times a week that follows the same steps every time?
- What report, list, or email gets rebuilt by hand on a schedule?
- Where does work sit waiting for one specific person to touch it?
- What's the thing you'd check every hour if checking were free?
- What lives only in your head, or only in one employee's head?
Notice none of those questions mention AI. That's the point. The problem gets named first, in plain language, with a real cost attached. An hour a day. A missed invoice a month. A customer who didn't get a follow-up. Then, and only then, does the fix get picked.
Three honest outcomes
A real diagnostic can land three different ways:
- AI fits. The work involves judgment or language. Drafting replies, summarizing documents, sorting incoming requests by what they actually need. This is real, and when it fits, it's a beautiful thing.
- A plain automation fits. The steps never change. Move this file, fill this cell, send this email when that thing happens. No AI required, and honestly, no AI wanted. Boring automation is cheaper and it doesn't hallucinate.
- A better process fits. The steps are fine, they just live in somebody's head or happen in the wrong order. Writing them down fixes it. Cost: an afternoon.
I'll go deeper on choosing between these in another post. The short version is that most problems sold as AI problems are actually number 2 or number 3 problems wearing a fancier price tag.
Sometimes the answer is "keep doing it by hand"
This one surprises people. There are tasks where the honest recommendation is: change nothing.
The task that happens twice a year. The judgment call where you're the judgment. The customer conversation where showing up personally IS the product. Around here especially, plenty of business runs on somebody knowing your name and your order. You don't automate that. You protect it.
If a task costs you twenty minutes a quarter, any build I could sell you costs more than the problem. I'll tell you that. It's a short conversation and I don't make a dime on it.
What the honesty costs me, and why I do it anyway
Let's be direct. Telling a prospect "you don't need AI for this" costs me a project fee that week. I know. I've done it plenty of times, and I'll do it again.
I was raised with the idea that integrity and honor are two things a man should hold dear. So the math here was settled a long time before AI showed up. I would rather tell you not to buy something, and lose the money, because it's the honest thing to do, than pad my wallet on a build you didn't need. People who know me will tell you: I say no all the time. Always have.
My reputation is worth more than a one-off project. I live in Fredericksburg. This is my market and these are my neighbors.
The test you can run on anyone, including me
Next time somebody pitches you AI, ask them one question: "What would make you tell me NOT to buy this?"
If they can't answer, they're not diagnosing, they're closing. A real answer sounds like: "If the task is rare, if the steps require your personal judgment, or if a checklist would fix it, I'd tell you to keep your money."
That's my answer. It's on the record now. Hold me to it.
Ranch photo: brian.perkes (flickr.com/photos/57253465@N03)