Chad’s Blog
Pragmatic Technologies for Life and Business Success®
Last week I did something in a live AI Insider Lab session that I could not rehearse, could not edit, and could not rescue if it went badly.
I put a real business on the screen in front of a room of experienced consultants and let AI take its marketing plan apart.
Patrick and Barbara Lyons volunteered: Barbara Lyons Mulvey Interiors. Barbara has spent twelve years in interior design. She is a color expert, and she can walk into a room and tell you exactly why it is not working for the people living in it. Patrick handles the marketing. Their challenge was simple to state and brutal to solve: they did not know how to reach the buyers who needed Barbara most.
Before I typed a single word about their business into AI, I set the terms of the conversation. This is the part most people skip, and it is the part that determines everything that follows.
I told Claude to keep every response under thirty seconds because the room was watching live. I told it to ask one question at a time, with a small number of distinct options, so everyone could see the choices. I told it to rate its confidence level on every answer it gave me. And I told it this:
Your job is not to help me feel better about anything.
That last instruction is the whole article. Everything you are about to read is downstream of it.
Patrick Made His Case
After a dozen questions about Barbara’s history, her client list, and Patrick’s actual bandwidth, AI produced ten ranked marketing priorities. Google ads landed at number ten. Dead last.
Patrick disagreed, and he had reasons. Someone who feels something is wrong with their home goes to Google and types “interior designers near me.” Google locks in on the exact keywords their buyers would use. And Google is geographically targeted, which matters enormously when your entire market sits inside a twenty-mile radius.
Both reasons are sound. That is what makes what happened next interesting.
So, I put it directly to AI. Patrick believes Google should be marketing option number one. Here are his two reasons. I need your assessment, pushback, or agreement. If pushback, why.
The Pushback
Pushback, not agreement.
AI granted the logic and rejected the sequencing. Google ads capture people already searching for an interior designer. That is a small slice of this audience. Barbara’s buyers do not know why their home feels off, and they actively distrust designers before they ever open a search bar. You are not competing for existing demand. You have to create it.
Then it went somewhere I had not anticipated.
Paid clicks would land on a website carrying five testimonials and no case studies. You would be paying to send skeptical buyers to a page that cannot answer their skepticism. That is an expensive proof of concept for a landing page that is not ready.
And then, unprompted, it argued against itself. Weakness in my position: if Barbara’s competitors are already ranking on Google, waiting costs share of search over time. Meaning every month Barbara stays out, the designers already there get harder and more expensive to displace.
That is not a machine trying to please me. That is a machine showing me the edge of its own confidence.
One Asset, Four Objections
Sitting at number three on that ranked list was the idea that made me stop the session and highlight it on screen.
Turn Barbara’s five testimonials into before and after visual case studies. Real homes, real people, specific colors, specific flow problems solved.
AI’s note on it: this is the asset that kills all four objections at once.
I asked it to name them. It quoted the objections back to me from the brief I had given it. Someone else’s opinion imposed on my home. A large unpredictable bill. No idea when this ends. Embarrassment about letting anyone see the house as it is today.
One asset. Four objections. That is leverage.
Then it flagged the dependency. This only works if Barbara has usable before photos and will let Patrick publish them. Confirm that access before building.
So, I asked.
The Floor Gives Way
Patrick pointed to the photographs already on their website. Dining rooms, kitchens, family rooms, all documenting real work.
Then Barbara said the thing nobody made her say.
The bulk of their inventory of successes is their own home. Their own kitchen. Their own dining room. Their own living room. Some of the remaining images are rooms they did not design at all, purchased for the site.
She volunteered that, unprompted, in front of a room of consultants. It took more courage than anything else that happened in that hour.
I had already been circling the same problem from another direction. Five testimonials, first names only, no faces, no homes, no way to verify a single one. I said it as gently as I could and I meant every word of it: I could manufacture a list like that overnight. What I cannot manufacture is a real client, with a real name, in a real home, saying what actually changed.
The strategy was excellent. There was nothing to build it with.
What the Room Did Next
This is where AI stopped being the most valuable participant.
Becky Morgan, who has spent decades advising manufacturers, reached back into the list of ten and pulled out item eight, which nobody had read closely. Position Barbara as a diagnostician, not a decorator.
Becky saw what that idea was actually worth. If Barbara is diagnosing rather than decorating, she does not need a single before and after photo. Film her walking into any kitchen, asking four questions, and explaining why the room is not working. It can be a kitchen she has never touched.
The photo problem disappears. And the diagnostician frame does something the case studies never could: a diagnostician is not there to impose her taste on your home. She is there to tell you why it is not working. That answers the objection Barbara’s buyers actually have.
AI generated that idea. It ranked it eighth. A human in the room recognized it was the answer to a problem AI had raised three screens earlier.
While that was happening, Colonel John Boggs quietly Googled designers near me on his own machine and reported back what he found. Beautiful photography everywhere. Warm reviews everywhere. Before and after case studies almost nowhere. Independent verification, from outside the room, while AI was still on screen.
Not AI delivering answers to a passive audience. AI produced the raw material, Patrick defended his position, and the room found the idea that made the plan buildable.
The Mechanism
None of that happened because AI is clever.
It happened because I configured the session to argue. I asked for a confidence rating on every answer, and it used them honestly. Medium on the photo question, because client willingness was untested. Medium on the forty-hour weekly plan, because hours available are not the same as skills applied. Those ratings told the room exactly where to push.
I did not watch this session. I directed it. I wrote the constraints, framed the case, decided when to accept an answer and when to force a fight, and pulled the room in at the moment AI had gone as far as it could. And I did not describe any of this to my members beforehand. I ran it live, on someone else’s business, with no ability to edit the outcome, in front of people who would have noticed immediately if it failed.
This is what I mean when I talk about leveraging AI as a thinking partner. Not just learning to ask better questions. Setting the conditions under which my own thinking gets challenged hard enough to sharpen my judgment.
One clarification, because it matters. Disagreement is not proof you were wrong. An AI instructed to argue will argue, including about the things you got right. AI ranked the diagnostician idea eighth and Becky knew it was first. The point is not to make AI argue with you. It is to make AI capable of arguing with you, and then to bring your own judgment to whatever comes back. That is the difference between being agreed with and being advised.
Patrick and Barbara walked out of that hour with a ranked plan, a weekly schedule, and a clear picture of the one asset they have to build before any of it works. That is a starting point, not a finished result, and I told the room as much while we were still in it.
That is why the line at the top of this article is not a provocation. It is a diagnostic. A mirror is not broken. It works perfectly. It gives you back exactly what you brought to it, in better light, with nothing added. If every AI session you run ends with your own idea confirmed and more neatly worded, the AI did not fail you. It reflected you, because that is all you asked it to do. You configured it that way.
That session ran the way it did because I was directing it. The Lab is where you learn to direct it yourself. One of our AI Insider Lab members, Candida Marques, joined with very little AI experience. A few months in, she said this:
“I wound up getting two new clients to work with!”
Registration for the next AI Insider Lab group opens later this month. If you want to see what this looks like on your own business rather than mine, book thirty minutes with me: https://calendly.com/thechadbarrgroup
Bring one real decision or problem you have been circling. That is the only requirement.
Watch the full session here.
This Week’s Carousel
If AI Has Never Told You You’re Wrong, You’re Using a Mirror.
