The $100,000 Testimonial Came From a Man Who Didn't Know How to Turn On His Computer
By Atiba de Souza
The wrong ICP screening question
At one of Atiba's recent AI events, Darvin asked a fair question before the doors opened: could a coffee-shop owner who doesn't even know what AI is be a target for this?
Atiba's answer wasn't a theory. It was a man who had already come. His entire prior experience with AI, in his own words, was signing up for the course. Another attendee that same year showed up with a brand-new computer and asked how to turn it on.
By any normal screening logic, that man was the least likely person in the room to get anything out of two days of AI training. The obvious ICP was the opposite: people already using AI, people who'd get it faster, people who'd make the workshop look good on camera.
Two days later, he put a number on it.
A hundred thousand dollars this year. That's what a man whose entire prior experience with AI was signing up for the course said he expects it to add. Not from what he built in the room. From how he now thinks about solving problems.
Atiba is explicit about where that number came from, because it's the whole point: it didn't come from the tool, it came from the rewiring.
Sit with that before you go any further. If you'd screened your ICP off "already comfortable with AI," you'd have filtered out the best testimonial in the file.
Why "ask AI what your customers would rave about" is a tactic pretending to be a strategy
Here's the instinct almost everyone has, because it's fast and it feels like progress: open a chat window, ask AI to generate the pain points, the testimonials, the "what would delight our customers" list, and ship it into the marketing plan by lunch.
That's a tactic. And in the 7 Practices of Effective Marketing, no tactic gets proposed until three things exist first: a defined Win, the steps mapped in order to reach it, and a narrowed focus on the one thing that matters right now. Skip those and you're not doing marketing, you're doing decoration. Fast, busy, and going nowhere.
The Win. Every game has a way to score, and everyone playing knows what scoring looks like. Before you go hunting for a rave, in AI or in a real conversation, you need to define what a rave sounds like in terms you can measure. Not "positive feedback." A specific shape.
Rave: an unprompted sentence, in the customer's own words, about the return they got. Not a compliment you wrote for them to approve.
An AI-generated testimonial can never meet that definition, no matter how well it's written, because it was never unprompted and it was never theirs.
Thinking steps, not the prompt
The step that gets skipped is the one before you ask anyone anything: who is this customer, what outcome did they actually buy, what problem were they solving when they came to you. That's the foundation the man with the brand-new computer exposes. Nobody would have mapped him as the customer. He was mapped as the outcome he wanted and got: a different way of thinking about problems, worth six figures to him this year.
If Atiba had asked AI "what would our attendees rave about," fed it a profile of confident early-adopters, and built the marketing copy around that guess, this testimonial never gets collected, because the person who gave it never gets invited.
Narrow the focus, then go get the sentence
There's usually more than one thing you could put in front of a customer at any given moment. The discipline is picking the one. Not the feature list. One sentence.
This is exactly what's playing out with the pharmacy owner Atiba has been teaching AI to since June. She's the owner of a seventeen-year pharmacy who opened a wellness clinic, and a few months ago she could barely get Claude installed. She's the person the usual arc predicts will quietly stop opening the tool. Instead, she's now exhausting the $100-a-month plan every week. Say that plainly: the same person who struggled with the install is the one now maxing out the subscription. That's the turn, not a general trend, hers specifically.
The instinct most people would follow here is to write the rave for her, or ask AI to draft something flattering about "ROI." Atiba does the opposite: when a client says they're burning through their plan, congratulate them, then have them say out loud what the spend is returning. That sentence is the testimonial. Nothing gets added to it.
Was to help you understand that this can be something you integrate in so many different ways, and… as long as, as well, I think you're going to say from what you've said already, that I'm spending $100 a month on this, but I'm getting way more value than $100 a month out of this.
The usual arc runs the other way entirely: someone struggles with the install, decides the tool isn't for them, and quietly stops opening it. She's the exception, and the exception is what you narrow your focus onto and quote directly.
The scale you're actually standing on
There's a scale for measuring how far a company has actually travelled with AI, against how far it believes it has. Most companies that say "we use AI agents" are describing the first two steps of a much longer distance, and calling that arrival is exactly what stops them going further.
The same scale applies to your marketing. Asking AI to guess a rave and dropping the output into your site copy is step one. It feels like arrival because it produced words fast. It isn't. Arrival is a real customer's own sentence, checked against the Win you defined before you went looking, and used because it's true, not because it's polished.
| Asking AI to guess the rave | Asking the customer to say the return out loud |
|---|---|
| Generic enough to describe any product in the category | Specific, uses their number, their words |
| Feels productive, ships the same afternoon | Feels risky, might come back as silence |
| Confirms whatever you already believed about your ICP | Sometimes hands you the customer you'd have screened out |
| Marketing built on a guess | Marketing built on a sentence you could put in front of a judge |
Name the fear, because it's the reason people default to AI
AI always answers. It never sits in silence after you ask it a hard question. Ask a real customer to name the return they got and you might get nothing. You might get "I don't really rave about you." You might find out the ICP you built the whole campaign around was never the one buying the outcome, the way the man with the new computer wasn't supposed to be your ICP either.
That's the real cost of doing this correctly, and it's not a process problem, it's an ego problem. You have to be willing to ask ninety customers and hear silence from most of them, to get the one sentence that's actually a rave. Asking AI to invent the sentence instead is not a shortcut through that discomfort. It's an escape from it, and it produces marketing that sounds like marketing because that's exactly what it is.
Rank these, because they are not equal
The Win matters most. Skip it and every rave you collect, real or generated, is unmeasured against anything, which means you can't tell a good customer's sentence from a flattering one. Narrow the Focus is the one people treat as an afterthought, an intern's job, and it's the one that turns a wall of quotes into a single line that closes deals. Thinking Steps Not Tactics is the discipline that stops your hand from reaching for the prompt window first.
The Win is also the one governing what you do Monday. It's the filter you check the sentence against once you have it, not something you circle back to later. If you skip that check, the fear above wins by default, because a vague, flattering sentence is easier to accept than a true one that doesn't quite land.
And here's when not to do any of this: if you don't have a real customer who's actually gotten the outcome yet, don't manufacture one to fill the slot. Wait. A counterfeit rave, AI-written or otherwise, is the fastest way to burn the one asset that was going to work.
The thing to do Monday
Pick one customer this week. Not your best guess at your ICP, the one who's actually using the thing and getting a return, even if they're the person you'd have screened out, like the man with the new computer or the pharmacy owner who once couldn't install the tool.
Ask them to say, out loud, in their own words, what that return is worth to them. Expect silence from most of them if you try this across your customer list; ask anyway, because the cost of that silence is smaller than the cost of a sentence you made up. When you get one that lands, check it against the Win you defined before you asked, not after. Don't edit it. Don't ask AI to tighten it up. That sentence, exactly as they said it, is your next piece of marketing.
Questions