|
Well, hello, and welcome to this week's edition of the eCommercer.
Most of us treat AI like it's confident and correct at the same time. It answers fast and sounds sure of itself, so we rarely stop to check whether sure also means right.
My guest this week had a clear example of what happens when confidence outruns context, and it's got me thinking about what we actually feed our own AI before we trust it with anything that matters.
The Ad Budget That Came From Nowhere
Marissa Rodriguez has spent twenty years building and scaling ecommerce brands, first as a fractional CMO and COO, now as founder of Through Experience, a mentorship platform for ecommerce founders. She's also training her own AI on everything she knows, which is why she can spot exactly where generic AI goes wrong.
A founder she was mentoring, pre-revenue and about to launch, asked ChatGPT to build a financial model. ChatGPT decided the business needed to spend 20% of revenue on paid advertising, with no sales history to go on and no questions asked first.
“You’ve never launched. You’re pre-revenue,” Marissa said when she saw it. “Where did ChatGPT get that 20% was the right number?”
That's the whole problem in one example. A generic AI doesn't know what it doesn't know, so it answers anyway, and if you're a first-time founder, you have no way of spotting that the answer is nonsense.
Six Videos Became Fifty
Marissa's fix is AI trained by someone who's actually done the job, not AI that's industry agnostic and built to work for everybody. She's doing exactly that with her own frameworks, and it's taken far longer than she expected.
She started with an email marketing course she already had, six videos, 45 minutes. Simple enough to feed into an AI, she thought. It became 50 videos.
“There was so much more than I anticipated that I actually had to sit down and give it the context for understanding,” she told me. “When does a business need this versus when does a business need that. The answer isn’t always black and white and the same for everybody.”
Her advice on where to start is simple. Get the numbers straight first, the P&L and the financial model, because most ecommerce businesses don't actually have one. Then look at email marketing, where she thinks AI delivers the clearest win for a small or mid-sized team right now.
And underneath all of it, she keeps her own focus narrow.
| |
“There's only three actions a person can take on the internet that matter to an ecommerce business. Click, subscribe, and buy.”
Marissa Rodriguez
|
The AI Board Of Directors
I've done a version of this myself. Everything we've learned at Aurion, our frameworks, our playbooks, our way of thinking through a problem, is in a layer we call SAM, which sits behind Claude for our team. The output only got good once we'd done the unglamorous work of writing down what we actually know, the same work Marissa's six videos went through on the way to fifty.
I've also built something a bit sillier with the same instinct behind it, an AI board of directors. Jeff Bezos, Steve Jobs, Warren Buffett, and a few others, each briefed on everything they've actually said and written. When I bring them a decision, they push back on each other, and on me. Good fun, but the reasoning is real, because it only works when it's given something specific to argue from.
Generic in, generic out. Specific in, useful out. Marissa and I landed on the same lesson from two very different directions.
| 
