Someone asked me last month what my favorite AI use case is, expecting me to say content generation or email drafting. It’s neither. My favorite use case is arguing with it, the digital equivalent of a devil’s advocate who never gets tired or takes it personally.
I’ll lay out a strategy I’m genuinely convinced is right, then ask it to find every hole in it. Not to write the strategy. To break the one I already wrote. That’s a completely different job than
“make this faster,” and it’s the one I think most small business owners never try, mostly because arguing with a chatbot feels a little unhinged the first time you do it. It is a little unhinged (it’s okay, we’ll be fine). It’s also extremely useful.
Fast is not the same as smart, and here’s what that costs you.
Most people use AI the same way: as a typing accelerant. Draft this email faster. Write this caption faster. Summarize this document faster. All useful, none of it is actually using the tool for what it’s uniquely good at.
Here’s what gets missed. AI has effectively read more marketing strategies, more failed launches, more customer complaints, and more competitor positioning than any single consultant could get through in ten lifetimes. That’s an enormous asset for stress-testing an idea, and almost nobody uses it that way, because stress-testing feels slower and less satisfying than watching a finished draft appear like a magic trick.
There’s a story I keep coming back to. EY reportedly ran an experiment across hundreds of executive sessions, different industries, different continents, and asked each group to use AI to invent a brand-new snack from scratch. Independently, team after team landed on the exact same answer: matcha, monk fruit, adaptogenic herbs, minimalist eco-packaging, reported by Kampaign Lab. Every team walked away thinking they’d invented something original. Collectively, they’d invented one snack, forty times over. That’s what happens when you ask a tool to generate an idea instead of asking it to attack one you already had.
Sit with why that happens, because it explains a lot about how to actually use these tools well. When you ask AI to generate, it heads for the center, the same statistical middle we talked about with voice. Everyone asking the same open question gets pulled toward the same crowded consensus, which is how forty smart teams independently invent one snack. But when you ask AI to critique something you already made, the dynamic flips. Now it’s not reaching for the average answer, it’s reacting to your specific answer, poking at your specific weak points. You’ve given it something particular to work against, and particular is where it stops being generic. Generation pulls toward sameness. Critique pushes away from it. Same tool, opposite direction, entirely because of how you pointed it.
Try this instead of your usual prompt. Take a decision you’ve already made, a price increase, a new offer, a change in who you’re targeting, and ask AI to argue the other side as convincingly as it can. Not “is this a good idea,” which invites a polite yes and nothing else. Actually ask it to build the strongest case against what you’re planning. Half the time you’ll find a real gap you missed. The other half, you’ll come out of it more confident, because your idea survived a genuine attack instead of a rubber stamp from something that’s intentionally designed to be agreeable.
This is the difference between “more content” and “better thinking,” and it’s the gap between AI as a content mill and AI as an actual creative partner. A content mill produces volume. A creative partner makes your original idea sharper before it ever goes out the door.
Fast gets you forty average posts by Friday. Smart gets you one post that’s actually right, tested against the best objection anyone could raise, before a single customer ever sees it.
Takeaway: The best AI prompt isn’t “write this for me,” it’s “tell me why I’m wrong.”
Try this: Take one decision you’ve already made this month and ask your AI tool to argue against it as convincingly as possible. See what survives. Then, if you’re feeling brave, ask it to argue against its own argument. The third round is usually where the genuinely useful stuff shows up.
Next issue I want to talk about something a little more behind the scenes: how much of what I send you is AI-assisted, and why I think that’s completely fine as long as one condition is met. Talk soon, Tiffany