The best AI still doesn't make a good campaign

"Our AI is the best." After twenty years of technology and data promises, that's just the next one being paraded through town.

Here's what that looks like in practice. A campaign automatically tested several ad creatives against each other, and after a few days a system dropped the one that got the fewest clicks. But that was the creative that actually worked best with the audience that mattered, just with a smaller group of people. Without someone catching that and stepping in, exactly the right creative would have disappeared.

That shows where the limit sits. A system can try many variants at once. It can't know which audience actually matters for a campaign if nobody defined that beforehand.

That's why we stay tool agnostic. We use whatever genuinely makes sense for a campaign, whether a learning machine sits behind it or a classic, hand managed setup.

Next time someone shows an AI slide, ask one question: which decision did it improve, and how can you tell? Everything else is packaging.