A friend of mine in California is helping organize a big event for her kid's school right now. The committee has leaned hard on AI for a lot of the planning, including the creative parts, not just the logistics. When she reviewed what the group had come up with, her reaction surprised me. Nothing was wrong exactly. Every idea worked. But she couldn't point to one that felt alive. "It's like the ideas were already average before anyone even voted on them," she said.

There's a reason for that. AI is trained on what's already been written, designed, and pitched by nearly everyone. Ask it for ideas and it hands you the most common answer sitting in that pile. Fast, safe, competent. Also the last thing you want when the goal is a theme, a tagline, or a campaign that people actually remember.

Here are three prompts I use to push it off the average instead of accepting whatever list it gives me first.

Ask for the common answer, then ask past it. Literally: "Give me the most common approach to this problem. Now give me 2 to 3 ideas that would appear on this list with only 5% probability." That second question forces the model out of the safe cluster in the middle of the distribution and into the tail, where the interesting stuff sits.

Anchor it with a paired example before it generates anything. Show it what "average" and "weird" look like in your world, not just in general. For a corporate merchandise gift, the average idea is a battery pack. A 5% probability idea is a bar of chocolate or even a reed diffuser. Hand it one pair like that and its next round gets noticeably sharper.

Let it interview you first. Before you ask for ideas, tell it: "Before giving me any ideas, ask me questions to understand what 'creative' means to me here." It'll ask about the audience, the vibe, what's already been tried, what's off limits. Answering those questions shapes the output more than any single clever prompt, because now the model has your taste in the context window instead of the internet's.

None of this is specific to school committees. I run the same three moves before a client pitch, a headline for a launch, a subject line, anywhere the first draft matters more than the tenth. The prompt takes maybe 30 extra seconds. The gap between "technically fine" and "someone in the room reacts to it" is usually right there.

If my friend's committee had run this year's event through those three moves before anyone voted, I'd bet at least one idea would have made someone laugh, or argue, instead of everyone just nodding along.

Alex