Ideas are cheap. Campaigns are not.
Ask any model for campaign ideas and you get twenty in thirty seconds. That feels like progress and almost never is, because an idea is not a campaign. A campaign is an idea plus a reason it will work, a channel logic that follows from it, and an artefact somebody can say yes to.
The gap between those two things is where the actual job lives, and it is the part that does not get faster just because generation did.
The inputs that decide the output
Output quality tracks four inputs far more than it tracks how you phrase the request.
- A specific person. Not a segment. "Parents of a first child under one" beats "millennials" every time, because only one of them has a life you can write about.
- The tension. What does this person believe or fear that makes the category awkward? An idea with no tension underneath is decoration.
- The constraint. Budget, channel, legal, a client who hates humour. Constraints are generative. They collapse the space fast.
- What success looks like as a behaviour. "Awareness" is not a behaviour. "Switches provider before renewal" is.
Telling a real idea from a plausible one
Generated ideas are fluent, which is exactly what makes them hard to judge. Fluency reads as quality until you try to sell it.
Six questions kill most of them quickly. Most ideas will not survive all six, and that is the point. The value is in the cut, not the pile.
- Could a competitor run this tomorrow with their logo on it? If yes, it is a category truth, not an idea.
- Does it survive being said out loud in one sentence? If it needs a paragraph, it is not finished.
- What does it ask the audience to do differently? If nothing, it is an ad about a product rather than a campaign.
- What does it cost to make? Beautiful and unmakeable is the same as no idea.
- Who does it annoy? An idea nobody objects to is usually one nobody notices.
- What happens in month three? If there is no answer, it is a one-off rather than a campaign.
From idea to campaign
Once one idea survives, the work is expansion with discipline. The channel plan has to follow from the idea, not sit beside it: if you could swap the idea out and keep the same channel list, you have a media schedule, not a campaign.
Then it needs to become an artefact. A deck, a case board, a channel-by-channel plan — something with edges that a client can approve or reject. Ideas that never reach an artefact never get bought, and "we discussed it in a workshop" is not an artefact.
Where the human stays
Judgement about the room. A model can tell you an idea is structurally sound; it cannot tell you this client just lost a pitch with something adjacent, or that their CMO has been burned by influencer work twice this year.
Use generation to get to a defensible shortlist faster, then spend the time you saved on the argument. If you are choosing tools rather than running a brief, the companion piece is our guide on what an AI campaign strategy tool must do.