The part of creative work I enjoy is the moment when a familiar task suddenly looks different. That is where I would look for help from AI: a conversation that opens another direction and gives me something to think about.

When asking for alternatives, I want them to differ in the idea itself. One might use a situation, another a sound, and a third a dialogue. That gives us a meaningful choice.

Let's try a fictional 20-second radio ad for a bicycle workshop.

First, what should the listener understand?

I would start with a short brief. Here, the main message is that people can book a bicycle check. It also sets the limits: we have no information about repair speed, prices or discounts.

Audience: people who ride a bicycle every day.
Situation: a small fault is postponed until it disrupts a journey.
Message: the workshop takes bookings for bicycle checks.
Action: call to arrange a visit.
Format: a 20-second radio ad.
Limits: do not invent prices, discounts or repair deadlines.

For a real task, I would add the approved name and contact details. I would read the finished script aloud: twenty seconds on paper still needs checking in delivery.

Three directions I would like to explore

I would give AI a more specific task:

Use this brief to develop three ideas:
- an everyday situation involving a postponed bicycle problem;
- a sound contrast;
- a short dialogue where someone else notices the problem.

For each, write the idea in one sentence, a script outline,
a call to action and one possible weakness.

The everyday situation could feature someone saying, for the third day, “I'll fix that squeak later.” It is a recognisable habit. The setup just needs to leave enough time for the rest of the ad.

The sound direction could use the rhythm of a ride, an intrusive noise and a pause before the message. I would watch the implied promise here: the sound should not suggest that every fault will be fixed immediately.

In the dialogue, one person says the bicycle is fine. The other asks: “Then why are you pushing it?” That seems like a starting point worth developing. We still need a way to make the workshop memorable, since the joke could overshadow its name.

These are outlines, but they already let us choose a structure to develop and see where each might run into trouble.

Before polishing the words

I would take one direction and check four things: whether the offer is understandable, whether the next action is clear, whether we can produce it, and whether it makes any unsupported promises.

Then I would play it to a few people who have not seen the brief. “What is being offered?” and “What would you do next?” are simple questions I would want answered.

That does not establish the whole campaign's effectiveness. It can reveal where the idea gets lost, though. That is the role I would give AI in creative work: helping us reach an idea worth testing and developing further.