A chatbot gives you what you ask for, not what you hoped for. Ask for “some ideas” and you get three safe ones. Ask for a range — from the obvious to the bold — and you get the directions a good creative team would put on the wall.
Give me some ideas for our summer campaign.
Give me 10 directions for our summer campaign, ordered from the safest to the boldest. At least three should be ones our category would never run. One line each.
The prompt
Copy-ready prompt
Give me [NUMBER — e.g. 10] directions for [CAMPAIGN / BRIEF]. Order them from the safest to the boldest. At least [NUMBER] should be ideas our category would never run. For each: one line for the idea, and one line on what it would ask the brand to risk. Do not stop at the obvious ones.
Why it works
Anthropic’s guide is direct about this: if you want “above and beyond” behaviour, “explicitly request it rather than relying on the model to infer this from vague prompts.” Its own example turns “Create an analytics dashboard” into a request to include as many relevant features as possible — the same move, applied to ideas.
Use it for
Concept rounds, naming, headlines, campaign territories. Asking for the risk of each direction keeps the bold ones honest, and gives you the sentence you will need in the client meeting.
Sources
- Anthropic — Claude prompting best practices — platform.claude.com (accessed 29 Sep 2026)
More tricks and templates in the Prompt Lab. Want this applied to your brand’s brief? Start a project.