Act as a [RELEVANT EXPERT] helping me brainstorm options for [OBJECTIVE]. The audience or user is [AUDIENCE]. These are the facts, constraints, and non-negotiables: [CONSTRAINTS]. Existing attempts, competitors, or ideas to avoid repeating are: [WHAT ALREADY EXISTS]. Judge success by [SUCCESS CRITERIA]. Work in three passes. 1. Frame the problem: state the job to be done, the most important constraint, and 3 assumptions that could materially change the answer. Do not treat assumptions as facts.
2. Generate 18 ideas across distinct approaches. Group them into 4-6 named territories so the list does not contain minor variations of one concept. For each idea, provide a short name, a 1-2 sentence description, the audience benefit, and the main implementation or adoption risk. Include at least 3 conservative ideas, 3 unconventional ideas, and 3 ideas that reduce scope or remove a step rather than adding a feature.
3. Prioritize: score the 6 strongest ideas from 1-5 for expected impact, effort, confidence, and fit with [SUCCESS CRITERIA]. Explain each score briefly. Recommend the best 3 to test first, with a low-cost first experiment, the signal to measure, and a reason the test could mislead us.
Do not invent user research, market size, legal clearance, technical capability, or results. Avoid vague ideas that cannot be tested within a defined time or budget. If my objective has an unstated decision, surface it rather than choosing silently.
Before answering, self-check that the ideas are genuinely different, every recommendation respects the supplied constraints, and the shortlist is not merely the most familiar option. If a required input is missing, ask up to 3 clarifying questions; otherwise, provide the full brainstorm.Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [RELEVANT EXPERT] | Enter the expertise needed, such as B2B product strategist, curriculum designer, or event producer. | B2B SaaS onboarding strategist |
| [OBJECTIVE] | State the problem or opportunity you want ideas for. | increase the percentage of new workspace admins who invite a teammate in their first week |
| [AUDIENCE] | Describe the people the ideas must serve. | operations managers at 20-200 person professional-services firms |
| [CONSTRAINTS] | List budget, timeline, team, technical, brand, legal, or operational limits. | two engineers for two weeks; no in-app chat; cannot offer monetary incentives; must work on desktop and mobile |
| [WHAT ALREADY EXISTS] | List current approaches, failed attempts, competitors, or ideas you do not want repeated. | a generic welcome email, a checklist with five setup tasks, and a persistent invite banner that has low click-through |
| [SUCCESS CRITERIA] | Define what a successful idea must achieve and how you will judge it. | a measurable lift in qualified teammate invites without increasing support tickets or delaying first-value completion |
How to use
- Write one decision-focused objective and include the constraints that would make an otherwise good idea unusable.
- List existing attempts so the model can explore new territories instead of renaming familiar ideas.
- Inspect the scores and experiments for assumptions that need evidence before work begins.
- Follow up with: “Take the top three ideas and write a one-week experiment brief with owner, metric definition, guardrail, and stop condition.”
Variations
Naming workshop
Use this to develop names for a product, feature, program, or event.
Generate naming directions for [THING BEING NAMED] for [AUDIENCE]. It should signal [DESIRED ASSOCIATIONS] and avoid [UNWANTED ASSOCIATIONS]. Brand, linguistic, and legal constraints: [CONSTRAINTS]. Existing names or competitors to avoid resembling: [EXAMPLES]. Create 30 candidates in 5 distinct naming territories. For each, give the name, pronunciation if unclear, rationale, and one likely confusion or risk. Then shortlist 8 using memorability, clarity, distinctiveness, and fit. Do not claim trademark availability or invent language meanings. Self-check for duplicated constructions, hard-to-spell names, and names that contradict the constraints.
Campaign concepts
Use this to explore an audience campaign before choosing creative execution.
Brainstorm campaign concepts for [OFFER OR CAUSE] aimed at [AUDIENCE] to achieve [BEHAVIOR CHANGE]. The evidence, insight, and product truth we can use are [FACTS]. Constraints are [CONSTRAINTS], and prior campaign territory to avoid is [WHAT EXISTS]. Generate 15 concepts across at least five creative territories. For each, give the core insight, one-line concept, execution examples for [CHANNELS], and the risk of misunderstanding or backlash. Prioritize five based on credibility, distinctiveness, and feasibility. Do not invent audience research or promise performance. Self-check that the concept works without a misleading claim and each territory is meaningfully different.
Decision options
Use this when you need to choose a direction under real tradeoffs.
Develop decision options for [DECISION] for [STAKEHOLDERS]. The objectives are [OBJECTIVES], constraints are [CONSTRAINTS], and known facts are [FACTS]. Produce 8 materially different options, including a status-quo option and one that reduces scope. For each, state what it entails, upside, downside, prerequisite, cost or effort range if supportable, and the condition under which it should be rejected. Then build a comparison table using [DECISION CRITERIA] and recommend the 2 options worth validating first. Do not create estimates or stakeholder preferences not supported by my facts. Self-check that the recommendation follows the stated criteria rather than novelty.
Tips
- Start with the decision or behavior to change, not a request for “ideas”; a clear outcome gives the brainstorm an evaluable center.
- Require idea territories and risks so the model cannot pad the list with cosmetic variations.
- Include the status quo and scope-reduction options; they often reveal a simpler route than adding a new initiative.
- Treat prioritization scores as hypotheses until they are tied to evidence, owners, and a testable signal.
FAQ
How many ideas should I ask AI to generate?
Ask for enough ideas to cover several distinct approaches, then force a shortlist. Eighteen ideas grouped into territories is usually more useful than an unstructured list of 100.
Can AI validate which idea will work?
It can identify assumptions and propose tests, but it cannot establish demand or feasibility without evidence. Use experiments, customer conversations, and technical review for validation.
How do I stop AI from giving generic ideas?
Provide constraints, failed approaches, a specific audience, and a measurable success criterion. Require each idea to include its mechanism and main risk.