Business prompts that hold up — marketing, sales and HR
Most prompt collections are lists of one-liners that produce generic output. What actually works in a business context is a structure — role, context, constraints, examples, output format — applied consistently.
AI Tools Hub Editorial · published · updated
The structure
- Role — who the model should write as, specifically enough to matter
- Context — the situation, the audience and what has already been tried
- Constraints — length, tone, what must not appear
- Examples — one or two pieces of real output you consider good
- Format — exactly how you want it back
The example is the part people skip and the part that changes the result most. One real sample of your own work does more than a paragraph describing it.
Marketing
You are writing for [audience] who already know [X] and are sceptical about [Y]. Write three subject lines for an email announcing [thing]. Constraints: under 50 characters, no exclamation marks, no superlatives. Here are two subject lines that performed well for us: [examples]. Return as a numbered list with a one-line rationale for each.
Sales
Here is a transcript of a discovery call: [transcript]. Extract: the stated problem, the unstated problem, who the decision maker appears to be, the objection most likely to end this, and three questions to ask next. Be concrete; do not speculate beyond what is in the transcript.
HR
Rewrite this job description for [role] to be specific about what the person will actually do in the first ninety days. Remove anything that could appear in any job description. Keep it under 400 words. Current version: [text].
What makes prompts reusable
- Keep the structure and swap only the bracketed parts
- Store them where the team works, not in individual chat histories
- When one produces something good, save the output as a new example
- Review them quarterly — model behaviour shifts and prompts drift out of calibration
Do these work the same in Japanese?
The structure does. Specify the register explicitly — polite form, plain form, who the reader is — because Japanese drifts between them more easily than English.
Should we build custom GPTs for these?
Once a prompt is used by several people regularly, yes. Until then the overhead exceeds the benefit.
Related
More articles
Opinion
Why AI rollouts fail in Japanese companies — four structural reasons
For business
Enterprise AI rollout checklist — SOC 2, ISO 27001, DPAs and personal data
For business
Japanese-developed LLMs for business — ELYZA, Sakana AI, NTT tsuzumi and PLaMo
Comparison
AI coding tools compared — Cursor, Copilot, Claude Code, Windsurf and Codeium