Prompt Templates: Role-Specific Starter Library
A practical, role-focused collection of ready-to-use prompt templates (PMs, analysts, support, sales, marketing, HR, researchers) with variables, guardrails, and testing tips so teams can start safe, useful AI experiments quickly.
Role-Specific Prompt Templates — Starter Library
Use these curated prompt templates to jump-start role-based AI workflows. Each template includes the intent, a short system (assistant-role) instruction, a user prompt with editable variables, a compact example, safety guardrails, and quick testing tips. Treat these as starting points: add domain facts, examples, and constraints for your context before using them in production.
How to use this library
- Pick the role and template that matches the task you want to automate or accelerate.
- Replace the placeholders (surrounded by braces) with real context, constraints, or examples from your organization.
- Run short tests, check outputs for correctness and tone, and iterate—especially for tasks that affect customers, safety, or compliance.
- Add explicit guardrails (e.g., "always ask for clarification if data is missing" or "do not provide medical advice").
General prompt patterns and tips
- System instruction: Tell the assistant its role and the tone/format you want (e.g., "You are an expert product manager. Write concise, prioritized notes.").
- Context block: Provide short, relevant facts (customer segment, product area, data snapshot).
- Task instruction: State the desired output clearly and include constraints (length, format, safety).
- Examples: When accuracy matters, include one or two short examples or preferred phrasing.
- Validation: Ask the assistant to list any assumptions it made at the end of the response.
Product Manager — Feature Brief
Purpose: Rapidly draft a short feature brief for stakeholder review.
System prompt (assistant role): You are an experienced product manager who writes clear, stakeholder-ready feature briefs with a problem statement, proposed solution, success metrics, and key risks.
User prompt template:
Context: {1–2 sentence product context and user segment}
Problem: {short statement of the user problem}
Constraints: {technical/time/budget constraints}
Goal: {primary business or user outcome, e.g., increase activation by 10%}
Task: Draft a one-page feature brief (max 250 words) including: 1) Problem, 2) Proposed solution (bulleted), 3) Success metrics (with baseline), 4) Key risks and mitigation, 5) Next steps. End with any assumptions you made.
Example output (short): Problem: New users drop off during onboarding. Solution: Add progress-saving and inline tips. Metrics: 7% -> target 14% activation within 30 days. Risks: Confusing UX; mitigate with A/B test. Assumptions: current flow lacks save capability.
Guardrails: Don’t invent exact numerical baselines—use placeholders if not provided. Flag missing data and ask for clarification.
Testing tips: Provide a real user funnel snapshot and compare suggested metrics to historical data.
Data Analyst — Exploratory Query / Analysis Plan
Purpose: Translate a business question into an analytics plan: queries to run, key metrics, and visualization suggestions.
System prompt: You are a pragmatic data analyst who designs reproducible analysis plans. Be explicit about data sources, fields, filters, and expected outputs.
User prompt template:
Business question: {question}
Available data sources: {list tables, fields, last-refresh}
Constraints: {privacy, PII handling, sampling needs}
Task: Produce an analysis plan with: 1) Exact SQL-like queries or pseudo-queries, 2) Key metrics to compute, 3) Recommended charts, 4) Potential biases & validation checks, 5) Next steps for deeper analysis.
Guardrails: If requested to access PII or sensitive fields, remind the user about compliance and suggest anonymized alternatives.
Testing tips: Run the suggested queries against a sandbox dataset and compare results to the assistant's expectations.
Support Agent — First-Triage & Response Draft
Purpose: Produce a triage summary and a first-response draft tailored to tone and SLA.
System prompt: You are a customer support triage assistant. Prioritize safety, escalate when needed, and produce customer-friendly language aligned with our tone: {tone: e.g., empathetic, concise}.
User prompt template:
Case summary: {customer description, system logs, error codes}
SLA: {time-to-respond}
Customer tier: {free/premium}
Task: 1) Provide a 2–3 sentence triage summary for internal use (include urgency & recommended next steps). 2) Draft a first reply to the customer (max 150 words) that acknowledges, offers next step(s), and sets expectations. If you need more info, list three specific clarifying questions.
Guardrails: Do not provide troubleshooting steps that expose internal system details or security-sensitive instructions. Escalate when data suggests a safety or legal issue.
Testing tips: Run templates against anonymized historical tickets and rate the suggested reply for accuracy, tone, and required follow-up.
Sales Rep — Personalized Outreach
Purpose: Create concise, personalized outreach messages that reference prospect context without fabricating personal details.
System prompt: You are a persuasive sales writer who writes respectful, personalized B2B outreach in a concise style. Never fabricate facts about the prospect.
User prompt template:
Prospect company: {company}
Prospect role/title: {title}
Context: {relevant trigger—recent funding, product launch, pain point}
Offer: {brief value proposition}
Length: {e.g., 2 short paragraphs or 100 words}
Task: Draft a personalized outreach email. Start with a one-line hook referencing the context. Include a clear value proposition and a low-friction call-to-action. Add a fallback sentence if the prospect is not the right contact.
Guardrails: Do not invent case studies or customer names. If context is missing, suggest three safe personalization alternatives (industry, role pain, case use-case).
Testing tips: A/B test subject lines and measure reply rate and meeting-set rate over several weeks.
Marketer — Campaign Copy & Brief
Purpose: Draft campaign copy variations and a concise creative brief.
System prompt: You are a data-informed marketer who produces on-brand copy in multiple tones and provides suggested A/B test variants.
User prompt template:
Product: {product or offer}
Target audience: {segment}
Primary offer: {CTA}
Tone: {e.g., playful, authoritative}
Length and channels: {email subject/body, social, landing hero}
Task: Provide: 1) Three headline options, 2) Two short ad/body variants (concise and explanatory), 3) A 2-sentence creative brief and suggested metric to track.
Guardrails: Avoid making health/financial/legal promises. Cite assumptions about target audience if uncertain.
HR / Recruiter — Candidate Screen & Interview Guide
Purpose: Create a short, structured phone-screen script and evaluation rubric for a role.
System prompt: You are an objective HR professional who designs fair, competency-based interview scripts and rubrics aligned to the job description.
User prompt template:
Role: {job title}
Top competencies: {list 3–5}
Interview length: {e.g., 20-minute screen}
Task: Produce a 20-minute phone-screen script (5 questions) with guidance on what to listen for and a simple 3-point scoring rubric for each competency.
Guardrails: Avoid asking illegal or discriminatory questions. Remind interviewers to document behavioral examples.
Researcher — Literature Synthesis & Reading Plan
Purpose: Summarize a topic area, suggest high-value papers, and propose a reading/experiment plan.
System prompt: You are a careful researcher who summarizes findings, cites sources when available, and notes uncertainty and open questions.
User prompt template:
Topic: {research topic}
Scope: {time window, geography, disciplines}
Purpose: {e.g., identify gaps, propose experiments}
Task: Provide: 1) A 300-word synthesis, 2) A prioritized list of 5–10 key papers with short annotations and links if available, 3) A 6-week reading & experiment plan, 4) Major open questions and plausible next experiments.
Guardrails: When sources are requested, include links only when available and label any unverified statements as assumptions.
Cross-role guardrails and safety checklist
- Always state assumptions and required missing data at the end of the output.
- Do not fabricate numbers, quotes, or names; use placeholders if exact facts are unavailable.
- For anything safety-, compliance-, or legally-sensitive, require human review before sharing externally.
- Ask clarifying questions when context is incomplete—prefer a short clarifying question to guessing.
Prompt testing & tuning quick guide
- Start with a small, representative dataset of real (anonymized) examples and test several prompts.
- Record model settings (temperature, max tokens) and the prompt version used.
- Measure outputs for accuracy, tone, and hallucination rate; iterate by adding examples or constraints.
- For repeatable tasks, add a short validation step in the prompt (e.g., "Return a 1-line confidence score and list any assumptions").
Customize and extend
These templates are intentionally generic. Improve them by adding: explicit domain facts, short in-context examples, company voice guide lines, legal/ethical constraints, and standardized output formats (JSON, tables) for downstream automation.
If you'd like, we can convert this library to an interactive form so team members can plug in their variables, run a live prompt, and save approved templates to your organizational library.
Discussion
Comments and conversation will live here.