Production-ready prompts for PMs at every stage: discovery, spec writing, prioritization, and stakeholder communication. Each prompt is structured so Claude or ChatGPT produces a usable deliverable, not a generic text block.
Works with Claude, ChatGPT, Gemini, or any LLM. Copy, fill in the brackets, paste.
Paste any of these prompts into the improver →
Tailor for your product, team, and specific situation.
Builds a hypothesis-driven guide that avoids leading questions.
Create a user interview guide for the following research context:
Product: [describe your product in 1-2 sentences]
User segment: [e.g. "B2B ops managers at 50-500 person companies"]
Core hypothesis to test: [e.g. "Users abandon the workflow because the approval step is too opaque"]
Interview duration: [30 / 45 / 60 minutes]
Generate:
1. A 1-sentence screener question to confirm the participant fits the segment
2. 2 warm-up questions (context-setting, not about the product)
3. 5 open-ended discovery questions that test the hypothesis without leading
4. 2 "show me" tasks if the session includes a prototype or existing product
5. 1 closing question to surface what participants wish you had asked
Do not ask yes/no questions. Avoid "would you" framing. Flag any question that could bias the response.
Extracts themes and evidence from raw interview notes.
Synthesize the following user interview excerpts into research findings.
Excerpts (paste verbatim quotes or paraphrased notes from [N] interviews):
```
[paste your notes here — one interview per block or label each quote with a participant ID]
```
Produce:
1. Top 5 themes, each supported by at least 2 quotes. Format: [Theme title] — [1-sentence description] — Supporting quotes: [Q1, Q2]
2. The single most surprising finding
3. The top user job-to-be-done implied by the data
4. 3 open questions this research raised that require follow-up
Do not add interpretation beyond what the quotes support.
Turns a vague problem into a crisp, testable statement.
Sharpen this problem statement for a product brief:
Current statement: "[paste your draft]"
Context:
- Who has this problem: [user segment]
- Evidence we have: [e.g. "22% drop-off at step 3, 15 support tickets/week about X"]
- Current workaround users use: [describe it]
- Why the current workaround is inadequate: [describe it]
Rewrite the problem statement to meet these criteria:
1. Specific user (not "users")
2. Specific situation (when/where the problem occurs)
3. Specific pain (consequence of the problem)
4. No implied solution
5. Under 40 words
Provide 3 alternative versions, ordered from most conservative to most ambitious framing.
Turns a feature description into Given/When/Then criteria.
Write acceptance criteria for this feature in Given/When/Then format.
Feature: [describe the feature]
User story: As a [user type], I want to [action] so that [outcome]
Key constraints: [e.g. "must work on mobile, must support 10k concurrent users, must not change the existing API contract"]
Out of scope: [list what this feature explicitly does not do]
Generate:
- The primary happy-path scenario (1 Given/When/Then)
- 3 edge cases with their expected behavior
- 2 error states and expected UI/UX response
- 1 non-functional requirement (performance, accessibility, or security) relevant to this feature
Format as a numbered checklist that an engineer can use for a PR description.
Surfaces stakeholder questions before your review meeting.
I am writing a PRD for [feature/product]. Here is a summary:
[Paste your PRD summary or key sections — problem statement, proposed solution, success metrics]
Generate the 10 most likely questions that stakeholders (engineering lead, design lead, legal, finance, executive sponsor) would ask in a review meeting.
For each question:
- Label which stakeholder is most likely to ask it
- Suggest a 2-sentence draft answer I can add to the PRD
Order by the questions most likely to block approval first.
Finds the gaps in your acceptance criteria before engineering does.
Review these feature requirements and identify missing edge cases.
Requirements:
```
[paste your current requirements or acceptance criteria]
```
Product context: [1-2 sentences]
User types who will interact with this feature: [list them]
Integrations or dependencies: [e.g. "payment provider, email service, mobile app"]
Find:
1. User behaviors the requirements don't handle (what happens if the user does X?)
2. System states the requirements assume but don't specify (what if the user is offline / session expired / data is missing?)
3. Permissions or role-based access gaps
4. Data validation edge cases (empty input, malformed input, max length)
5. Race conditions if multiple users interact with the same data
List each gap as a one-line question the engineering team will ask if not addressed.
Scores your backlog and challenges its own estimates.
Score the following feature candidates using RICE prioritization.
Feature list:
1. [Feature A — 1-sentence description]
2. [Feature B — 1-sentence description]
3. [Feature C — 1-sentence description]
[add more as needed]
Context:
- Product type: [SaaS / consumer app / internal tool]
- Monthly active users: [N]
- Team size: [engineers + designers]
- Current top metric: [e.g. "weekly active users", "NPS"]
For each feature, estimate:
- Reach (users affected per quarter, 1-10 scale)
- Impact (effect on the top metric, 1-3 scale: 3=massive, 2=significant, 1=minimal)
- Confidence (% certain about estimates: 100% / 80% / 50%)
- Effort (person-weeks)
- RICE Score = (R × I × C) / E
After scoring, identify: (a) which estimate you're least confident in and why, (b) which features are likely undervalued by RICE because they're infrastructure or retention work.
Prepares you for the pushback before the roadmap review.
Prepare me for stakeholder objections to my proposed roadmap.
My top 3 priorities for next quarter:
1. [Priority 1 — brief description and rationale]
2. [Priority 2 — brief description and rationale]
3. [Priority 3 — brief description and rationale]
What I'm deprioritizing: [list 2-3 things stakeholders might expect to see]
Stakeholders attending the review: [e.g. "CEO, Head of Sales, Head of Engineering, Head of Design"]
For each stakeholder, generate:
- Their most likely objection to my priorities
- The strongest version of that objection (steelman it)
- A 2-sentence response that acknowledges the concern and explains the tradeoff
Then: what is the single most likely thing to derail this roadmap that isn't on this list?
Condenses a complex initiative into a 5-bullet exec update.
Write an executive summary for this product initiative.
Initiative: [name]
Background (for context, not for inclusion): [2-3 sentences on what this is and why]
Current status: [e.g. "in discovery", "in build", "launched 2 weeks ago"]
Key metrics so far: [paste numbers if available]
Decision or input needed from leadership: [what specifically do you need from this audience?]
Write a 5-bullet executive summary:
1. What we're doing and why (1 sentence)
2. Current status (1 sentence)
3. Key results or learnings so far (1-2 sentences with numbers)
4. Top risk or open question (1 sentence)
5. Ask: what decision or resource is needed (1 sentence)
Tone: direct, no jargon, no passive voice. Under 120 words total.
Writes internal and external announcements that drive adoption.
Write two launch announcements for [feature name]:
Context:
- What it does: [1-2 sentences]
- Who it helps: [user segment]
- The main pain it solves: [describe it]
- How to access it: [where in the product, any required setup]
- Any caveats or known limitations: [list them]
Version 1: Internal Slack announcement (to customer-facing teams)
- Under 100 words
- What it is, who it's for, how they can try it
- One concrete example or screenshot description
- Link to docs (use [docs-link] as placeholder)
Version 2: In-app or email announcement (to end users)
- Under 80 words
- Lead with the user benefit, not the feature name
- One clear CTA
- No jargon
Format as two separate blocks labeled "Internal" and "External".
What are the best AI prompts for product managers?
The best PM prompts share three qualities: (1) They give context upfront — product type, user segment, current metrics, and what decision needs to be made. (2) They ask for structured output — a prioritized list, a framework, or a template the PM can fill in, not a wall of prose. (3) They constrain the scope — "3 hypotheses" or "5 interview questions" produces a usable deliverable; open-ended requests produce generic text. Prompts that include the business context (company size, growth stage, user type) consistently outperform those that treat the AI as a general-purpose text generator.
How can product managers use AI for user research?
PMs can use AI for three user research tasks: (1) Interview guide creation — provide your hypothesis and user segment, ask for 10 open-ended questions that test that hypothesis without leading the participant. (2) Synthesis — paste verbatim quotes from 10+ interviews and ask the AI to identify the top 5 themes with supporting evidence. (3) Persona sharpening — describe your current persona and ask what critical details are missing that would change product decisions. AI accelerates research synthesis and guide creation, but cannot replace talking to real users — it has no access to your specific customers or market.
How do I use AI to write better PRDs?
For PRD writing: (1) Use AI to draft the problem statement — describe the user pain and your evidence, ask for a 3-sentence problem statement that a non-technical executive could understand. (2) Use AI to stress-test your requirements — paste your acceptance criteria and ask "what edge cases or user behaviors would break these requirements?" (3) Use AI to generate the FAQ section — after drafting, ask "what are the 10 most likely questions stakeholders will ask about this spec?" AI works best as an editor and devil's advocate on PRDs you write, not as the author. Specs that require judgment about tradeoffs and business context need PM ownership.
What AI prompts help with roadmap prioritization?
Useful prioritization prompts: (1) RICE scoring — paste your feature list and ask the AI to score each on Reach, Impact, Confidence, and Effort on a 1-10 scale with brief justification. Then challenge the scores with "which of these RICE estimates is most likely wrong and why?" (2) Dependency mapping — paste your backlog and ask "which items are blockers for other items? Identify 3 sequences where shipping A enables B enables C." (3) Stakeholder objection prep — describe your top 3 roadmap priorities and ask for the strongest argument against each. AI is good at structured frameworks; you supply the domain knowledge about what the numbers actually mean.
How can PMs use AI for competitive analysis?
AI helps with competitive analysis structure, not data collection — it cannot browse live competitor sites or access pricing pages. Use it for: (1) Building comparison frameworks — "I need to compare 5 project management tools across PM workflows. Generate a comparison matrix with the 8 most decision-relevant dimensions for a 50-person tech team." (2) Positioning articulation — paste competitor messaging and ask "what positioning gaps does this leave that we could own?" (3) Battlecard drafts — give the AI factual differences between your product and a competitor and ask it to generate the top 3 objections and responses. Fill in real data yourself.