The best ChatGPT and Claude prompts for product managers do one specific job each: draft a PRD from a one-line idea, split a feature into user stories with Given/When/Then acceptance criteria, score a backlog with RICE, or turn interview notes into a persona. This guide gives you 25 of them, grouped by the work an AI product manager workflow actually covers, plus four reusable templates (PRD, user story, now/next/later roadmap, persona) that you can paste into any model. Every prompt uses [brackets] for the fields you fill in. They work in ChatGPT, Claude, Gemini and Copilot without changes.
PRD generator
AI PRD Generator Skill
Turns a feature idea plus your context into a complete PRD with problem statement, goals, requirements, metrics and open questions.
$29
Get the skill →
How to get useful output from these prompts
Three habits make the difference between a generic answer and a document you can actually ship.
-
Give the model the same context you would give a new hire. Product, users, business model, the constraint you are working under. One paragraph is enough, but it has to be there.
-
Ask for a structure, not a vibe. "Write a PRD" gets you an essay. "Fill this PRD template, mark anything you had to assume with [ASSUMPTION]" gets you a draft you can review in ten minutes.
-
Iterate in the same thread. Once the model has the context, follow-up prompts ("now write the user stories for section 4") are cheap and consistent. Do not start a new chat for every artefact.
If you want the model to hold this behaviour permanently instead of re-pasting instructions, that is what a skill file does. Browse the Claude skills collection for role-based options, or keep reading for the free prompts.
Group 1: Discovery and research (prompts 1 to 5)
Prompt 1: Interview synthesis
Here are notes from [number] customer interviews about [topic]: [paste notes]. Group the findings into themes. For each theme give a name, how many interviewees mentioned it, one representative quote, and the underlying job the user is trying to get done. End with the three themes you would investigate further and why.
Prompt 2: Persona from evidence
Using this research: [paste interview themes, survey data or support tickets], write a persona for [segment]. Use this structure: name and role, context (company size, tools, team), goals, top three frustrations, how they evaluate a solution, what would make them churn, one verbatim quote. Mark anything not supported by the evidence with [ASSUMPTION].
Prompt 3: Problem statement
We are considering building [feature] for [persona]. Write five alternative problem statements in the form "[persona] needs a way to [job] because [insight]". For each, say what evidence would confirm it is a real problem and what evidence would kill it.
Prompt 4: Competitor teardown
Compare how [competitor A], [competitor B] and [competitor C] handle [use case]. Build a table: capability, how each product does it, where it falls short for [our persona]. Then list three positioning options for us. Only state facts you are confident about and flag anything you are unsure of.
Prompt 5: Interview script
Write a 30-minute discovery interview script for [persona] about [problem area]. Open with two warm-up questions, then eight open questions that ask about past behaviour rather than hypothetical preferences, then two closing questions. Add a note under each question on what a strong answer would tell us.
Group 2: PRDs and specs (prompts 6 to 10)
Prompt 6: PRD from a one-liner
Write a PRD for [feature] in [product]. Context: [users, business model, constraint]. Use these sections: problem, goals and non-goals, success metrics, user stories (summary), functional requirements (numbered, each testable), non-functional requirements, dependencies, risks, open questions. Mark assumptions with [ASSUMPTION]. Keep it under 900 words.
Prompt 7: PRD review
Review this PRD as a sceptical engineering lead: [paste PRD]. List every requirement that is ambiguous, untestable or contradicts another one. For each, quote the line and propose a rewrite. Then list the questions you would ask before estimating.
Prompt 8: Non-goals and scope cuts
Here is the scope of [feature]: [paste]. Propose the smallest version that still tests the core hypothesis "[hypothesis]". List what moves to non-goals, what moves to a later release, and what risk we take by cutting each item.
Prompt 9: Edge cases
For the requirement "[paste one requirement]", list edge cases across: empty or missing data, permissions, concurrency, offline or slow network, localisation, accessibility, and abuse. For each, say the expected behaviour or write "needs decision".
Prompt 10: Technical brief for engineers
Turn this PRD into a one-page technical brief for the engineering team: [paste PRD]. Include the user-facing behaviour, data that must be stored or changed, integrations touched, and the acceptance tests that define done. Do not propose an architecture, only what the system must do.
Group 3: User stories and acceptance criteria (prompts 11 to 15)
Prompt 11: Epic to stories
Split this epic into user stories: [paste epic]. Format each as "As a [persona], I want [action] so that [outcome]". Each story must be deliverable in one sprint and independently valuable. Add a one-line note on why it is a separate story.
Prompt 12: Given/When/Then criteria
Write acceptance criteria for this story: [paste story]. Use Given/When/Then. Cover the happy path, at least two failure or edge scenarios, and one permissions scenario. Each scenario must be verifiable by a tester without asking the PM.
Prompt 13: INVEST check
Evaluate these stories against INVEST (independent, negotiable, valuable, estimable, small, testable): [paste stories]. Give each a pass or fail per letter, explain any fail in one sentence, and rewrite the failing stories.
Prompt 14: Story map
Build a user story map for [journey, e.g. onboarding a new team]. Row one: the backbone activities in order. Under each activity: the tasks a user does. Then mark which tasks belong to release 1, 2 and 3, with the outcome each release proves.
Prompt 15: Bug to story
Rewrite this bug report as a user story with acceptance criteria and a severity rating: [paste bug report]. Separate what the user experienced from what you think the cause is. Include the reproduction steps as a Given/When/Then scenario.
Group 4: Roadmaps and prioritisation (prompts 16 to 20)
Prompt 16: RICE scoring
Score these backlog items with RICE: [paste items with any data you have]. For each, estimate reach (users per quarter), impact (0.25 to 3), confidence (percentage) and effort (person-weeks), show the calculation, and rank them. Flag every estimate with low confidence and say what data would raise it.
Prompt 17: ICE quick sort
Give an ICE score (impact, confidence, ease, each 1 to 10) for these ideas: [paste list]. Return a table sorted by total, with one sentence of reasoning per idea. Then tell me which two items you would challenge the scores on and why.
Prompt 18: Now/Next/Later roadmap
Turn this prioritised list into a now/next/later roadmap: [paste list]. Now: what we are committing to this quarter. Next: what we expect to start after. Later: directions we believe in but have not committed to. Write each item as an outcome, not a feature name, and add the metric that tells us it worked.
Prompt 19: Trade-off memo
We have to choose between [option A] and [option B] for [quarter]. Context: [team size, goals, constraints]. Write a one-page decision memo: the decision, the options considered, the criteria, how each option scores, the recommendation, and what would make us reverse it.
Prompt 20: Saying no
A stakeholder ([role]) has asked for [request]. It does not fit the current roadmap because [reason]. Draft a reply that acknowledges the need, explains the trade-off in plain terms, offers what we can do instead, and states when we will revisit. Keep it under 150 words and not defensive.
Group 5: Launch, metrics and stakeholder comms (prompts 21 to 25)
Prompt 21: Success metrics
For [feature] aimed at [persona], propose one north star metric, three supporting metrics and two guardrail metrics that would show harm. For each: definition, how it is measured, the baseline we need first, and a target for 90 days after launch with your reasoning.
Prompt 22: Launch checklist
Create a launch checklist for [feature] with owners and timing: two weeks before, one week before, launch day, one week after. Cover engineering readiness, QA, support enablement, documentation, marketing, sales, analytics and rollback. Mark items that block launch.
Prompt 23: Release notes in three voices
Write release notes for [feature] in three versions: one for end users (benefit first, no jargon, under 80 words), one for internal sales and support (what changed, known limitations, how to answer common questions), and one for the changelog (precise, technical). Changes: [paste].
Prompt 24: Weekly stakeholder update
Write this week's product update from these notes: [paste notes]. Structure: headline (one sentence), shipped, in progress with expected dates, blocked and what we need, metrics that moved, decisions needed from the reader. Under 250 words. No filler.
Prompt 25: Post-launch retrospective
Here is what happened after launching [feature]: [paste metrics and feedback]. Compare results against the PRD goals: [paste goals]. Write a retrospective: what we predicted, what happened, why the gap exists, what we would do differently, and the next decision to make (double down, iterate or stop).
Templates you can paste into any prompt
These four templates are the structures the prompts above refer to. Paste one into the chat, add "fill this in for [feature]" and the model will keep the headings.
PRD template
-
Problem: who has it, how we know, what it costs them today
-
Goals: two or three outcomes with a metric each
-
Non-goals: what this release deliberately does not do
-
Users and personas: primary and secondary
-
User stories: the short list, detailed stories live in the backlog
-
Functional requirements: numbered, each one testable
-
Non-functional requirements: performance, security, accessibility, localisation
- Dependencies and risks
-
Open questions: with an owner and a date
User story template
As a [persona], I want [action] so that [outcome].
Acceptance criteria:
-
Given [starting state] when [action] then [observable result]
-
Given [edge case state] when [action] then [expected handling]
-
Given [user without permission] when [action] then [access denied behaviour]
Roadmap template (now/next/later)
| Column |
Meaning |
What each item includes |
| Now |
Committed, in progress this quarter |
Outcome, owner, success metric, target date |
| Next |
Expected to start after Now, not yet committed |
Outcome, open questions, dependency on Now |
| Later |
Directions we believe in, subject to learning |
Outcome, the hypothesis it rests on |
Persona template
-
Name and role, plus the context they work in (team size, tools, budget authority)
-
Goals: what they are trying to achieve this quarter
-
Frustrations: top three, each tied to evidence
-
How they evaluate a solution and who else is involved in the decision
- What would make them leave
-
One quote from real research, not invented
Which tool for which task
Prompts are free and fast. A generator gives you a structured output every time without re-explaining the format. A role skill changes how the model thinks for the whole session. Here is where each fits.
Isolde - Product Owner AI Skill ($29): a product owner persona that writes and refines backlog items, acceptance criteria and sprint goals in a consistent voice.
Cassius - AI and Data Product Manager AI Skill ($29): built for PMs shipping data and AI features, covering model evaluation criteria, data requirements and metric design.
Business and Operations Prompt Library ($27): a broader set of prompts for memos, updates, meeting notes and process documents that sit around product work.
If you also run delivery, the prompts for project managers page covers status reports, risk logs and timelines. For general-purpose prompt writing, see ChatGPT prompts.
Frequently asked questions
Can ChatGPT or Claude replace a product manager?
No. The model drafts and structures; you supply the judgement, the context and the decision. The prompts above are useful because they remove the blank-page problem and enforce a structure, not because they know your users.
Do these prompts work the same in ChatGPT and Claude?
Yes. They are plain text with no tool-specific syntax. Claude tends to follow long structures closely, ChatGPT tends to be more concise, so you may need to add "keep all sections" or "shorter" depending on the model.
What is the difference between a prompt, a generator and a skill?
A prompt is a single instruction you paste. A generator is a skill file tuned for one output type, such as a PRD, so the format is consistent every time. A role skill, such as a product owner persona, changes how the model reasons across a whole session. Skills from KissMySkills are one-time purchases (.md file plus README) that work in Claude including Claude Code, ChatGPT as a custom GPT, Gemini as a Gem, and Copilot.
Should I use RICE or ICE for prioritisation?
Use ICE when you need to sort a long list quickly and do not have reach data. Use RICE when the items compete for the same engineering capacity and you can estimate reach and effort with some confidence. Both are inputs to a decision, not the decision.
What if a skill does not fit my workflow?
Every skill comes with a 30-day money-back guarantee, no questions asked.
The bottom line
Pick the five prompts that match the artefacts you produce most often, keep the templates in a note, and run them in one thread with real context. That alone will save you a few hours a week. When you find yourself pasting the same instructions every day, move that job into a generator or a role skill so the model holds the structure for you.
Bundle
Product Manager Tool Bundle - 1 AI Skill + 3 Tool Skills
One product manager role skill plus three tool skills for PRDs, stories and roadmaps, ready for Claude, ChatGPT, Gemini and Copilot.
$74.99
Get it now →
Browse all Claude skills at KissMySkills.
