AI Prompts for Product Managers

CONTEXT
We are researching [USER SEGMENT] to understand [RESEARCH QUESTION].

EVIDENCE
Use only the interview notes or transcript excerpts below as evidence:
[INTERVIEW EVIDENCE]

TASK
Cluster recurring user problems, current workarounds, and contradictions. Keep each cluster traceable to the supplied evidence.

DECISION CRITERIA
Prefer clusters supported by multiple independent observations, but preserve high-severity minority signals instead of hiding them.

CONSTRAINTS
- Do not create or rewrite user quotes.
- Do not infer frequency beyond the supplied sample.
- Separate direct evidence from interpretation.
- Mark anything not supported by the input as an assumption or open question.

OUTPUT
Return a table with: problem cluster, supporting evidence, contradictory evidence, workaround, confidence in the synthesis, and next research question.

CONTEXT
The team needs to decide [DECISION TO MAKE].

EVIDENCE
[CURRENT EVIDENCE]

ASSUMPTIONS ALREADY IDENTIFIED
[KNOWN ASSUMPTIONS]

TASK
Identify the most decision-relevant evidence gaps. Distinguish what is known, inferred, assumed, and still unknown.

DECISION CRITERIA
Prioritize gaps that could materially change the decision, not questions that are merely interesting.

CONSTRAINTS
Do not fill gaps with synthetic personas, imagined customer preferences, fabricated benchmarks, or generic market claims.

OUTPUT
Return: known facts, supported interpretations, assumptions, critical unknowns, and the smallest research step that could reduce each critical uncertainty.

CONTEXT
Product context: [PRODUCT CONTEXT]

EVIDENCE
[EVIDENCE]

TASK
Critique this problem statement:
[PROBLEM STATEMENT]

DECISION CRITERIA
A strong statement should identify the affected user, situation, observed problem, and meaningful impact without prescribing a solution.

CONSTRAINTS
- Do not invent additional user evidence or impact.
- Flag claims that are not supported by the evidence.
- Identify solution language disguised as problem language.

OUTPUT
Return: what is clear, what is unsupported, hidden assumptions, solution leakage, a tighter evidence-preserving rewrite, and the single most important question to answer before prioritization.

CONTEXT
We need to compare the following opportunities:
[OPPORTUNITIES]

EVIDENCE
[EVIDENCE]

DECISION CRITERIA
Use only these criteria:
[CRITERIA]

CONSTRAINTS
[CONSTRAINTS]

TASK
Compare the opportunities using the supplied criteria. Show where evidence is strong, weak, or missing and explain the major trade-offs.

RULES
- Do not assign numerical scores unless the required inputs were provided.
- Do not invent reach, impact, confidence, effort, revenue, or user demand.
- Do not choose a winner solely because one description is more detailed.

OUTPUT
Return a comparison table, the strongest argument for each option, uncertainty that could change the ranking, and questions the PM should resolve before deciding.

CONTEXT
Business and product context:
[BUSINESS CONTEXT]

EVIDENCE
[EVIDENCE]

STRATEGY TO REVIEW
[STRATEGY]

KNOWN NON-GOALS
[NON-GOALS]

TASK
Stress-test the strategy. Identify the target user, chosen problem, strategic choice, advantage or mechanism, trade-offs, non-goals, and assumptions.

DECISION CRITERIA
A strategy should make meaningful choices and exclusions, connect to real evidence, and remain coherent under stated constraints.

CONSTRAINTS
Do not invent market size, customer demand, competitive advantage, company capabilities, or competitor facts.

OUTPUT
Return: strategy map, contradictions, missing choices, assumptions that need evidence, strongest alternative, and what evidence would make the current strategy less credible.

CONTEXT
Objective: [OBJECTIVE]
Candidate initiatives: [INITIATIVES]
Known dependencies: [DEPENDENCIES]
Constraints: [CONSTRAINTS]

TASK
Organize the initiatives into candidate outcome-oriented themes, then challenge the proposed sequence.

DECISION CRITERIA
Evaluate strategic fit, dependency logic, evidence strength, reversibility, risk, and whether the sequence creates useful learning.

CONSTRAINTS
- Do not invent deadlines, capacity, commitments, or customer demand.
- Do not turn every initiative into a top priority.
- Treat sequencing as a hypothesis unless dependencies make it mandatory.

OUTPUT
Return candidate themes, sequencing rationale, overloaded areas, missing rationale, dependency risks, and decisions that still require PM judgment.

CONTEXT
Approved scope: [APPROVED SCOPE]
Known constraints: [KNOWN CONSTRAINTS]

DOCUMENT TO REVIEW
[PRD DRAFT]

TASK
Review the PRD for unclear requirements, unsupported claims, hidden assumptions, solution details presented as user needs, missing edge cases, and unresolved decisions.

DECISION CRITERIA
A reviewer should be able to distinguish problem, evidence, scope, non-goals, requirements, success criteria, risks, and open questions.

CONSTRAINTS
Do not invent customer evidence, business rules, technical constraints, baseline metrics, or stakeholder decisions.

OUTPUT
Return findings by severity, the exact section affected, why it matters, a suggested clarification question, and optional wording improvements that preserve uncertainty.

CONTEXT
Actor and situation: [ACTOR / CONTEXT]
Approved scope: [APPROVED SCOPE]
Known rules: [KNOWN RULES]
Known failure or edge states: [FAILURE STATES]

TASK
Draft candidate user stories and observable acceptance criteria for the approved slice. Also identify decisions that remain unresolved.

DECISION CRITERIA
Each story should represent coherent user value or behavior and criteria should describe observable outcomes, not implementation detail.

CONSTRAINTS
If a behavior, permission, business rule, or error state is not defined by the inputs, flag it as a question instead of choosing a rule.

OUTPUT
For each story return: actor/context, need, outcome, acceptance criteria, edge cases, and unresolved question.

CONTEXT
Product context: [PRODUCT CONTEXT]
Desired outcome: [PRODUCT OUTCOME]
Expected user behavior change: [USER BEHAVIOR]
Known risks: [KNOWN RISKS]

TASK
Suggest candidate success, diagnostic, and guardrail metrics. Explain what each metric would actually measure and where it could mislead us.

DECISION CRITERIA
Prefer metrics that connect to the stated outcome and behavior, are interpretable, and can change a product decision.

CONSTRAINTS
Do not invent baselines, benchmarks, expected lift, sample size, conversion rates, causal relationships, or available instrumentation.

OUTPUT
Return a table with metric candidate, role, why it may help, failure mode, segmentation question, instrumentation needed, and what evidence is required before choosing it.

HYPOTHESIS
[HYPOTHESIS]

EXPERIMENT DESIGN
[EXPERIMENT DESIGN]

METRICS
[METRICS]

KNOWN LIMITATIONS
[KNOWN LIMITATIONS]

TASK
Critique whether the experiment can answer the stated question. Identify ambiguous success/failure criteria, plausible confounders, instrumentation risks, and counter-explanations for possible outcomes.

DECISION CRITERIA
The design should connect intervention, mechanism, observable outcome, and decision rule without claiming more certainty than the design supports.

CONSTRAINTS
Do not invent statistical power, significance, baseline values, expected lift, or causal validation.

OUTPUT
Return: strongest part of the design, risks to interpretation, missing decision rules, alternative explanations, and questions to resolve before launch.

SOURCE UPDATE
[SOURCE UPDATE]

AUDIENCE
[AUDIENCE]

DECISION OR ASK
[DECISION / ASK]

TASK
Rewrite the update so the audience can quickly understand outcome, evidence, risk, unresolved uncertainty, decision/ask, and next step.

CONSTRAINTS
- Preserve every factual claim and number.
- Do not add confidence, progress, commitments, dates, causes, or outcomes not present in the source.
- Keep unresolved risks visibly unresolved.

OUTPUT
Return a concise update with: status, evidence/result, risk/uncertainty, decision/ask, and next step. Then list any source statement that is too ambiguous to rewrite safely.

LAUNCH PLAN
[LAUNCH PLAN]

KNOWN OWNERS
[OWNERS]

DEPENDENCIES
[DEPENDENCIES]

PAUSE / ROLLBACK RULES
[ROLLBACK / PAUSE RULES]

TASK
Review the plan for missing readiness areas, unresolved dependencies, unclear ownership, weak monitoring, communication gaps, and decisions that still need an explicit owner.

DECISION CRITERIA
A launch is inspectable when scope, audience, rollout, owners, dependencies, success/guardrail signals, support readiness, monitoring, and pause/rollback logic are explicit.

CONSTRAINTS
Do not assume readiness because a checklist item is mentioned. Do not invent completed work, owner agreement, test results, dates, or operational coverage.

OUTPUT
Return: confirmed inputs, unresolved items, owner questions, launch risks, and go/no-go decisions that require human confirmation.

CURRENT BULLET
[CURRENT BULLET]

WHAT I ACTUALLY DID
[WHAT I ACTUALLY DID]

VERIFIED OUTCOME OR LEARNING
[VERIFIED OUTCOME / LEARNING]

TASK
Critique the bullet for vague scope, unclear decision-making, inflated ownership, and unsupported impact. Then propose up to three truthful rewrites.

DECISION CRITERIA
A strong PM bullet should make context, contribution/decision, and verified outcome or learning clear without pretending collaboration was sole ownership.

CONSTRAINTS
Do not invent metrics, users, revenue, team size, seniority, ownership, experiments, or business outcomes.

OUTPUT
Return: evidence gaps, wording risks, and truthful rewrite options. If the supplied evidence is too weak for a quantified result, write a non-quantified version instead.

INTERVIEW QUESTION
[INTERVIEW QUESTION]

MY ANSWER
[MY ANSWER]

TARGET LEVEL
[TARGET LEVEL]

INTERVIEW MODE
[INTERVIEW MODE]

TASK
Act as a skeptical interviewer. Identify assumptions, weak trade-offs, vague evidence, missing success measures, or places where I avoided making a decision. Ask one follow-up at a time.

CONSTRAINTS
- Do not rewrite my experience or invent company context.
- Do not provide a polished answer before I attempt the follow-up.
- Challenge reasoning, not trivia.

OUTPUT
First return a short critique tied to the interview skill, then ask the single highest-value follow-up question.