Feedback analysis prompt examples

Use these examples in Agent Sessions where available, or adapt them to the Wisdom experience in your workspace. They are requests for analysis, not guarantees about the data available or the correctness of a generated answer.

Replace the placeholders before sending. Include the product or issue, reporting dates, customer segment, relevant sources, and desired output. When asking for a number, specify whether you mean feedback records, users, accounts, or another available measure.

Get an overview of feedback

Summarize feedback about [product area] from [start date] through [end date] for [segment and sources]. Group the main positive themes and pain points. State the searched scope, include supporting record links, and distinguish recurring evidence from isolated examples. Explain any coverage limits.

For a numerical breakdown, add: “Count distinct [feedback records/accounts/users] for each theme, explain the filters and denominator, and note whether themes overlap.”

Investigate possible causes and customer impact

Review complaints about [issue] within [scope and dates]. Describe what customers were trying to do, the obstacles they reported, and the consequences they described. List possible explanations with supporting evidence and unresolved questions. Separate customer statements, observed patterns, and hypotheses; do not treat association as proof of cause.

Use the result to guide investigation. Feedback can suggest an explanation without establishing a technical root cause or predicting the effect of a product change.

Create a report, FAQ, or shareable draft

Draft a voice-of-customer summary for [product area] over [dates and scope]. Include the main themes, supporting quotes and record links, positive feedback, and open questions. Mark unavailable evidence explicitly and explain how any numerical claims were calculated.

Draft an internal FAQ about [feature] from the questions and confusions in [feedback scope]. Separate answers supported by the feedback from product facts that need verification. Include source references and mark unanswered questions.

Draft a short Slack update about [product area] for [audience], based on [scope and dates]. Include the main complaints, requests, positives, and evidence links. Return the draft for review; do not post it.

For the complete metrics-and-narrative workflow, see How do I create a voice-of-customer report? Output and export options depend on the experience and permissions available to you.

Support roadmap planning

Summarize requests and pain points for [product area] within [scope and dates]. Group candidate opportunities, show the supporting evidence and selected count or measure, and distinguish customer requests from your recommendations. Note sampling limits, overlapping themes, and any assumptions about customer impact. Do not invent revenue or outcome estimates.

Treat this as evidence for prioritization alongside product strategy, feasibility, and other business context.

Understand dissatisfaction and compare populations

Compare complaints about [workflow] between [segment A] and [segment B] over the same [dates and sources]. Use the same measured object and denominator definition on both sides. Show the supporting examples, coverage and sample sizes, and explain where the comparison is inconclusive.

For a period comparison, replace the segments with two explicit date ranges and keep the remaining conditions consistent.

Saved preferences and recurring work

If your experience supports saved instructions or preferences, use its available settings or guided tool flow and check what was saved, its scope, and when it applies. Merely writing “create a rule” does not confirm that a persistent instruction exists.

A workflow-specific preference can be handled differently from a general Rule. Keep the required product scope, segment, and dates explicit in each report request. Saved instructions are not enforced query filters or data-access policies.

For recurring analysis, use the available automation setup and confirm its trigger or schedule, scope, destination, authorized account, and status. A prompt asking for an alert is not by itself evidence of an active subscription or automation.

Choose a model with the available control

Use the model selector offered by your experience when you want to choose a model. Availability and defaults can change with your workspace and configuration; use the options and messages shown.

Typing a model name into an analysis prompt is not a reliable substitute for changing that selector. If selection is unavailable during active work or a pending action, finish or resolve that work and follow the control's instructions.

Verify citations, quotes, and numbers

Ask for supporting record links, then open them. Check exact quotations against source text, distinguish excerpts from paraphrases, and use speaker or timing attribution only where it is available. A citation does not guarantee that every generated statement is correct.

For numerical claims, inspect the measured object, filters, dates, denominator, and coverage. See Calculations on Quantify. No results means no results in the searched scope; it does not prove that an issue never occurred.

Refine the analysis in the same conversation

Use follow-ups to narrow or extend the analysis while restating constraints that must stay fixed. For example:

• “Keep the same dates and sources. Break these themes down by [segment].”

• “Show the records supporting this claim, and separate exact quotes from paraphrases.”

• “What evidence is missing, and what should we verify before sharing this conclusion?”

Review the final scope and evidence after each material change. For Agent setup and Session controls, see Getting Started with Enterpret Agent.

Daily reports, weekly digests, and recurring summaries

Start with Enterpret Agent to refine a reporting prompt, then use scheduled Agent automations for recurring delivery. Example: “Summarize feedback about [product area] from the last 24 hours. Compare with the previous 24 hours, identify significant changes, and include counts with definitions and quotes with source links.” Set the daily schedule, timezone, and Slack or email destination in the automation settings. Add connectors when the work needs external context or a supported action.