Find buried signals in years of customer feedback.

What the AI will do

• Find and organize customer feedback, then examine it from several perspectives for quieter recurring concerns.
• Check candidate signals against original records and counterexamples, then develop your selections into cited reports and recommendations.

What you can provide

• The product or business to investigate, if it is not clear from context.
• Access to feedback or a permitted export when needed; the AI investigates available sources and helps you supply what is missing. A small sample can start a preliminary review.

Where you'll take part

• If the AI cannot start separate review sessions itself, open them with the exact briefs it supplies and return the findings; the AI does the analysis and combines the results.
• If independent agents or sessions cannot be arranged, decide whether to accept the less independent sequential-review fallback.
• Choose which signals deserve fuller reports, then correct the findings and recommendations.

What you'll get

A traceable shortlist of overlooked signals and, for those you select, fuller cited reports with specific recommendations and clear evidence limits.

Activity context

The loudest complaint is not always the most useful discovery. Have AI search years of customer feedback for quieter, recurring concerns that may point to a change worth making.

For the user

Paste the prompt below and identify the product or business you want to investigate. The AI will look for accessible feedback and help connect or import more if needed. A large collection gives it room to find patterns that a small sample could miss.

You’ll review the quieter signals and their examples, then choose which deserve a fuller report with supporting evidence and product-change recommendations.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Find quieter, potentially useful signals in a large collection of customer feedback. Do the collection, analysis, and evidence checking yourself. Ask one focused question at a time when you need my input.

1. Establish the product and find the feedback.

Identify the product or business from context; ask if unclear. Check available connections and try relevant internal customer systems, such as a CRM or support platform. Search public reviews and other customer feedback where appropriate. Inspect the sources before asking me to gather data. If access is missing, investigate connection or export options, then guide me through the specific user action needed. Explain the value of read-only access and offer an IT request when needed.

Inventory the available sources, dates, products, record counts where knowable, and fields. Read example records to understand how the data is organized. Establish a scope broad enough to look for recurring secondary concerns, ideally across a substantial period. Tell me what you can cover and what is absent. If only a small sample is available, investigate routes to more data and explain that any patterns found so far are preliminary.

2. Prepare a traceable collection.

• Create a working index with a stable ID for each record and its source link or internal reference, date, product, available customer context, and original text.
• Keep ratings and source channels where available.
• Identify duplicated reviews, syndicated copies, repeated messages in one case, and templated material so they do not masquerade as independent customers.
• Preserve the originals and document how you group or exclude duplicates.
• Distinguish separate issues in the same feedback item.

Divide large collections into manageable batches that retain coverage of different dates, sources, products, ratings, and customer groups where known. Track exactly which records each pass reviews. If you cannot process the whole collection, state the sampling method and coverage rather than implying an exhaustive review.

3. Read from several independent perspectives.

Use several subagents when available. Give each the product context, its assigned records, source-reference requirements, and a distinct analytical task. Include passes looking for: secondary or incidental concerns; repeated workarounds and unmet needs; changes over time or differences between products and customer groups; and details of use, packaging, convenience, or reliability that seem actionable. Also include an open-ended pass for patterns these perspectives miss.

Treat these as lenses, not a fixed list of findings you expect to discover.

Require each reviewer to capture candidate signals, exact supporting excerpts and record IDs, context, and counterexamples. Have them distinguish the main complaint from incidental mentions. Do not let early hypotheses dictate every pass. If subagents are unavailable, organize separate sessions for these perspectives. Prepare a self-contained brief for each session with the product context, assigned source records, analytical task, and required evidence format. Keep initial sessions independent of one another's hypotheses.

Maintain a coordinator record of scope, coverage, source index, returned findings, and remaining work. Start the sessions yourself where your tools permit; if I must start them, provide the exact briefs and explain how to return the findings for you to combine. Keep the analytical work with the AI. If fresh sessions cannot be arranged, offer clearly labeled sequential passes as a less independent fallback and wait for my agreement before using it.

4. Combine candidates without losing the quiet details.

Merge signals that express the same underlying concern while preserving meaningful differences. For example, widespread demand for more protein in a food product might be a loud, familiar theme. A steady stream of customers mentioning cheap packaging as a secondary or tertiary concern might be quieter and easier to act on. This is an illustration, not a finding to impose on the data. A secondary concern need not be literally the third sentence.

For each candidate, explain what customers are experiencing, how it differs from dominant themes, and what change it might suggest. Preserve examples from varied records. A signal can be quiet because it is rarely the main complaint, concentrated in a smaller group, described in varied language, or obscured by broader categories; it need not be the least frequent topic.

5. Check each candidate against the underlying records.

Search the full accessible collection for the idea using synonyms, paraphrases, and product-specific language. Read matching records in context and examine nonmatching samples to check what the search misses. Look for contradictory experiences and alternate explanations. Have a separate reviewer challenge candidates where available.

• Count independent supporting records using a stated definition, and report denominators only for collections actually examined by a comparable method.
• Compare time periods or segments only after checking whether source mix, product mix, or collection practices changed.
• Distinguish recurrence from proof of prevalence or causation.
• Rank useful signals by evidence strength, persistence, affected customers, likely significance, and plausible ease of improvement.
• Keep speculative opportunities labeled as such.

6. Present discoveries and let me choose the deeper work.

Give me a concise shortlist. For each signal, include the finding, why it was easy to overlook, representative quotations with source references, the extent and limits of supporting evidence, contradictory evidence where present, and a possible product or service response. Include a short coverage note so I know what was examined. Ask which signals deserve a fuller report and wait for my selection.

For each selected signal, investigate further as needed and produce a thorough report with traceable citations, examples, counts where supportable, alternative explanations, and specific product-change recommendations. Tie each recommendation to customer evidence, distinguish known feasibility or cost from estimates, and explain what should be tested before treating the change as a solution. Deliver the report and incorporate my corrections.

Done when

You have a traceable shortlist of quieter customer-feedback signals. For the ones you select, you have a fuller cited report and specific recommendations grounded in the feedback.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.