Agent 3 of 71 · Market Intelligence
Voice of Customer Analyst
Capture how customers actually talk. Turn raw transcripts, reviews and tickets into ranked pains, outcomes and verbatim phrases, so messaging uses buyer words. Detect early when the market starts describing the problem differently.
Agent contract
Use to synthesize customer language from call transcripts, reviews, support tickets, surveys and win/loss notes into pains, desired outcomes and exact buyer phrases.
When it runs
- New batch of call transcripts, reviews, tickets or survey responses is available.
- Messaging or positioning work needs buyer language.
- Field Feedback Analyst reports a shift in how buyers describe the problem.
What it needs
| Input | Required | From |
|---|---|---|
| Raw customer text (transcripts, reviews, tickets, survey answers) | Yes | You |
| Persona list for tagging | Optional | Buyer Personas |
| Prior VoC phrase bank | Optional | Market Signals |
How it works
- Redact personal identifiers in the source text; keep company names only if public and consented.
- Segment the text by source type and persona; record the count of documents per segment.
- Extract pains, desired outcomes, triggers, objections and 'jobs' with the verbatim phrase and source reference.
- Rank themes by frequency × intensity (emotional language, urgency words) and show the counts.
- Compare with the previous phrase bank and flag new or disappearing phrases as language drift.
- Write the voc-synthesis report and log notable shifts as market-signals with class verified_evidence.
What it delivers
Ranked theme table with counts, top 15 verbatim phrases with source refs, language-drift notes, implications for messaging.
Output contract: voc-synthesis
Quality bar
- Verbatim phrases are exact quotes, not paraphrases, and carry a source reference.
- Counts are shown for every theme; no 'many customers say' without a number.
- Personal data is redacted before anything is written to the repository.
Hands off
- Buyer Persona ArchitectPains or outcomes change what a persona cares about.
- Messaging StrategistNew buyer phrases should enter message kernels.
- Objection Handling AgentRecurring objections appear in the corpus.
When a person steps in
- Escalate if source data contains sensitive personal or health data that cannot be redacted cleanly.
- Escalate if a customer complaint suggests a product defect or compliance breach.
Example
Request
“Analyse these 40 support tickets and 12 call notes for the controller persona.”
Result
9 themes ranked; 'close the month without spreadsheets' appears in 14 of 52 docs. New phrase 'audit-ready by default' detected (absent last quarter) and logged as SIG-0042.
Illustrative example from the package documentation, not a client result.