Atul Singh logoAtul Singh
← Back to prompts
MarketingProduct Marketerintermediate 120 min saved

Customer Voice Synthesis & Messaging Architect

Turn raw customer feedback into high-signal messaging by clustering authentic verbatim language.

Customise & copy

Add the context you have. The prompt updates as you work.

What output or research artifact are you informing with this language?

Who is the specific segment whose language you are analyzing?

Paste raw customer interview notes or transcripts.

Paste customer support logs or public review data.

Paste open-ended survey exports.

Live prompt

5 fields
Analyze the following qualitative customer data sources to extract, cluster, and analyze recurring verbatim customer language. Your goal is to identify high-signal phrases to inform landing page headline and hero copy for our target audience of B2B SaaS product managers.

Here are the customer data sources to analyze:

1. Customer Interviews / Transcripts:
Customer A: 'We spend hours manually stitching CSVs together every Monday.' Customer B: 'I just need a dashboard that doesn't break when our schema updates.'

2. Support Tickets / Product Reviews:
Review 1: 'Exporting large data sets always times out without warning.' Ticket 2: 'Where is the documentation for setting up SSO?'

3. Survey Responses:
Survey 1: 'We switched from legacy software because it was too rigid for our custom workflows.'

Strictly adhere to the following execution rules:
- Do not fabricate, smooth out, or alter customer quotes. Use authentic verbatim text only.
- Strictly separate direct customer verbatims from your AI synthesis and interpretation.
- Group expressions into distinct clusters: Problems/Pain Points, Desired Outcomes, Objections, and Alternatives.
- Highlight high-signal phrases that are supported by multiple distinct input sources.
- Provide actionable recommendations for applying the extracted language directly to landing page headline and hero copy.

Proof before you use it

A real example

Tested on gemini-3.1-flash-lite on 2026-09-06

{
  "target_audience": "finance automation directors",
  "survey_responses": "Response 44: 'We needed a tool that talks directly to our ERP without custom middleware.'",
  "messaging_objective": "b2b product positioning and website copy",
  "support_and_reviews": "Ticket 128: 'Reconciliation rules keep deleting themselves after an update.' Review: 'Saved us 10 hours on closing, but setup took weeks.'",
  "interview_transcripts": "CFO 1: 'Month-end close is a black box until day five.' CFO 2: 'Our auditors hate spreadsheets with manual formulas.'"
}

A small ritual that works

How to use it

  1. 01Gather raw data from your CRM, support tickets, and interview transcripts.
  2. 02Paste the data into the corresponding input fields in the prompt.
  3. 03Define your specific messaging objective and target audience segment.
  4. 04Review the AI-clustered verbatim output to identify high-signal phrases for your copy.

Why it works

The prompt forces a rigorous separation between authentic customer verbatims and AI synthesis, while grounding messaging objectives directly in raw qualitative data like interviews, support tickets, and surveys. This prevents generic copywriting by anchoring every copy recommendation strictly in documented user friction and terminology.

Where it fails

INSUFFICIENT_EVIDENCE

Customisation tips

Ensure the input data variables (interview transcripts, support tickets, and surveys) contain a sufficient volume of overlapping verbatims if you intend to isolate high-signal phrases supported by multiple distinct input sources.

Watch-outs

  • Do not allow the model to invent or smooth out customer quotes; enforce strict verbatim usage to maintain signal integrity.
  • Watch out for unverified feature claims sneaking into the actionable recommendations that are not explicitly backed by the provided customer data.
  • Ensure implicit alternatives or missing clusters are handled transparently rather than hallucinating text to fill empty data categories.

Works in

ChatGPTClaudeGemini

Was this useful?

Submit your variation

If your version helps, it may be published with your name and a link back to you.

Want to make AI useful across your team? Explore Atul's training.