SupportCustomer Success Managerintermediate 15 min saved

Customer Support Root Cause Analyzer

Distinguish surface-level complaints from systemic root causes to resolve customer issues permanently.

Customise & copy

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

Stated Complaint *

The customer's initial complaint or ticket description.

Conversation Transcript *

Transcript or text log of the customer-agent conversation.

Business Context *

Select the industry or product type for better contextual framing.

Live prompt

3 fields

Analyze the following customer support interaction to uncover the true root, underlying problem driving the interaction, distinguishing it from the surface-level complaint.

Initial Stated Complaint: The customer is upset that their invoice is higher than expected.

Conversation Transcript:
Agent: Welcome to support. Customer: My bill is $50 higher this month, fix it now! Agent: Let me check. You added two user seats on the 12th. Customer: I didn't authorize that! Team lead told me adding seats was free during the trial. Agent: The trial ended last week. Customer: Nobody told me when it ended, now my accounting department is going to reject this expense.

Context/Environment: SaaS / Subscription Software

Provide your analysis using the following strict structure:

1. STATED COMPLAINT VS. UNDERLYING PROBLEM
- Stated Complaint: Summarize the surface-level issue presented by the customer.
- Underlying Problem: State the actual systemic, technical, or psychological blocker driving the interaction.

2. CONVERSATIONAL EVIDENCE
- Provide 2-3 direct quotes or specific exchange points from the transcript that justify why the underlying problem is the true root cause.

3. STRATEGIC RESOLUTION
- Recommend a specific, actionable strategy to resolve the underlying issue rather than just patching the surface symptom.

Proof before you use it

A real example

Tested on gemini-3.1-flash-lite on 2026-09-04

{
  "stated_complaint": "Customer claims they were overcharged on their monthly bill.",
  "context_environment": "SaaS / Subscription Software",
  "conversation_transcript": "Agent: Welcome to support. Customer: My bill is $50 higher this month, fix it now! Agent: Let me check. You added two user seats on the 12th. Customer: I didn't authorize that! Team lead told me adding seats was free during the trial. Agent: The trial ended last week. Customer: Nobody told me when it ended, now my accounting department is going to reject this expense."
}

A small ritual that works

How to use it

01Paste the customer's initial complaint into the 'Stated Complaint' field.
02Copy and paste the full conversation transcript into the 'Conversation Transcript' field.
03Select your business environment to ensure context-aware analysis.
04Review the generated root cause analysis and implement the suggested strategic resolution.

Why it works

The prompt enforces a rigorous analytical framework by demanding a strict separation between surface-level symptoms and systemic root causes. By requiring direct conversational evidence to justify the underlying problem, it prevents unsupported speculation and forces the AI to anchor its strategic resolutions directly to transcript details. The use of context/environment variables also frames the analysis to fit specific industry realities.

Where it fails

INSUFFICIENT_EVIDENCE

Customisation tips

To adapt this prompt for other domains, expand the context_environment selection options to include industries like logistics, hospitality, or education. You can also add a variable for company policy constraints to ensure that the strategic resolutions align with what support agents are actually authorized to do.

Watch-outs

Ensure transcripts provided in the variables contain enough dialogue depth to reveal an underlying problem; very short exchanges may force the model to hallucinate psychological or systemic blockers.
Watch for AI models overcomplicating simple transactional errors into massive organizational failures when the root cause is merely a localized miscommunication.