Reconfirmation

The “Reconfirmation” tracker detects instances where a representative restates or verifies a customer’s details during the conversation. Understanding these moments helps teams measure attention to detail, confirm adherence to verification protocols, and ensure that customers feel accurately understood.

Who Benefits:

• Project Managers & Team Leads: Verify that agents are consistently confirming customers’ details, improving data accuracy and reducing follow-up issues.
• Sales & CX Teams: Identify patterns in how and when representatives reconfirm information, ensuring customers feel acknowledged and validated.
• Business Managers & Strategists: Spot trends in reconfirmation requests—if customers often need details repeated, process improvements may be required.
• Compliance Officers: Ensure that representatives follow necessary verification steps to comply with security or regulatory standards.

Value Proposition:

• Accuracy & Trust: Detecting reconfirmation moments highlights agents’ diligence in ensuring the correctness of information, fostering trust.
• Efficiency: By monitoring these occurrences, organizations can refine scripts or training to streamline verification processes.
• Continuous Improvement: Insights from reconfirmation data feed into coaching and procedural updates, reducing errors and misunderstandings.

Data Dictionary

Objective:

Detect phrases where a representative reconfirms a customer’s information. Rather than a numeric score, the tracker returns occurrences of matched phrases along with timestamps and confidence. This data helps understand how thoroughly agents verify details.

Proposed JSON Schema:

{
  "trackers": [
    {
      "name": "Reconfirmation",
      "matches": [
        {
          "text": "<string>",
          "timestamp": "<ISO 8601 or relative time>",
          "confidence": <number> // 0.0 to 1.0
        }
      ]
    }
  ]
}

Data Dictionary:

If No Matches Found:
If no reconfirmation phrases are detected, return an empty matches array for the "Reconfirmation" tracker or omit the tracker. If no trackers match at all, return:

{
  "trackers": []
}

Determining Confidence & Thresholds

How the Reconfirmation Tracker Works:

• Seed the tracker with phrases that represent verification or reconfirmation steps (e.g., "just to confirm your details," "to verify, your address is," "let me restate your order number").
• The system interprets these phrases and identifies semantically similar statements.
• Confidence indicates how closely a detected phrase suggests reconfirmation of information.

Confidence & Calibration:

• Initial Testing: Validate the tracker on known calls to ensure detected phrases truly represent reconfirmation.
• Refine Phrases: Add or remove seed phrases for better coverage or reduced false positives.
• Thresholding: Downstream users can filter low-confidence matches if necessary.
• Stakeholder Feedback: Involve compliance or training leads to refine what constitutes a meaningful reconfirmation.

Prompt Construction & Instructions

Role Specification & Reiteration:

• Present the system as a “highly experienced assistant” detecting reconfirmation phrases.
• Reiterate instructions: return only JSON, follow the schema, no extra commentary.

No Hallucination:

• Only detect reconfirmation if supported by transcript evidence.
• If uncertain, assign lower confidence or omit matches.

Strict Formatting:

• Return only JSON with trackers.
• If none found, return empty trackers.

Example Prompt for Implementation

System Message (Role: System):
"You are a highly experienced assistant that identifies 'Reconfirmation' mentions in a conversation. You have:

• A transcript
• A 'Reconfirmation' tracker defined by seed phrases indicating verifying or reconfirming customer details.

Your task:

1. Detect phrases where the representative reconfirms information.
2. Return text, timestamp, and confidence for each match.
3. Return only JSON as defined. No extra commentary.

Instructions & Variables (example):

{
  "tracker": {
    "name": "Reconfirmation",
    "phrases": [
      "just to confirm",
      "to verify your details",
      "let me restate your order",
      "correct me if I'm wrong, your account is"
    ]
  }
}

User Message (Role: User):
"Analyze the following transcript and detect 'Reconfirmation' mentions:

[TRANSCRIPT_JSON][TRANSCRIPT_JSON]"

Expected Output Format:

{
  "trackers": [
    {
      "name": "Reconfirmation",
      "matches": [
        {
          "text": "Just to confirm, your address is 123 Main Street?",
          "timestamp": "2024-07-10T09:55:00Z",
          "confidence": 0.9
        }
      ]
    }
  ]
}