Call Analysis

Overview

The Call Analysis feature in VoiceCreator.ai allows you to summarise and evaluate calls, providing valuable insights into their effectiveness. This feature leverages the latest models, primarily Anthropic’s Claude Sonnet, and uses OpenAI’s GPT-4o as a fallback.

Customisable options include prompts, schemas, and rubrics to tailor the analysis to meet your specific needs. The analysis results are stored in the call logs under structured properties like call.analysis.summary, call.analysis.structuredData, and call.analysis.successEvaluation.

Customisable Features

You can customise the analysis behaviour in the assistant’s assistant.analysisPlan configuration.

1. Summary Prompt

The summary prompt generates a concise summary of the call, stored in call.analysis.summary.

Default Summary Prompt

You are an expert note-taker. You will be given a transcript of a call. Summarise the call in 2-3 sentences, if applicable.

Customising the Summary Prompt

To customise the summary prompt, update the summaryPrompt property:

{
  "summaryPrompt": "Custom summary prompt text"
}

To disable the summary prompt, set it to an empty string "" or "off":

{
  "summaryPrompt": ""
}

2. Structured Data Prompt

The structured data prompt extracts specific pieces of data from the call and stores them in call.analysis.structuredData.

Default Structured Data Prompt

You are an expert data extractor. You will be given a transcript of a call. Extract structured data per the JSON Schema.

Customising the Structured Data Prompt

To set a custom structured data prompt, update the structuredDataPrompt property:

{
  "structuredDataPrompt": "Custom structured data prompt text"
}

3. Structured Data Schema

The structured data schema enforces the format of the extracted data using JSON Schema standards.

Customising the Structured Data Schema

To define a custom schema, set the structuredDataSchema property:

{
  "structuredDataSchema": {
    "type": "object",
    "properties": {
      "field1": { "type": "string" },
      "field2": { "type": "number" }
    },
    "required": ["field1", "field2"]
  }
}

4. Success Evaluation Prompt

The success evaluation prompt determines whether the call met its objectives. The results are stored in call.analysis.successEvaluation.

Default Success Evaluation Prompt

You are an expert call evaluator. You will be given a transcript of a call and the system prompt of the AI participant. Determine if the call was successful based on the objectives inferred from the system prompt.

Customising the Success Evaluation Prompt

To set a custom success evaluation prompt, update the successEvaluationPrompt property:

{
  "successEvaluationPrompt": "Custom success evaluation prompt text"
}

To disable the success evaluation prompt, set it to an empty string "" or "off":

{
  "successEvaluationPrompt": ""
}

5. Success Evaluation Rubric

The success evaluation rubric defines how the call’s success is measured. The following rubrics are available:

• NumericScale: A scale of 1 to 10.
• DescriptiveScale: A scale of Excellent, Good, Fair, Poor.
• Checklist: A list of criteria with statuses.
• Matrix: A grid evaluating multiple criteria across performance levels.
• PercentageScale: A scale of 0% to 100%.
• LikertScale: A scale of Strongly Agree, Agree, Neutral, Disagree, Strongly Disagree.
• AutomaticRubric: Breaks down evaluation into multiple criteria, each with its own score.
• PassFail: A simple pass or fail evaluation.

Customising the Success Evaluation Rubric

Set a custom rubric using the successEvaluationRubric property:

{
  "successEvaluationRubric": "NumericScale"
}

6. Combining Prompts and Rubrics

Prompts and rubrics can be combined for detailed call analysis:

{
  "successEvaluationPrompt": "Evaluate the call based on these criteria:...",
  "successEvaluationRubric": "Checklist"
}

How to Use Call Analysis

1. Configure Prompts and Rubrics: Tailor the summaryPrompt, structuredDataPrompt, and successEvaluationPrompt properties as per your requirements.
2. Test and Iterate: Run calls and review the generated summaries, structured data, and evaluations. Adjust configurations to optimise results.
3. Extract Insights: Use the generated outputs to assess call quality, improve workflows, and better meet user needs.

Example Configuration

Here’s an example of a complete call analysis configuration:

{
  "assistant": {
    "analysisPlan": {
      "summaryPrompt": "Summarise the call in 2-3 sentences.",
      "structuredDataPrompt": "Extract user details like name, email, and phone number.",
      "structuredDataSchema": {
        "type": "object",
        "properties": {
          "name": { "type": "string" },
          "email": { "type": "string" },
          "phone": { "type": "string" }
        },
        "required": ["name", "email"]
      },
      "successEvaluationPrompt": "Evaluate if the user received adequate support during the call.",
      "successEvaluationRubric": "DescriptiveScale"
    }
  }
}

Conclusion

VoiceCreator.ai’s Call Analysis feature provides powerful tools to evaluate and summarise calls, offering actionable insights to improve performance. By customising prompts, schemas, and rubrics, you can tailor the analysis to suit your business needs.