Skip to main content

Setting up your AI Call Scoring templates

Building on your top performer or your own methodology, create templates that will take your team’s performance to the next level.

Updated this week

This article covers how to set up your AI Call Scoring templates that will allow you to verify at scale how well your Sales representatives perform against your methodology and to identify the most relevant coaching opportunities to boost your team’s performance.

Prerequisites:

  • You must be a Modjo Administrator/Manager

  • You must have enabled the Modjo AI on your organisation

1. Create a model

To create an AI Call Scoring template, go to Settings > AI Call Scoring.

Here, you’ll find your existing templates to edit, and the option to create new ones.

You can build a Modjo template using 3 methods:

  • From top performers: create a template based on how your best reps conduct their calls. Our AI analyzes their common patterns to generate a model, great to leverage existing expertise.

  • From scratch: the best choice if you know exactly what you want to assess.

  • From a use case: need guidance or inspiration? Start from one of our 4 use cases, including BANT methodology, gathering key context, following meeting structure, or executing a client business review.

A. Create a model from top performers

Use this if you’ve identified high performers in specific contexts:

  • A rep who quickly disqualifies bad leads

  • An AM who consistently drives upsell

Goal: Generate a model based on their calls to replicate what works.

Steps:

  • Click Create a model, then select Leverage top performers.

  • Choose the performers to analyze and apply filters (team, tags, deal stage, etc.). The AI will analyze up to 50 recent calls matching your filters.

  • Click Generate questions.

  • You can edit or add questions, then move on to testing.

💡 You can include as many questions as needed.

B. Create a model from scratch

Build a fully custom model:

Steps:

  • Click Create a model, then select Start from scratch.

  • Choose whether to target specific calls or not.

  • Click Create questions and write your first one.

  • ⚠️ Testing the question is mandatory, otherwise it won’t be validated.

  • Use Improve my question to let the AI optimize the clarity and impact.

  • Indicate whether the response is correct to help the AI refine results.

  • Once tested, click Validate question.

  • Then click Finalize template.

C. From a predefined use case

Jumpstart with ready-to-use templates.

Steps:

  • Select a predefined use case.

  • Add or adapt questions to match your needs.

  • ⚠️ Test each question (mandatory).

  • Validate them to finalize your model.

2. Test your questions (applies to all template types)

To validate and publish a model, each question must be tested on at least 5 calls.

How it works:
The AI reviews each call and scores the response:
 ✅ True: the question was clearly addressed
 ❌ False: the question was not addressed
 ⚠️ Partial: partially addressed, could have gone further
 ➖ Not applicable: this label is used when the topic was mentioned during the call, but not in the expected way.
For example, if the question is “Did the Sales Rep ask for the prospect's budget?” and the prospect brings it up spontaneously, the AI considers the question not applicable.
This value is neutral and not included in the overall scoring.

  • Audio timecodes are available so you can replay the exact excerpts justifying the score.

  • You can confirm or reject the AI’s analysis with a 👍 or 👎.

  • If you edit a question and rerun the test, the same calls will be used so you can compare the impact.

💡 For the “Leverage Top Performers” model:

  • The 5 calls used for testing are always the most recent ones from the top performers.

  • If you edit a question and rerun the test, the same calls as before the edit will be used so you can compare the impact.

3. Write / improve your questions

Use Improve my question to rewrite your questions for clarity and accuracy, improving AI analysis reliability.

4. Final configuration

Once all questions are tested and validated:

  • Name your model

  • Choose the AI output language

  • Click Publish

Your model will automatically apply to relevant calls and results will be visible in your analytics and call review pages.

Related articles:

Did this answer your question?