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.
Table of content
Prerequisites
You must be a Modjo Administrator/Manager
You must have enabled the Modjo AI on your organisation
Creating a template
To create an AI call scoring template, go to Settings > AI call scoring. In this page you will see your existing AI call scoring templates that you can edit and you will be able to create new templates. An AI call scoring template in Modjo can be built from 3 methods:
Build from top performers: create a template based on how your top performers do their calls. Our AI will analyze the most common patterns in your top performers’ calls and create a template out of it. Great to save time and capitalize on your existing expertise!
Start from scratch: the best way to start if you already know exactly what you want to assess
Start from a use case: need inspiration and guidance about the type of things he can assess with AI call scoring? Start from one of our 4 use cases including BANT methodology, gathering key context information, following meeting structure standards or executing well a business review with a client.
💡 Your templates can have as many questions as you want!
Build from top performers
Use cases you can address with the top performer flow
When creating a template from the top performer flow, the AI will analyze the most common patterns in your top performers’ calls and create a template out of it. This creation method is more especially relevant when you know one or several of your Sales representatives to be particularly good in specific contexts. Here are a few examples:
One of your Sales representative is particularly good at disqualifying opportunities that have less chances to close at the beginning of the Sales cycle. Building on his calls to create your templates will help you highlight the top questions he asks in qualification calls and replicate this method across your team.
One of your Account Managers generates more upsell opportunities than the rest of the team: build on his / her business review calls to identify questions that are always asked by this account manager and check if the rest of your team also asks them.
Building a template from the top performer flow
From the AI call scoring settings page, click on create a template and select the “build from top performers” flow.
Select the top performer(s) based on who you want the template to be built. You can also filter the calls that the AI will analyze using teams, tags, deal stage, etc. For example, if you want to build on the performance of an account executive at a specific stage of the sales cycle, it would be better to filter the calls that will be analyzed based on the deal stage.
💡 The AI will be able to analyze up to 50 calls from the top performers. It will select the most recent calls of the top performers that match filtering criteria (teams, tags, etc.).
When you are all set, click on “generate questions”. The AI will analyze the calls to create a template based on the identified patterns. It can take up to 1 or 2 minutes.
Your template built from your top performs is created! From there you can still add questions, edit the questions suggested by the AI and most importantly test your questions
Testing your questions (all creation methods)
No matter how you create your AI call scoring template (from top performer, from scratch or from use case), you will always have to test the questions of your template before publishing it.
The testing experience goes as follow:
For a given call, the AI shows the answer it would provide to the question you asked. The answer can be true, partial, not applicable or false
True: if the answer to the template question is true (example: the question is “Did the Sales Rep ask for the budget?”. If the Sales Rep did ask, then the AI will say true)
False: if the answer to the template question is false
Partial: if the AI assesses that the Sales actually did what was asked in the question but could have gone further. It usually happens for more complex questions.
Not applicable: if the topic was mentioned during the call but not necessarily as framed in the AI call scoring question. For example, if the question is “has the Sales Rep asked the propsect budget” and that the Sales hasn’t done it but this is because the prospect proactively started talking budget then the model will send a not applicable value. The not applicable value is neutral in the scoring (it won’t be taken into account in the computation of the AI call scoring)
The timecodes referring to where the AI found the answer are displayed so that you can more easily check if the AI answer is right by playing the call snippet justifying the AI answer.
Based on this you assess whether you agree or disagree with the AI by using 👍 or 👎
For a question test to be validated, you need to assess the AI answer on at least 5 calls.
Once done with the 5 tests you will see how accurate the AI was with a grade on 5 based on your assessments. From there you can either:
Run a new test (to test on a bigger sample)
Edit the question (if not satisfied with AI accuracy for instance)
Validate the question.
💡 How are the calls selected
In the case of top performer flow:** the AI will display the 5 most recent calls of the top performers for the test.
In the case of from scratch or use case flows: the AI will display the 5 most recent calls matching the filtering criteria selected during setup step.
💡 What happens if a user edits a question and goes back to testing You will be exposed to the same calls than before editing the question and will hence be able to assess if the AI assessment works better on the same sample of calls.
Writing or editing your question using AI
When writing questions, you can also benefit from the “Improve with AI” feature to help rephrase the question and make it more accurate and better understandable by the AI: this will improve the accuracy and relevance of the AI in the analysis of your calls!
For more tips about how to write good questions, click here.
Final setup
Once your questions have been tested, all you have to do is give your template a name and select the output language you want for the AI call scoring. You can then finally publish it!
Once published, your template will start apply on the calls matching your filtering criteria and you will find the scores whether from the analytics (for global view) or from the call details page (for a view at call level).