Using Auto Call Scoring To Choose The Right Calls For Manual Review

If you have ever done manual QA, you know that finding a call to review can take up to 30 minutes (and sometimes longer). Tools like call randomizers can help somewhat, but you still don't know if you are reviewing the right calls. As you want to automate your QA processes with automated call scoring, you might wonder: "Can I choose Auto QA-scored calls to be reviewed by human evaluators? Isn't this going to just recreate my problem?"

I hear you. I was a contact center supervisor until a few years ago, and I definitely feel your pain. Since joining MiaRec, I have talked to hundreds of contact center managers, asking themselves the same questions. 

In this article, I’ll explain why it's essential to have Auto QA evaluate most, if not all, of your calls. I’ll also show you how to leverage QA scores and advanced Voice Analytics tools, such as Sentiment and Topic Analysis, to automatically and effectively identify calls that require further review. Let's dive in!

Using Auto Call Scoring To Pre-Score & Identify Calls For Human Review Or Follow-Up

Automatically scoring 100% of your relevant call recordings using Generative AI-powered Auto QA gives you a powerful way to pre-evaluate your calls. 

First and foremost, your Auto QA solution will provide you with a total score for each call. Depending on your scorecards and the total score an agent can achieve, you can set a threshold that determines when a call is good enough versus when it needs to be reviewed to find out why the score is so low and what actions need to be taken to improve the scores in the future. 

This is a great way to:

• Be aware of anything out of the ordinary happening in your contact center (for example, one of your agents is consistently rude to customers)
• Measure if training and coaching is effective (tracking of agents and teams before and after training to see the improvements)
• and much more.

Image: This screenshot shows an agent failing to meet the passing threshold with a 37% Auto QA score. It's a clear case for human review to identify areas for improvement.

Secondly, you can use Auto QA to check for routine requirements, such as script adherence, reading of compliance statements, and so on. Now, you have pre-scored all of your calls and have deep insights into the preliminary performance. This allows your human supervisors to dive deeper into a specific section to perform a more in-depth check. For example, you notice through the pre-scoring that agents become insecure in the middle section of the call. You review some of these calls to find the root cause and design training and knowledge base articles to help your agents. After the training, you can see how this section improves.

Combining Auto QA With Sentiment and Topic Analysis To Empower Your Human Evaluators

Auto QA has scored all or at least most of your calls, you can use powerful Voice Analytics tools to automatically identify calls that require human review or follow-up. This allows you to filter through all of your calls based on specific criteria to pick out the ones that truly matter. 

Imagine having a dashboard that neatly organizes all call scores, highlighting those that fall below a certain threshold (e.g., 50), show negative customer sentiment indicating dissatisfaction, or involve specific topics like "escalation requests" or "account cancellations." Let's take a closer look at these insights.

Using Topic Analysis To Identify Critical Calls

Topic Analysis is a powerful tool for efficiently identifying calls that require attention. By using Generative AI to automatically categorize calls based on their content, it allows quality assurance teams to quickly focus on interactions that involve critical issues or subjects of particular interest.

For example, if topics like "escalation request," "upset customer," or "cancel subscription" are detected, these calls can be flagged for immediate review. This approach is far more efficient than manually searching through calls or relying on random selection.

Image: Screenshot from the MiaRec Topics Dashboard showing 26 calls tagged "Problem with service," 18 with "Frustration," and 7 with "Broken Trust"—all ideal for QA analyst review.

Using Sentiment Analysis To Prioritize Calls With Negative Interactions

Sentiment Analysis provides another layer of insight by gauging the emotional tone of interactions. By identifying calls with negative sentiment scores, QA teams can prioritize reviewing potentially problematic interactions. This proactive approach allows for timely intervention, whether it's addressing customer concerns or providing additional support to agents.

Image: Screenshot from MiaRec application showing a call with negative customer sentiment.

Why Combining Topic Analysis & Sentiment Analysis With Auto QA Is Your New "Superpower" For Efficient QA

Combining Topic and Sentiment Analysis with automated call scoring gives you what I like to call "QA Efficiency Superpowers" because it dramatically enhances the efficiency and effectiveness of the QA process. Together, these tools allow QA teams to:

1. Quickly identify calls that need immediate attention without listening to hours of recordings.
2. Detect patterns and trends in customer issues and agent performance more easily.
3. Focus resources on the most critical interactions, improving overall service quality.
4. Improve customer satisfaction by addressing issues proactively.
5. Provide more targeted and timely feedback to agents.

These are just a few examples of how you can use those three tools to not only achieve significant time and cost savings, but also to actively start contributing to revenue generation by being able to eliminate busywork and focus on what drives business growth.

Conclusion: Balancing Comprehensive QA with Efficiency

In summary, automatically scoring your calls using AI-powered Auto QA significantly reduces the time spent on evaluating agents. This includes finding the right calls to review as Auto QA allows you to identify the right calls to review based on a set of defined criteria rather than random selection. It also allows you to overlay your call data with Topic and Sentiment Analysis which gives you superpower-like capabilities to drill down and hone into which calls require human review, follow up, or even escalation.