Indian Call Center QA Analytics using AI: Overcoming the cost barrier

India is home to one of the largest call center industries in the world, employing a vast workforce for both domestic and international operations. Domestic call centers alone employ hundreds of thousands of agents, with many more working in international processes.

?

QA Budgets in Call Centers

Call centers generally allocate around 5% of their total budget to quality assurance. Given the modest percentage, the bulk of this is spent on manual efforts, such as employing quality analysts and managers who monitor agent performance. Typically for every 100 seats, a domestic call center typically spends less than INR 2,00,000 per month on QA.

?

The Challenge of Manual QA

Each agent typically handles around 100 hours of calls per month. This varies depending on capacity utilization rates in the call center. Manually reviewing such volumes is a massive undertaking, and as a result, traditional QA processes only sample 1% of the total calls. This limited sampling leaves large gaps in ensuring consistent customer service quality. There are many undetected instances of poor service, indifferent handling and rude behaviour.

?

Economics of AI-Driven QA

AI-driven analytics presents an opportunity to revolutionize the way call center QA is conducted. However, the primary barrier to the widespread adoption of AI in this sector is economics. To make AI-driven analytics viable, the cost of analyzing calls needs to be significantly lower than manual alternatives.

?

For instance, considering the QA budgets, an AI-based analytics solution would need to analyze 1,000 hours of calls at a cost of less than INR 40,000 (10% sampling of 100 agents). This would allow the call center to reduce one QA analyst per 100 call center agents. This translates to around INR 0.66 per minute (~0.8 US cents/minute)—a figure AI solutions have struggled to achieve until recently.

?

Breaking the Cost Barrier

The arrival of accurate open-source algorithms (ASR, translation, LLMs) for Indic languages, such as those launched by Sarvam.ai, has lowered the cost of AI-driven call analytics. These algorithms allow for more efficient processing and analysis of regional language calls, which dominate Indian call center operations. This advancement could finally make AI-powered QA affordable for Indian call centers, enabling them to review a higher percentage of their calls rather than relying on small samples.

Phased Approach for AI Implementation

One approach to using AI for analyzing call center service quality is to migrate QA metrics selectively (one by one) to AI models. Simpler metrics can be migrated first:

• Mention of company branding in the first 15 seconds
• Proper disposal of the call using appropriate short codes (to ensure no repeat calls)
• Reasonable number of words per minute used by agent
• Avoidance of rudeness/ abusive behaviour -- detected using an emotion detection AI model

This can be followed by more complex QA metrics that require checking the objection handling/ rebuttals of the agent. Over time all the QA metrics that the call center uses can be migrated to AI.

The promising future

With these cost-efficient AI solutions, the industry stands on the brink of a major transformation. Domestic call centers can achieve comprehensive quality assurance, leading to better service, improved agent performance, and ultimately, enhanced customer satisfaction.