IBM uses GenAI to help Euro telco call-center team work better

Every call a customer makes to a call center is a learning opportunity for the enterprise. But when call center agents are too rushed to be able to read previously captured automatic transactions, valuable insight is going to waste. IBM found a fix for a large European telco using generative AI (GenAI).

The telco’s call center agents did not have time to read the transcripts recorded from previous customer contacts. Conversations with customers had not been fully captured in the company’s CRM, leaving gaps in understanding. The incomplete and lengthy nature of these transcripts meant they were hardly used.

Customer ops leaders are on a red warning to adapt quickly with GenAI

HFS identifies customer operations—including customer service functions—as one of the first areas generative AI (GenAI) will impact (see Exhibit 1). IBM’s solution involved using two GenAI foundation models, also known as large language models (LLMs).

Exhibit 1: Our red-amber-green scale shows with what urgency you should apply GenAI to redesign roles; customer operations (including customer services) are already in the red zone

Data sources: *Corporate Industry Service, Oxford Economics, McKinsey
Source: HFS Research, 2023

First LLM summarizes conversations into simple bullet points

In the first LLM instance, IBM applied GenAI’s natural language understanding capabilities to generate automatic call summarization, not simply transcribing the conversation as it happens but using the transcription to generate an easier-to-consume summary in three or four bullet points. The LLM was also trained to identify key topics and extract these.

It then updates the Salesforce CRM, so any agent handling the same customer in the future has the summary and critical topics in a much more concise format. Providing the summary cuts time to insight and makes that insight available to the person who needs it when they need it. The solution also captures actions taken in response and looks at how well aligned they were with what was said—supporting consistency between how an agent responds and the insights they are provided with.

A second LLM helps telco flag attrition risk from call content

The second LLM flags attrition risk by analyzing what is said in customer conversations. It analyses mentions of the telco’s competitors, sentiment, and emotion analysis. The analysis can identify when to respond with alternative offers or provide data to understand the trigger points for customer churn.

The outcome from analyzing six million conversations each month is around 30% savings in pre- and post-call operations, saving agents time and making them better able to provide an improved experience to customers. The client has also identified more than five million in annual operational improvements.

The Bottom Line: Focus on augmenting your people’s capabilities.

There is no mention of headcount reduction in this case study and no blunt measures of increased productivity. Instead, it focuses on improving how call center staff can do their job—delivering the insight and support they need in the form they need it when they need it. Enterprise leaders would be wise to apply a similar focus to their forays into GenAI.