Analysing and Enhancing Customer Service Calls with a Large Language Model: Findings Detailed

Hostcomm Interaction Analytics effectively converts phone and video interactions into written text, which is subsequently processed by expansive language models. This process yields valuable data that can be harnessed for enhancing agent training and boosting customer satisfaction levels. The following is a sample output generated by Interaction Analytics after processing a customer service call. Manually analyzing the call transcript would be a labor-intensive task. To maintain confidentiality, any information that could reveal the identities of the individuals or the company involved has been altered.

Post call summary report:

"Sales Agent Call Summary

The conversation is between John, a customer service agent (spk_0), and a customer (spk_1). John is calling to assist with an incomplete order that the customer started on AB's website. They discuss the customer's current situation, their needs, and John suggests a specific plan. After going through the details of the plan, the customer decides to proceed with the purchase. John also clarifies the installation process, confirms the customer's details, initiates a credit check, sets up security questions, explains terms and conditions, and arranges for paperless billing.

Upselling and Marketing Opt-In:


John recommended to the customer the 'Plan A' plan based on the customer's internet usage requirements. The plan offers a higher speed than the customer currently has with their current service.

Marketing Opt-In:

John asked the customer for permission to allow AB to contact him with offers, news, and rewards. The customer agreed to receive such notifications via email.

Sentiment Analysis:

John (spk_0): Positive sentiment. John was polite, informative, and helpful throughout the conversation. He explained the options available to the customer and managed to successfully assist in completing the order.

Customer (spk_1): Neutral to Positive sentiment. The customer was cooperative, responsive, and receptive to the agent's recommendations. They didn't express any strong positive or negative feelings but was generally agreeable and satisfied with the service.

Score out of 10:

Agent (John): 8.5/10

John was professional and patient, providing detailed explanations about the service and clarifying all terms and conditions. He demonstrated a good understanding of the customer's needs and guided the customer to an appropriate solution. The deduction in score is due to the lack of an explicit attempt to upsell more additional services or products beyond the initial plan.

Reason for the Score:

John was successful in assisting the customer and making the sale. He effectively managed to upsell the 'Plan A' plan. His communication was clear, and he ensured the customer understood all terms and conditions. However, the conversation might have benefited from exploring more additional services or products that could enhance the customer's experience with BT, hence the deduction.

How could John have scored higher?

John's performance was commendable overall, with effective communication, detailed product knowledge, and patient handling of the customer's queries. However, there were a few areas where improvements could be made:

Empathy and Personalisation: While John was polite and professional throughout the conversation, he could have made more of an effort to create a personalised experience for the customer. Understanding the customer's personal context and unique needs can significantly improve the interaction.

Upselling and Cross-Selling: John focused mostly on the broadband product. However, he could have explored other services or add-ons that AB offers and that could benefit the customer. This could include premium support packages, TV subscriptions, or mobile services.

Proactive Communication: John mentioned important information like credit checks and contract termination conditions as part of the signup process. However, these significant points could be highlighted earlier in the conversation to ensure transparency.

Efficiency: The conversation was quite long and could have been more succinct. John could improve by summarising key points more concisely and avoiding repetitiveness.

Closing the Sale: John could improve his closing skills. He could affirm the customer's decision by summarising the benefits of the chosen package and reassuring the customer that they've made a good choice, creating a more positive ending to the conversation.

Follow-up Commitment: John could provide the customer with a clear next step or follow-up action to ensure the customer is aware of the process following the conversation. He could also offer to personally follow up with the customer, which could create a more personalised customer service experience."


The provided summary accurately encapsulates the customer phone call, surpassing the capabilities of a conventional voice analytics service. Instead of merely generating rudimentary data like keyword identification, sentiment analysis, and transcription, the advanced language model enhances its value by comprehending the conversation and critiquing the service agent's behaviour. This feature proves to be highly beneficial for training a vast team of service agents as the system can deliver this information promptly, within minutes of saving the call recording.

If you would like to experience this technology please contact us at [email protected] and we can provide a demonstration and a free trial.

Trusted by

The Car Buying Group Logo 2 Case Study Beer 2 Case Study EE logo white2 Case Study David lloyd white Case Study Kantar white Case Study Hmrc white Case Study Lb redbridge white Case Study Peabody White Case Study Scottishpower white 150px Case Study Shelter 2024 Case Study