How AI is Elevating Customer Service: Learn to Prevent Escalations

Contact centers play a crucial role in many businesses, but their high turnover rates necessitate ongoing training for new agents in operational, behavioral, and compliance aspects. Among these, honing behavioral skills is paramount, as they directly influence customer experience and satisfaction. A skilled agent with these qualities can navigate difficult situations, build rapport, and leave customers feeling heard and understood. Satisfied customers are more likely to be loyal and repeat business. A highly behaviorally adept agent can often overcome shortcomings in operational knowledge by effectively gathering information and seeking appropriate solutions. They can also navigate situations where strict compliance might not be the best course of action for customer satisfaction. Gerald Zaltman, professor at Harvard Business School, claims 95% of purchase decisions are subconscious, driven largely by emotions.

Vonage Global Customer Engagement Report 2024 reveals that 74% of customers are likely to take their business elsewhere in case of poor interaction experience. This means that businesses rely on emotional intelligence for success, and that if contact center agents lack these skills, they could negatively impact CX outcomes.

In this blog, we are showcasing the potential of generative AI to transform agent training by creating more realistic and personalized training experiences that can help agents develop the skills to interact with a disparate range of customers in a variety of scenarios and emotions, thereby improving emotional intelligence and knowledge on how to prevent escalations and earn the customer’s trust in any situation.

De-escalation hurdles in contact centers

De-escalating difficult situations in contact centers is often challenging, especially when facing customers who are already upset, frustrated, or even angry due to issues they've encountered with the product or service. Limited call time, complex operations and policies, and lack of in-person interaction can bring further complications. Building rapport and effectively communicating empathy can be even more difficult over the phone, making it harder to calm the customer. These factors create a demanding environment where de-escalation skills are essential for maintaining positive customer interactions and a productive work atmosphere.

McKinsey’s 2022 State of Customer Care Survey has found that improving customer experience is the fastest-growing priority area for customer care leaders.

We believe a key factor in the decline in customer experience is the challenge in providing agents with de-escalation techniques within the constrained training timeframe. Traditional training methods, such as classroom instruction and role-playing exercises, can be time-consuming and expensive, and they are often not that effective in reaching all learners. New methods like simulation-based scenarios, in which agents practice handling customer interactions in a simulated environment through historical calls, have become a popular training method. However, they too are limited in effectiveness, as they often lack the realism and interactivity that are necessary to fully prepare agents for real-world interactions.

Consequently, limited training on emotional intelligence and de-escalation techniques can hinder agents’ abilities to manage difficult customer interactions. This can lead to poor customer experiences, especially for aggressive, frustrated customers or ones from diverse cultural backgrounds. Without support and tools, challenging situations can also reduce job satisfaction for agents.

How can AI help in enabling agents on de-escalation?

The power of generative AI and speech AI models can be leveraged to create realistic and engaging simulations of customer interactions, providing contact center agents with the opportunity to practice their skills and handle a wide range of difficult scenarios before facing real customers.

Here are the benefits of using such simulated calls for contact center training:

How can this be realized?

Generative AI and speech services (Text to Speech/Speech to Text) can create virtual contact center environments that provide agents with a completely immersive training experience. This ensures they are well-prepared for their actual jobs. Here is how it typically works:

Introducing Infosys Cortex learn to de-escalate situation, accelerated by NVIDIA AI Infosys Cortex is an artificial intelligence (AI) driven customer engagement platform that transforms contact center operations through purposeful communication and smart decision-making capabilities. Infosys Cortex adds value to all aspects of the contact center process, from agent onboarding and training to empowering agents during live calls and deriving insights from conversations for improved operations — helping improve the customer experience.

Infosys Cortex provides an AI-as-a-customer feature with a no-code configuration-driven studio to create any combination of scenario (e.g. billing issue, troubleshooting etc.) and customer characteristics (gender, age, language preference, emotion etc.), which can be used to provide simulated training environments on the agent desktop. Learn to de-escalate situation is one of the use cases in the AI-as-a-customer feature where agents can be trained on emotional intelligence through difficult customer conversations. For example, agents can easily practice conversation on “a customer calling third time for billing dispute.”

Infosys Cortex also provides enterprise integrations and proactive/on-demand assistance on the agent desktop during the conversation to help trainees become familiar with operational aspects. Furthermore, Infosys Cortex provides conversational analytics on these practiced simulated calls to provide personalized feedback to the trainee on performance, as well as recommendations for improvement in emotional intelligence.

Infosys Cortex also has a feature called Language Neutralization that enables effective communication between the customer and agent when they do not speak the same language, removing language barriers and facilitating smooth interaction. Learn to de-escalate situation is also supported with Language Neutralization to train agents on emotional intelligence when interacting with a difficult customer speaking a foreign language.

Infosys Cortex leverages NVIDIA RIVA, a collection of cutting-edge speech and translation AI microservices, to power learning de-escalation techniques, as well as, the newly released AI virtual assistant for customer service NVIDIA NIM™ Agent Blueprint, to scale operations.

The high accuracy of RIVA automatic speech recognition (ASR), neural machine translation (NMT), and engaging text-to-speech (TTS) empower accurate and natural communication. Powered by NVIDIA accelerated computing for model fine-tuning and processing, RIVA helps reflect emotion in speech synthesis, allowing agents to practice de-escalation scenarios with AI-generated customer voices that transition through emotional states, such as from anger to calmness. Infosys Cortex integrates the NVIDIA RIVA Conformer ASR for speech recognition and RADTTS for speech synthesis, providing a versatile platform for training and development.

Furthermore, integration is supported with the AI virtual assistant for customer service NVIDIA NIM™ Agent Blueprint for deploying large language models that can act as customer in the conversation while maintaining data integrity and governance.

Benefits of NVIDIA RIVA

NVIDIA Riva services have helped address some key challenges Infosys has faced in enabling emotional intelligence. The following are some key areas RIVA helps address:

• Accuracy: Fine-tuning support on domain language, different accents and pronunciation.
• Language barrier: Support for all major languages with consistent addition of support for new languages.
• Data privacy: On-premise deployment helps to keep sensitive data secure and private.
• Cost reduction: very cost-effective compared to hyperscalers pay-as-you-go pricing as volumes increase.
• Realism: Support for emotions like anger, calm, surprise etc. to make it sound like human interaction during difficult conversations. Also, emotional state transition makes de-escalation learning more effective.
• Low latency: On-premise deployment enables low latency thereby providing a better training experience. Here are the reference performance results for speech recognition and speech synthesis.

Key Takeaways

Learn to de-escalate situation is a transformative approach for contact centers, providing a scalable, cost-effective, and efficient solution for simulated learning with AI-as-a-customer.

The powerful learn to de-escalate situation offered by Infosys Cortex, based on NVIDIA RIVA speech AI and the AI virtual assistant for customer service NVIDIA NIM™ Agent Blueprint, enables contact center agents to improve their emotional intelligence, which is important for business success through better customer experience and satisfaction. Developers can experience NVIDIA RIVA NIM microservices and AI virtual assistant for customer service NVIDIA NIM™ Agent Blueprint on build.nvidia.com, or Learn more.

About the Authors

Samit Sawal

Samit Sawal is a Senior Architect with Infosys Center for Emerging Technology Solutions and has over 17 years of experience which includes incubating emerging tech, building IP, accelerators, platforms, and product engineering with a strong understanding of technologies such as Conversational AI, Generative AI, and domains like Customer Service and Core Banking.

Vishal Manchanda

Vishal Manchanda is a Senior Principal Architect with Infosys Center for Emerging Technology Solutions and has over 24 years of experience in the IT Industry. With expertise in Conversational AI, Vishal has been developing, architecting, and incubating various IT solutions/IPs around contact centers, personalized intelligent interfaces involving hyper contextual personalized videos, voice interfaces and comprehensive Conversational AI platforms.

Anand Santhanam

Anand is Senior Vice President and Head of Strategy for Communication, Media and Technology segment at Infosys. He is a 28 year globally tenured professional at Infosys, spanning varied roles and geographies with a specialization in the Telecommunications industry. He is leading a transformative Platform Incubation Engine group which brings Sentience, Learning and Business value into traditional Enterprise Services domains. He also heads a portfolio of large American Telecom and Cable provider accounts in the US. He has lived and worked in Canada, USA, Australia, NZ, Middle East and SE Asia (Singapore) in Engineering, Program Managing, Consulting, Client Partnership and Market Development roles. He is a Computer Science Engineer and an Infosys Constellation leader with programs at Stanford Graduate School of Business.

Prakruthi B Gowda

Prakruthi B Gowda is a Deep Learning Solution Architect at NVIDIA with five years of expertise in conversational and generative AI. She specializes in developing and architecting advanced solutions for contact centers and LLM-driven chatbots, utilizing cutting-edge technologies such as LLM, natural language understanding (NLU) in domains like customer service, Banking, Telco. Her work also involves integrating innovations in AI like automatic speech recognition, real-time language translation, and leveraging GPU computing to enhance the performance and scalability of AI-driven communication systems.

References

• NVIDIA. (2023, May 30).
• How Language Neutralization Is Transforming Customer Service Contact Centers
• Infosys Cortex (2021, January 20). Infosys Launches Cortex
• AI-as-a-Customer, The future of contact center training (2024, July). Infosys Cortex
• Harvard Business School (2003, 13 Jan). The Subconscious Mind of the Consumer (And How to Reach It) – Retrieved from hbswk.hbs.edu
• Vonage (2024). Vonage Global Customer Engagement Report 2024 – Retrieved from vonage.com
• McKinsey’s 2022 State of Customer Care Survey – Retrieved from mckinsey.com