In our latest blog, Tim Pearson caught up with Julian Williams at Anthropic to explore the science of conversations and how the increasing adoption of generative AI in customer service can improve consumer engagement.I had to call my broadband provider the other day, an activity that I always put off as it takes an eternity. Ages spent on hold listening to hits of yesteryear while being told my call is important and how much I am valued as a customer and then, if I get through without the call timing out, I invariably get diverted to another department as the operator doesn't have access to the data/ information/ fault tracking information etc. and round we go again. These conversations are usually pretty trying, made more so as you know full well that the agent on the other end of the phone is probably running against the clock to close your call as quickly as possible and is furtively typing away looking for the scripted answer to your problem.
That got me thinking.
When Is A Conversation Not A Conversation? At a customer level, are these really conversations? Is there really a dialogue where both parties discuss something, or is it simply just a processed conversation driven by a consumer need that's responded to by a human just reading out a carefully scripted response rather than applying any sentiment themselves? If so, I might as well use the other tool of choice by today's telecom providers, the chatbot. Can a virtual agent help me in ways where more traditional models are struggling?
That was a question I put to our good friends at Anthropic. Their creation of the Large Language Models (LLM) used in generative AI, such as Claude 3 as used in our own NAGRA Insight Negotiation Agent, are designed to deliver an engaging online experience. Their use cases include topics as diverse as troubleshooting tricky consumer tech, answering contact center queries or segmenting audiences to create marketing campaigns. Their company ethos is simple - they recognize the vast impact AI will have on the world and are dedicated to building systems with their partners that people can rely on while also generating research about the opportunities and risks of AI.
The goal isn't to simply replicate human conversation, but to create genuinely helpful interactions that adapt to each unique situation. When we talk about AI assistants like Claude, we're focused on moving beyond scripted responses to deliver meaningful, contextual help that truly addresses the customer's needs. The most effective AI conversations aren't just exchanges of information - they're collaborative problem-solving sessions where the AI understands both the explicit request and the underlying context to provide genuinely valuable help. Julian Williams, Telecommunications Industry Lead, Anthropic
Facilitating the Conversations of Tomorrow From a NAGRAVISION perspective, over 500 operators and 400m consumer devices use our technology daily across multiple distribution models. This allows us to really drill down on how we can help our customers optimize revenue and churn while delivering the most engaging customer experience possible.
Conversations that arise from today's multi-service operator offers can be complex. We want to empower consumers to get a full explanation from virtual agents of something they don't understand on their bill or be able to negotiate directly to get the package they want, subject to the operator's guardrails. Through our solution, consumers can resolve issues faster and our customers can improve customer satisfaction and when deployed at scale, realize significant revenue and cost benefits.
To do this, NAGRAVISION has gone beyond chatbot territory and developed a proprietary digital twin model that's designed to be used in the paid entertainment environment. We also selected the right LLM that fitted the conversation profile and output we wanted. Working with our strategic partner Amazon Web Services (AWS), we started working with Anthropic and identified Claude 3 as the best LLM for our purpose.
The NAGRA Insight Negotiation Agent The NAGRA Insight Negotiation Agent communicates in natural language with the consumer who is logged into their account on the operator's app or authenticated via a social media channel. When a consumer asks a question such as Why is my bill so high this month?', the Negotiation Agent looks up their account and responds with a list of their services broken down so that each service can be identified. The Negotiation Agent then continues the dialogue and can offer adjustments, new deals or upgrades based on the consumer's likely reactions predicted by their digital twin, albeit while constrained by operator-defined guardrails. The benefit for the consumer is that they have a results-focused conversation that gives them the package they want, while the operator benefits through reduced calls into the contact center.
The Negotiation Agent is a solution from the NAGRA Insight product line. Enabled through chatbots, social media and voice conversations, the solution relies on the integration of the operator's CRM data. The consumer's digital twin then directs a LLM that powers the NAGRA Negotiation Agent. This interprets consumer input and responds based on the guidelines defined by the operator. SaaS-based, the solution uses Amazon Bedrock which guarantees that no data is ever provided to Anthropic as the LLM provider. This ensures privacy and confidentiality while being able to scale to millions of conversations.
Maximizing Brand and Service Loyalty Ensuring consumers have access to tools that answer their questions, either as a physical conversation or digital dialogue is core to driving brand and service loyalty. The science of these conversations is complex but the latest advances in technology
