Guest Alex Kvamme, CEO @GetPathlight, talks about using AI to unlock market insights from discussions with customers and partners. They discuss the concept of conversational intelligence and its application in organizations, the trade-off between time and resources in gaining customer insights, where data in most organizations is located and how to get started, and the framework behind AI operations and lifecycle management.
Conversation intelligence platforms like Pathlight use AI to analyze and summarize millions of customer conversations, providing businesses with a comprehensive view of what their customers want and need.
Generative AI, specifically the transformer model, enables businesses to analyze and process large quantities of customer conversations, extracting valuable insights faster and more accurately than traditional methods, helping organizations make informed decisions based on real-time information and a deep understanding of their customers.
Deep dives
What is conversation intelligence and how does it benefit organizations?
Conversation intelligence is the analysis of customer conversations to extract insights, trends, and risks that are impactful to a business. This technology allows companies to have a better understanding of their customers' needs and make informed decisions. It covers various areas such as sales conversations, customer support, product quality, and supply chain issues. Conversation intelligence platforms like Pathlight use AI to analyze and summarize millions of customer conversations, providing businesses with a comprehensive view of what their customers want and need.
The power of generative AI in conversation intelligence
Generative AI, specifically the transformer model, has revolutionized the field of conversation intelligence. It enables businesses to analyze and process large quantities of customer conversations, extracting valuable insights in a much faster and more accurate way than traditional methods. Through transfer learning, generative AI models can quickly identify patterns, trends, and risks, providing organizations with real-time information and deep understanding of their customers. This technology reduces the time to implement and start seeing results, allowing businesses to make better decisions based on the knowledge extracted from their conversations.
Overcoming the challenges of scaling conversation intelligence
Scaling conversation intelligence poses unique challenges due to the complexity of the underlying infrastructure required. Building a system that can handle millions of conversations, process them, and generate meaningful insights at scale requires dedicated infrastructure and technology. Pathlight, for example, has developed its own middleware and infrastructure layers to manage models, prompts, and data processing pipelines. The focus is on solving customer problems and providing actionable insights rather than the intricacies of the backend infrastructure. This technology enables companies to uncover valuable insights, solve customer issues across various departments, and achieve better results with fewer resources.
Topic 1 - Welcome to the show. Alex, Tell us a bit about your background.
Topic 2 - What is the concept of conversational intelligence and how does it apply to most organizations today? What problem is it trying to solve?
Topic 3 - I would think there is a trade off between time and resources to get to a customer issue vs. the value of that insight. How does an organization weigh the opportunity cost? How do you keep the insights generated from being overwhelming
Topic 4 - Let’s move from the concept to practical. Where is the data in most organizations today that will yield results and solve problems? How would you suggest folks get started and what use case are they likely to implement first? Is this data that humans either can’t or won’t get too because it is an enormous amount or maybe too tedious to pay for?
Topic 5 - How does all of this work under the hood? Is this one model or multiple models working in parallel? Is there a framework for the operations and lifecycle managed by an organization?
Topic 6 - Let’s talk about what it takes to get an LLM into production. The the rise of LLM’s and foundational models such as Llama2, there is an interest for organizations to use LLM’s, but going from concept to production still has a high barrier to entry. It’s easy to download a model, it’s much harder to either fine-tune it or set up RAG with a vector database for your specific use case. Would you agree and what are your thoughts?