Unleashing the power of generative AI: Verisk’s journey to an Instant Insight Engine for enhanced customer support

This article describes Verisk's journey in developing an AI companion for enhanced customer support using generative AI on AWS. It covers how Verisk implemented a Retrieval Augmented Generation (RAG) approach with AWS services like Amazon Comprehend, Amazon Kendra, Amazon Bedrock, Amazon Rekognition, and Amazon Transcribe to build an "Instant Insight Engine" for their FAST platform.

Specifically, the article covers:

• The opportunity to scale customer support and free up staff for innovation
• The approach of using RAG over fine-tuning for access to dynamic data, multiple sources, reduced hallucination, and transparency
• The solution architecture involving various AWS services and Anthropic's Claude model
• Data harvesting, structuring, and retrieval techniques like chunking and multi-step querying
• Experimenting with LLM parameters and models for improving response quality
• Metrics, data governance, and techniques for evaluating accuracy
• The business impact of rolling out the AI companion to beta customers
• Conclusion on the potential of generative AI for transforming customer support