Technical Deep Dive: Leading Household Appliance Brand Achieves 79% Successful Answer Rate with GenAI Bot - Here’s How
In today’s competitive market, delivering exceptional customer service is crucial for maintaining brand loyalty and satisfaction. A prominent leader in the household appliance industry recognized this need and partnered with NeuraFlash to significantly elevate their contact center capabilities through the power of Generative AI (GenAI) with Amazon Bedrock, Amazon Connect, and Amazon Lex. This collaboration delivered outstanding results, achieving a 79% successful answer rate for FAQs, a 74% containment rate, and a 36% success rate for troubleshooting—significantly outperforming the 0% success rate of non-genAI bots.
The Challenge: Building a Future-Ready Customer Service Foundation
This household appliance company had several key objectives for their customer service operations:
- Improving Deflection Rates: They sought to enhance their self-service capabilities to reduce the number of customer inquiries handled by human agents, ultimately lowering operational costs.
- Increasing Accuracy and Resiliency: A solution was needed that could consistently deliver precise information across their expanding and evolving product line.
- Creating a Path to Higher Self-Service Resolution Rates: Their goal was to empower more customers to resolve their issues independently, reducing the need for live agent support.
In addition, they wanted to implement an analytics framework to identify trends and product issues based on customer inquiries. This would not only refine their customer support, but also provide valuable insights for ongoing product development.
The Solution: Leveraging Generative AI for Superior Customer Support
NeuraFlash designed and implemented a solution leveraging Bedrock, Lex, and Amazon Connect Chat.
Bedrock KB Support:
The solution integrates Bedrock KB support to manage the orchestration of both indexing and query retrieval processes. This allows for efficient handling of a wide range of content types, including converted ZingTree data (such as FAQs and Troubleshoot guides), Product Manuals, and Quick Start Guides. The system also enables retrieval filtering by specific product lines and document categories, ensuring relevant information is provided to users based on their queries.
Large Language Model (LLM):
The solution incorporates Anthropic Claude V3.5, accessible via Amazon Bedrock. This LLM is utilized through the Converse API, which includes tooling support for orchestration. This setup allows for sophisticated conversational capabilities, enhancing the solution’s ability to understand and respond to user inquiries with greater accuracy and relevance.
Testing and Prototyping Tools:
PromptFoo is integrated into the solution to facilitate testing, using the same code pipeline as Lambda (Lex) and Streamlit. This consistency ensures that the testing environment closely mirrors the production environment, reducing the likelihood of issues when deploying updates. Additionally, Streamlit is employed for rapid prototyping, enabling quick modifications and experimentation with inference models, retrieval filters, and other adjustable parameters, allowing the solution to adapt and evolve efficiently.
Deployment through Amazon Connect:
The solution was deployed within an Amazon Lex framework integrated with Amazon Connect, enabling a chat solution on the SharkNinja website for end user customers to interact with the solution. By leveraging Amazon Lex and Connect, the solution ensures that customer interactions are not only handled in real-time but also enriched with precise, contextually relevant information drawn from the knowledge base. The integration also allows for sophisticated query handling and dynamic content retrieval based on specific product lines and document categories, providing a highly personalized and efficient support experience.
Reporting and Analytics:
We utilized Amazon Connect to gather Contact Lens data, which includes sentiment analysis and call transcripts, and combined this data with Lex dialog hook logs and processed the transcripts using the Anthropic Claude LLM to score them. All this information was then fed into a data analytics solution built with Amazon Athena.
This analytics setup provides the customer with detailed insights, including the intent behind interactions (such as FAQs, troubleshooting, or support), the resolution state of each interaction, and a breakdown by product.
The comprehensive system was created to:
- Enhance Customer Self-Service: Fast, accurate responses empower customers to resolve issues independently, reducing the need for live support.
- Ensure Consistency in Documentation: Standardized product information delivers uniform, reliable support across all channels.
- Improve Feedback to Engineering: Captures customer interactions to provide actionable insights for product refinement.
- Provide Leadership Analytics: Offers leadership valuable insights into customer trends and service effectiveness through a robust analytics framework
The Impact: Transforming Customer Service with AI
The collaboration between NeuraFlash and this leading household appliance brand has delivered impressive results, setting a new standard for customer service:
- Achieved a 79% success rate for FAQs, drastically reducing human intervention.
- Empowered customers with a 74% containment rate, cutting costs and boosting satisfaction.
- Increased troubleshooting success to 36%, up from 0% with non-genAI bots.
- Enhanced brand reputation with 24/7 accurate support, driving customer loyalty.
Looking Ahead: Expanding the Partnership
NeuraFlash’s partnership with this household appliance leader marks the start of an exciting journey. The focus is on continuously integrating more AWS solutions to further streamline customer service, lower costs, and elevate the customer experience through cutting-edge self-service and case deflection.
NeuraFlash at a Glance:
NeuraFlash is a leading AWS and Salesforce partner, building AI-powered experiences that transform customer service. Our mission is to help businesses reimagine how they deliver personalized service and sales experiences that scale, empower teams, and uncover revenue. Working across industries, NeuraFlash enables intelligent business automation that drives real-time ROI.
