Peter Jang
January 18, 2025 10:38
ElevenLabs has enhanced customer support through AI-based solutions by successfully implementing voice agents that resolve more than 80% of user inquiries every day.
According to ElevenLabs, the company achieved a resolution rate of over 80% by introducing voice agents designed to efficiently handle user inquiries related to documents. The voice agent has had significant success handling user queries, handling approximately 200 calls per day.
Performance and Evaluation
Speech agents based on large language models (LLMs) were evaluated for their ability to effectively resolve or redirect inquiries. Human verification of 150 conversations showed an 81% agreement between LLM and human evaluators for successfully resolved inquiries. Agents also demonstrated 83% agreement to adhere to the knowledge base.
Additionally, 89% of relevant support questions were answered or correctly redirected by document agents, demonstrating their ability to manage simple queries.
Strengths and Limitations
strength
LLM-based agents excel at addressing specific questions that match well with available documentation. It’s useful for questions like API endpoints, language support, and federated queries because it effectively directs users to relevant pages and provides initial guidance for complex queries.
To optimize performance, ElevenLabs recommends targeting users with clear questions and utilizing redirects for more complex inquiries to increase the efficiency of your support process.
Restrictions
Despite these advantages, agents struggle with vague or account-related inquiries that require deeper investigation. Since voice medium is not suitable for sharing code or handling complex technical issues, ElevenLabs suggested redirecting users to documentation or suggesting a support channel for such queries.
Development and Configuration
The voice agent consists of system prompts that guide you through responses, allowing you to stay focused on your ElevenLabs products. A comprehensive knowledge base, including summarized versions of all documents, helps LLM provide accurate answers.
Three basic tools are integrated into agent functionality: redirect to external URL, email support, and documentation to provide multiple paths for user inquiries. The agent’s evaluation tool evaluates conversations against predefined criteria to ensure continuous improvement and stability.
continuous improvement
ElevenLabs recognizes the limitations of LLM in solving all types of queries, especially in a fast-evolving startup environment. However, the company emphasized the benefits of automation, freeing up teams to focus on complex challenges as the community expands the potential of AI audio technology.
Powered by ElevenLabs conversational AI, the agent serves as an effective tool for exploring product and support questions, constantly improving through automated and manual monitoring, reflecting the company’s commitment to improving the user support experience.
Image source: Shutterstock