Many companies rely on support chatbots that simply perform static knowledge base searches—barely scratching the surface of what AI can achieve in customer support.
At Relevance AI, we've taken a fundamentally different approach with our Customer Support Agent. We've built an intelligent multi-agent system that goes far beyond the typical AI chatbot connected to a knowledge base—it transforms support operations through sophisticated automation that enhances both customer experience and product development.
Building our Customer Support Multi-Agent System
Our Customer Support AI Agent uses a sophisticated multi-agent approach tailored specifically to automate the understanding of customer issues, respond to complex support requests and proactively share feedback to our product development team. Here's how it works:
Context is Everything
When a support ticket comes in, our agent doesn't immediately fire back a templated response. Instead, it first gathers comprehensive context about the issue. It captures the complete conversation history, including images or videos. This means it can process screenshots of the Relevance platform or other integrated platforms, error messages, transcribe and understand Loom videos, and even extract information from links to agents that users are having trouble with.
This depth of context gathering fundamentally improves the quality of responses we can provide. Our agent builds a complete picture of the problem upfront. This enables our agent to more accurately problem-solve and eliminates the frustration of fragmented support conversations.
Ticket Routing Intelligence
At the heart of our multi-agent system is an intelligent routing process powered by specialized sub-agents working together. The Ticket Router Agent acts as the coordinator, determining which specialized agent should handle each request based on the conversation history.
Each request is directed to the most appropriate, specialized agent:
- Integrations: When a customer asks about integrating with specific apps/software or AI models, the AI agent verifies our support of them and provides tailored guidance.
- AI Agent Use Cases: If someone needs help building a specific agent that they have in mind, our system invents the agent crafts the core instructions and the exact tools required for their use case, and they receive the link to the newly built agent
- Error Messages: When users share an error message they’ve received, our system identifies and provides the specific troubleshooting steps to guide the user to a full resolution
- All Other Enquiries: For general inquiries, the agent system searches through our knowledge base of product documentation and FAQs to deliver precise, relevant information to the user
What sets our approach apart is its ability to go beyond the one-dimensional search of a static knowledge base. Instead of relying on generic responses, our multi-agent system intelligently analyzes context and coordinates specialized sub-agents to deliver nuanced, tailored support.
Closing the Loop: Respond, Monitor and Improve
After gathering context and routing the request to the appropriate specialized sub-agent, our support agent synthesizes all insights to generate a personalized response for the user. But, the process doesn't end there.
Every interaction is automatically logged in Airtable, creating a robust system for analytics, performance monitoring, and assessing the overall health of our customer support operations.
More importantly, these support interactions feed directly back into our product development process. The detailed context, specific issues, and patterns of requests become a goldmine of product improvement opportunities.
What's Next
We're continuously building on our Customer Support Agent's capabilities. We’ll be exploring adding billing management automation and a Discord integration to expand available support channels. We also plan to build a proactive error message handling feature that will notify users within the platform when an error occurs, often before they even need to contact support. Perhaps most exciting is our work on browser automation capabilities that will allow our agent to perform hands-on troubleshooting of customer issues in real-time.
In Summary
Our Customer Support Agent improves both the customer experience and our internal product development process by capturing rich context, intelligently routing requests, and systematically logging interactions. We believe the future of AI-powered support goes far beyond generic AI chatbots. Our multi-agent system redefines customer service by delivering deep, context-aware insights that not only resolve issues but also fuel continuous product improvement.
Watch it In Action
