Voice agent system improves public service calls with natural speech
VISAGENT: An Agent-Based End-to-End Voice Interaction System
Computation and Language
Summary
Many public phone services still use rigid menus that are hard to navigate. The authors created VISAGENT, a voice system that talks naturally, understands what you want, and sends your request to the right expert agent quickly. It remembers what you told it before, so you don't have to repeat yourself. They tested this system on a busy Chinese hotline and showed it works fast and sounds natural.
What this means in practice
- •For public service operators: Deploy flexible voice interfaces that replace menu navigation with natural speech across multiple service domains.
- •For call center developers: Integrate a scalable agent routing system to route customer requests dynamically to specialized domain agents within a large agent ecosystem.
- •For smart speaker manufacturers: Enhance devices with a marketplace model for adding and updating conversational agents dynamically to extend functionality and personalization.$Commercial implications: Enables products selling dynamic agent onboarding and personalized voice services through an extensible agent ecosystem.
Authors
Xiang Wang, Zhaoxiang Liu, Xiang Chen, Liqiang He, Kai Wang, Shiguo Lian
Abstract
Voice-based public services have evolved from manual switchboards to intelligent platforms to meet modern user needs. However, current systems remain constrained by rigid menu-driven interaction, fragmented vertical services, and limited extensibility. To address these issues, we present VISAGENT, an agent-based end-to-end voice interaction system for large-scale public services that integrates end-to-end speech interaction, hierarchical agent routing, a scalable voice-agent marketplace, and user memory support. The system replaces menu-driven workflows with natural speech and dynamically dispatches user requests to either a general conversational model or specialized domain agents via a two-level routing paradigm that achieves sub-second intent-to-agent selection over hundreds of agents. User memory informs preference-aware agent selection and supplies relevant preferences and context to selected agents, supporting personalized task execution and reducing the need for users to repeat previously provided information. A voice-agent registration mechanism converts task agents into conversational entities, enabling dynamic onboarding and ecosystem-level extensibility. We validate VISAGENT through a real-world deployment in China's 114 hotline across multiple service domains. Experimental results demonstrate low-latency full-duplex interaction, robust routing scalability, and competitive speech naturalness. More broadly, this work advocates a shift from isolated voice assistants toward agent-centric conversational infrastructures, providing a general blueprint for building extensible voice-agent systems in public services and beyond.