ElevenLabs adds synthetic voice detection to ElevenAgents: government hotlines can tell whether a caller is human or AI before deciding how to respond

Government hotlines can now judge during a call whether the caller is a person or an AI agent calling on someone's behalf, and route accordingly: human callers move to the front of the queue, AI callers go to a dedicated agent with specific instructions and limits, and sensitive actions such as password resets or account access can be blocked. Citizens were increasingly turning to AI agents to handle interactions with government, while most governments were not equipped for this shift in volume; complex submission requirements and written documentation had long suppressed demand, and phone menus and transfers made the experience worse. Researchers tracked 84 cases across 11 countries where public services faced surges in filings, appeals and complaints, with officials increasingly pointing to AI as the cause; the researchers call this AI-driven rise in the volume or complexity of requests "agentic flooding."

Responses split into two options: adding friction through fees, IP blocks and stricter forms, or adding capacity by meeting agents with agents. Friction can work but decreases accessibility for the people who need the services most; adding capacity can significantly reduce administrative burden and increase throughput, and organizations that build it gain more signal and a higher-resolution picture of what citizens actually want and need. In Midland, Texas, a city serving 138,000 residents and receiving 3,000 calls a day, the city used ElevenAgents to build Jacky, an AI civic concierge that answers commonly asked questions and supports the multilingual population, and added a website chat function to further reduce phone volume.

In the Czech Republic, ElevenLabs supports national benefits and employment hotlines with AI agents that handle roughly 5,000 calls per day. These agents independently resolve around 85% of calls while dramatically increasing answer rates from about 60% to near-universal coverage. This deployment and Midland's Jacky are both public-service uses of ElevenAgents, one for a national hotline and one for a single city.

The material shows government public-service hotline settings: Midland, Texas, with 3,000 calls a day serving 138,000 residents, and Czech benefits and employment hotlines handling roughly 5,000 calls per day. It addresses government agencies and public-service operators that need to handle AI-agent callers, and suggests starting with simple use cases such as sentiment analysis of public comments or fraud detection before expanding to intake screening, triage and automated correspondence.

The material describes "agentic flooding" as researchers' characterization drawn from 84 cases across 11 countries, with officials "increasingly" attributing surges to AI rather than case-by-case confirmation. The Czech hotline's roughly 85% independent resolution rate and the rise in answer rates from about 60% to near-universal coverage are deployment figures given in the material, without stated measurement definitions or time range.