Webinar Recap: How Insurely Brought Voice Agents to Its Customer Center

Insurely introduced ElevenLabs Voice Agent to reduce the slow, frustrating experience customers face when calling insurance companies. In a CCW Digital webinar, ElevenLabs' GTM leads Jack Piunti and Josh Pimentel, along with Insurely CTO Colby Garland, shared the real-world implementation case.

In insurance, phone consultations remain a critical channel. But existing IVR systems or simple chatbots only process what the customer says, without picking up on emotional context — like whether someone is panicked about a roof leak or worried about a lapsing policy. Insurely used ElevenLabs' proprietary speech-to-text model to annotate emotional signals such as frustrated / relieved / confused from calls, enabling the agent to respond not just to words but to emotion.

The first agent, Emily, calls customers who started but didn't finish a rental home insurance quote, guiding them through the form step by step and breaking down insurance questions into plain language. As a result, customers completed quotes they would have otherwise abandoned, and Colby explained that it was Emily's encouraging conversational style, more than simple answers, that contributed to the conversion rate.

Insurely then expanded into the customer support area, and the agent now operates 24/7. As a result, support case volume decreased, revenue increased, and human agents were freed to focus on more valuable conversations instead of repetitive inquiries. This created a structure for growing throughput without adding headcount.

The most technical part is handoff and call routing. Business hours are Monday–Friday, 8 AM–5 PM, and when a customer wants a human agent, the AI checks the current availability of in-house staff.

• During business hours, the call is transferred via Twilio, and the full transcript is sent to internal systems in real time so the agent can pick up the context immediately.
• Outside business hours, the AI books a callback meeting within HubSpot, and automatically leaves a transcript and summary notes during the conversation.

The operating principles shared in the webinar were also clear. An agent should focus on only one role, the knowledge base must be up-to-date and specific, and data issues should be solved before the conversation, not during it. For example, instead of having customers spell out their email during the call, existing contacts should first be pulled in via CRM integration, and time zone should be explicitly set before scheduling an appointment.

Finally, ElevenLabs made it possible to define success criteria and categories for each conversation type in the analytics section, and Colby also built a separate analytics tool using the API. The key point is that structured measurement leads to structured improvement.