ElevenLabs AI Updates: November 3, 2026

1. ElevenLabs Production Data Shows Voice Agents Beating Chat on Conversations Longer Than Two Minutes

ElevenLabs compared voice (phone and WhatsApp) with chat across production ElevenAgents deployments that each handled at least 50 conversations on both channels between May and July 2026, judging each agent against success criteria its business defined. Chat did about 9 percentage points better in conversations under a minute, the two were roughly even between one and two minutes, and past two minutes voice success was 10.9 points higher (a 31.7% relative increase); successful long conversations averaged 4.6 minutes on voice versus 7 minutes on chat. Across 315 agent conversations that moved to a faster LLM, a 16% cut in average end-to-end latency coincided with a 26% lift in success rate, and the share of calls ending within 15 seconds fell from 21.1% to 3.8%, which ElevenLabs reads as callers hanging up when the first response is slow. In a separate 98-person study of a simulated cancelled-flight rebooking, 85% preferred the voice agent over the chat agent; the figures are vendor-reported, and the post notes that conversation length is an imperfect proxy for complexity, especially for chat. Source

2. KPN Builds Its Mobiel AI Business Phone Assistant on ElevenLabs and Lifts Dutch Verification Accuracy to 90%

ElevenLabs published a write-up on KPN, the largest telecom operator in the Netherlands, whose Mobiel AI add-on for KPN EEN MKB business mobile subscriptions runs on ElevenLabs voice AI inside KPN’s existing mobile app. Business owners keep their number and plan and get voicemail transcription, a transcript and summary of every call with follow-up actions captured, a call menu that routes callers to a colleague or department, and a status setting that controls when they are reachable. To handle Dutch compound words and regional accents, the two teams gathered more Dutch-language data and fine-tuned the transcription model for short, high-stakes sequences, raising verification accuracy from roughly 50% to 90%, an improvement ElevenLabs says now applies to every Dutch-language deployment on its platform. Mobiel AI runs on KPN’s network and in KPN’s own data centers with voice data kept in the Netherlands, customer data is not used to train models, and the partnership also covers KPN’s agentic customer care, internal tools and product development. Source

3. Angle Health’s ElevenAgents Pilot Resolves More Than 90% of Provider Calls Without a Transfer

ElevenLabs published a case study on Angle Health, which administers health plans for small and midsize employers and whose provider team fields thousands of calls a day, about 80% of them from provider offices. Two voice agents built on ElevenAgents answer in eight languages and connect directly to Angle Health’s core administration system: one verifies callers, pulls live claim and benefit data and submits prior authorization requests, while the other handles eligibility, benefits, claim status and member ID requests, and the agents call offices back when an escalated claim is resolved. The deployment runs under a business associate agreement with HIPAA safeguards including ElevenLabs’ Zero Retention Mode, and Angle Health says ElevenLabs gave the most accurate transcription and task execution on clinical language and diagnosis codes among the vendors it tested. During the pilot the agents handled about 80% of total provider call volume, resolved more than 90% of those calls end to end, and freed nearly 9,000 minutes a day for the provider team; Angle Health plans to expand the pilot and to use ElevenAgents on its commercial, brokerage and account management sides. Source