BUSINESSDevelopments · · 2 min read · Source: ElevenLabs
ElevenLabs data: voice beats chat on long customer requests
ElevenLabs published an analysis on November 2, 2026 showing that AI voice agents outperform text chat in longer customer conversations. For short requests, chat comes out ahead.
Key facts
- ElevenLabs published the analysis on November 2, 2026, based on customer service agents from May to July 2026 and a user study with 98 participants.
- In conversations over two minutes, voice agents resolved 10.9 percentage points more cases than chat, per the vendor; under one minute, chat led by about 9 points.
- This is a vendor study with no published raw data.
What ElevenLabs compared
The voice AI company analyzed customer service agents that ran both on phone or WhatsApp calls and in chat, using data from May to July 2026. Each agent was compared only against itself: voice versus chat, same company, same definition of success. A user study added 98 participants who had to rebook a canceled flight once with a voice agent and once via chat.
The key numbers
- Under 1 minute: chat was about 9 percentage points more successful.
- 1 to 2 minutes: both channels were roughly even.
- Over 2 minutes: voice was 10.9 points ahead, which ElevenLabs puts at 31.7 percent more resolved cases.
- Speed: successful long conversations took 4.6 minutes on average by voice and 7 minutes by chat.
- Preference: 85 percent of study participants preferred voice.
The analysis also found that when agents switched to a faster language model, average response time fell by 16 percent and the success rate rose by 26 percent. That is a correlation, not proof of cause and effect.
How to read it
The numbers come from a vendor that sells voice agents. Raw data and statistical tests have not been published. Conversation length stands in for "difficult request", which ElevenLabs itself calls imperfect. The user study used a simulated scenario, not real customers. The direction is still plausible: when people have a tricky problem, they would rather explain it than type for a long time.
What this means for you
- For short, simple requests, chat remains the better channel according to the analysis.
- For complicated or stressful requests, voice agents may resolve more cases and do it faster.
- Response speed plays a big part in whether voice AI succeeds.
How to use this
Sort the last four weeks of customer requests into "short and simple" and "long or frustrating". Only the second group is worth testing with a voice agent.
Luigi's take
The numbers are interesting, but they come from the seller. I would read them as a hypothesis, not proof. The core message still matches what anyone who has run customer service knows: people want to type short questions quickly, and they want to talk when they have a real problem. Treat both the same and you lose customers.
My advice: sort your customer requests by type before choosing a channel. Opening hours, order status and invoice copies are fine for chat. Complaints, rebookings and anything involving frustration are worth a voice test, with a human easy to reach behind it. Then measure it yourself: how many cases actually get resolved?
Source
This article was researched and written with AI assistance. Responsible: Luigi Lugmayr.
