When a problem is complex or stressful, our research suggests that people want to talk it through, not type it out. To test that, we compared voice calls (over the phone and WhatsApp) with chat sessions across hundreds of production agents running both, and ran a user research study for an additional point of comparison.
We found that voice was best for complex issues. In conversations longer than two minutes, voice succeeded 32% more often than text-based chat. When those longer conversations did succeed, voice resolved them about one third quicker.
People also preferred voice when things got stressful. In a simulated scenario with a canceled flight, 85% of participants chose it.
That doesn’t make chat obsolete. In short conversations, chat has a slight edge. But the longer a conversation runs, the more voice helps.
Voice stands out for complex support issues
We used conversation length as a stand-in for complexity, and saw a pattern emerge. In conversations under a minute, chat did better by about 9 percentage points. Between one and two minutes, the two channels were fairly even. Past two minutes, voice pulled ahead: success rates were 10.9 points higher — a 31.7% increase — and nearly two-thirds of agents did meaningfully better on voice.
Voice also resolves longer conversations faster. Among successful conversations over two minutes, voice averaged 4.6 minutes compared to 7 minutes for chat, and 78% of the 90 agents with enough successful long conversations on both channels resolved them faster on voice.
Duration is an imperfect measure of complexity, especially for chat, where customers may pause between messages. But the pattern is clear: the longer the conversation, the more voice helps.
Part of the reason may be that voice lets customers cover more ground. People speak about three times as fast as they type, and a spoken conversation keeps both sides engaged instead of waiting between messages. When a conversation gets complex, voice is the faster path to resolution.
The winning formula in voice: speed plus empathy
Not every voice conversation succeeds, and one of the biggest reasons why is latency. Across 315 agent conversations that switched to a faster LLM for lower latency, a 16% reduction in average end-to-end latency correlated with a 26% lift in success rate. At the same time, the share of calls ending within 15 seconds fell from 21.1% to 3.8%.
That suggests latency matters most at the start of a call, when people decide whether to stay on the line. If an agent is slow to respond, callers are likely to assume the line is dead or the system is broken, and hang up. Faster responses keep them on the line long enough to get help.
Speed is only part of it. When a conversation is stressful, how an agent sounds matters too. In a user study we ran, 98 participants worked through a simulated canceled-flight scenario, similar to the one inthis video, rebooking with both a voice agent and a chat agent in randomized order. In a survey after completing their scenarios, 85% responded that they preferred voice, naming speed of resolution and empathy among their top reasons.
Why getting voice right matters
Customer loyalty is won or lost when the stakes are highest: the canceled flight, the disputed charge, the problem that has already taken up too much of a customer’s time. Whatever the channel, failing to resolve the problem costs you the next conversation.
For the hardest conversations, voice gives you the best chance at a resolution, if it’s built well. That starts with an agent that can finish the job. Long conversations require an agent that keeps track of context and connects to your systems, so it can rebook the flight or issue the refund itself, instead of handing the customer off. Low latency matters just as much, because it keeps the conversation natural and callers on the line. And the voice itself should sound calm and empathetic, especially when the customer isn’t.
That’s what we’re building with ElevenAgents. It’s one agent, with the same knowledge and integrations across voice and chat, so customers can resolve their issue in whichever channel they choose. But the longer a conversation runs, the more voice helps, and the faster it gets there. The companies that resolve the hardest conversations are the ones customers come back to.
Our methodology
We looked at agents that handled at least 50 voice and 50 chat conversations between May and July 2026. Each was judged against success criteria the business defined for its own use case.
To keep the comparison fair, we compared each agent’s voice and chat performance against itself, rather than combining all conversations. Businesses differ in use case, volume, and how they define success. If we combined every conversation, a business that mostly uses voice and happens to have high success rates would make voice look better, even if voice made no difference for that business. Comparing each agent against itself removes that problem.
