Sendbird launches voice AI agents to bring human-grade conversation to customer calls

Artificial intelligence communications and conversation platform Sendbird Inc. today announced the launch of Sendbird Voice AI, a voice automation solution for customer service designed to take the anarchy of real conversations and handle them with care.

Customer service is the one place consumers expect a high level of attention across all their devices, especially phone calls.

“We focused on building for voice chaos, which means that when you are texting someone, it’s unlikely that you will hear somebody else’s texts,” Shailesh Nalawadi, vice president of product at Sendbird, told SiliconANGLE in an interview.

Human agents are very good at picking out voices; they won’t get confused if they hear someone else talking in the background, but a voice agent might get interrupted or taken off track. Nalawadi said Sendbird’s voice agents have been tuned to avoid this sort of problem in real-world situations where conversation is messy.

“We spent a whole lot of time on accents, emotions, interruptions and background noise, mumbling and kind of everything that makes phone conversations real and human,” Nalawadi added.

Customers don’t have the time for a slow or creaky experience with a voice AI agent. They want to get on with their lives – or maybe talk to another human being if the machine isn’t working for them. What can worsen is their concern about getting disconnected and having to explain their problem to a bot all over again.

Sendbird fixes this by embedding the entire context of a customer’s experience into their system across every channel. So that if a customer starts their experience on text, web, or even a phone call, they can pick right back up from where they left off.

For instance, if a customer is speaking with a voice AI agent to resolve an issue regarding an incorrect address. Imagine that they had something shipped and it was in packaging and wasn’t sent yet but they needed that address changed. They had just gotten through explaining the problem when they were disconnected.

With Sendbird, all of that explanation is summarized already by the agent and it can pick up where they left off. “What we’ve always had is this concept of for memory, so that when you reconnect with an AI agent, you are then able to pick up the conversation where you left off,” Nalawadi said.

The agents can operate globally using text-to-speech and speech recognition in 29 languages. Call centers and customers can also customize them for their specific brand and tone, as well as adjust their speech patterns and accent to fit clientele.

In addition, Nalawadi said, every agent is tested against realistic scenarios to ensure it understands intent, adapts on the fly and resolves issues as expected. If users are experiencing problems that an agent cannot handle, or are emotional and frustrated, for example, they’re becoming upset or belligerent, the agent can quickly escalate to a real person to handle the call.

Phone conversations aren’t the only place Sendbird expects to meet users on voice. The platform will expand in the future to voice over IP as well. For example, someone might make a call over WhatsApp audio, so the company is working on supporting that capability for markets outside the U.S.

“In the United States, we kind of presume that it’s going to be a phone call, but in many markets, actually, the experience starts with a visit to a website or an app,” Nalawadi said. “Globally, voice remains important; it’s just distributed in different ways.”

Image: SiliconANGLE/Microsoft Designer