Yelp said on Tuesday that it’s working on deploying AI-powered “voice agents” to help service providers and restaurants handle calls, answer basic questions, and accomplish tasks like adding a customer to a restaurant waitlist.
Yelp says its agents don’t require a complicated setup or API integrations, and can draw from existing metadata along with data from businesses, such as pronunciation guides, customized voice greetings, and call forwarding rules. In the case of a restaurant, for example, Yelp’s agents can connect to the restaurant’s management software to send reservation details to a customer post-call.
Yelp’s agents can also handle things like automatic spam filtering and call analytics. They’ll hand the conversation off to a human for more complex requests, and after each call, businesses get a summary of the call along with a transcript and the recording.
“Often, pros are working in difficult conditions and can’t take a call,” Yelp Chief Product Officer Craig Saldanha told TechCrunch in an interview. “We want[ed] to build a product to help convert the leads that [normally might be] missed.”
Yelp is using OpenAI’s Realtime API to handle end-to-end calling. This allows Yelp’s agents, augmented by the company’s knowledge graph, to ask and respond to follow-up questions.
Yelp said that it is constantly evaluating new models for its products to get the best results in terms of latency, speech recognition accuracy, and overall customer experience.
Saldanha thinks that voice tech will be commoditized over time, and that the point of differentiation will be the underlying data and the way AI handles customers’ queries. He added that, in those areas, he thinks Yelp has an edge over the competition.
