Voca.ai

The founders and the problem

Voca.ai was founded in 2017 by Einav Itamar (CEO), an entrepreneur who was among the founders of Corrigon, sold to eBay in 2016, and Dr. Alan Bekker (CTO). The company was headquartered in Herzliya, with an office in New York. Its target was phone-based call centers: according to the company's website at the time, seven out of ten customers still prefer to talk to a human agent, so it set out to build a computer system that sounds like such an agent.

The voice agent: from speech to intent

According to Calcalist, most systems in the field converted speech to text and only then analyzed it, while Voca developed an algorithm that identifies the speaker's intent directly from the speech. According to Globes, the company's agents were trained on earlier conversations between agents and customers, and paid attention not only to the words but also to pitch, tone and speaking pace. To make the conversation sound human, the system added pauses and filler words such as "umm." The agent handled simple requests and passed complex ones to a human agent, and was used, among other things, for debt collection, surveys, scheduling appointments and customer service.

Customers and funding

By the time of the acquisition, Voca.ai had raised about $6 million, including from American Express Ventures, which invested in it in October 2019, and from lool ventures and Flint Capital. The company took part in accelerator programs run by Citi and Mastercard, and won the Best of Show award at the FinovateSpring conference in 2019. According to TechCrunch, in October 2019 the company reported about 2 million calls a month in banking, telecommunications, insurance and law, and its customers included Toshiba and Amdocs.

The Snap acquisition, and two prices

In November 2020 Calcalist reported that Snap, the operator of Snapchat, was acquiring Voca.ai for about $70 million, citing a source who asked to remain anonymous. TechCrunch confirmed the deal through its own sources and reported $120 million, $70 million in cash and $50 million in stock. In an update to the article, it added that another source insists the total price, with all its components, is only $70 million. Snap declined to comment, and the amount was never officially published. About 30 to 40 of the company's employees (depending on the report) moved to Snap, and according to Globes they were to set up its development center in Israel. Globes linked the deal to the voice commands Snapchat launched in June 2020.

After the acquisition

In September 2022, as part of layoffs of about a fifth of Snap's employees worldwide, Calcalist reported that about 30 employees in Israel had been laid off, most of them from the development center in Herzliya, the center built on the Voca.ai acquisition, and that there was a plan to close it. According to the report, Snap's other centers in Israel, in Tel Aviv and Kfar Yehoshua, were only slightly affected.

What the story teaches

Voca.ai worked in a field that grew enormously a few years after it was sold: voice agents that handle entire service calls became, in the era of large language models, an active and competitive field. Its story shows that the attempt to build a voice agent that sounds human came before the generative wave, and that many of the ideas, such as intent detection, tone of voice and handing off to a human agent, were already on the table. It also shows the other side of a strategic acquisition: the acquired team was among the first to be hit when the buyer cut back.