Voysis raises $8 million to help it become the Twilio of voice AI

Voice-powered artificial intelligence is not something that’s easy to set up for just any business, even if it might have real benefits in terms of driving sales or improving customer experience. Voysis is a startup that wants to change that, with an AI platform that can parse natural language input, and that works effectively in specific domains including e-commerce, entertainment and more. Voysis doesn’t want to be Siri or Alexa – it wants to the perfect shop assistant, or the extremely insightful video store clerk.

“Voysis is a complete voice AI platform,” explained Voysis founder and CEO Peter Cahill in an interview. “What I mean by that is that the platform enables companies and businesses to rapidly stand up their own artificial intelligences that can be queried by voice or text.”

In other words, the company is focusing on building a platform for a few specific, high-value domains with a deep working knowledge and strong reliability, instead of a more shallow, but more general voice-powered assistant. The idea is to create something that works well enough in a given domain that it can be made available via APIs, letting business owners plug their operation into the platform with minimal fuss and set up time, so they can get started delivering voice-based AI interaction to their customers right away.

Taking e-commerce as an example, which makes sense because it’s one of Voysis’ primary targets as an industry, a retailer might like the idea of voice-based search, letting shoppers say clearly and with natural language what it is they’re looking for. Typically, building this kind of tool would involve a lot of time training the system and classifying inventory, and basically either assembling a series of not quite fit-for-purpose tools to try to tackle the problem or building a bespoke (and very expensive tool), which also involves recruiting talent that’s currently scarce and comes at a high premium.

“If we’re working with our first company or two within a vertical, we tend to be fairly hands-on, because we’re learning what their requirements are,”  explained Cahill, regarding how Voysis actually operates for customers. “But our focus is always really on making it available where people just use APIs, and just push their data to our servers and then have a voice assistant that’s relevant to them.”

In a live demonstration with a sample furniture store client, Voysis did indeed work well, responding to natural language input quickly and accurately. It even tracked context, letting you refine your search with inquiries like “Now show me only ones that are under $400” to get more specific, for example. And the information doesn’t need to be specially prepared or tagged for this to work, Cahill explained; their software reads and draws conclusions from existing material, including copy written to describe the item and its characteristics on customer-facing product pages.

Other companies, including IBM with Watson, have offerings that aim to deliver domain-specific AI. But Cahill, who holds a Ph.D, has been working on neural network and voice recognition technology in an academic capacity for the past 15 years. His focus, and the combined expertise of his 15-person team, which is set to double in 2017 thanks in part to this new funding and a new office opening in Boston, are the company’s competitive edge. The low friction for getting started is also key to the Voysis approach, which again differs a lot from the ample manual training required for most AI solutions.

Voysis’ Series A round of $8 million was led by Polaris Partners, which also provided Noel Ruane as a new executive at the company focusing on building out the business side of the young company.