The Silicon Review
Reach Your Customers at Scale with a Personal, Human Voice
Conversational AI is the holy grail of artificial intelligence. It has been a dream of computer scientists for over 50 years now. From movies to TV shows, conversational AI has dominated popular culture and has captivated the imagination of almost all demographics, making us wish we could actually speak WITH (not AT) machines. The recent development of deep neural networks (i.e. deep learning), allowed Voca.ai founders Einav Itamar and Dr. Alan Bekker to finally make that dream a reality, and in 2017, Voca was born. They set out on a journey with a mission to create a human-friendly virtual agent that by leveraging unique AI driven, human-like features, will have the ability to both increase revenue and dramatically improve customer satisfaction in call centers of various types and sizes. Two years out, they seem to have done it!
Discussing the Importance of Human at Scale in today’s Call and Contact Center with Einav Itamar, CEO & Co-Founder of Voca.ai.
Q. What made you explore and step into the call center industry?
Call centers around the world are going through massive changes recently. With over three million call center agents in the United States alone, there’s much room for innovation. On the one hand, people are always hungry for good service and they want to use their voice to get that service since voice is the fastest and most efficient way to communicate. On the other hand, companies keep trying to reduce operating expenses by driving people away from the voice channel to cheaper, less engaging alternative channels such as e-mail, chat and online forms. This frustrates the user and ultimately alienates him/her. So our message to the world is that you should stop compromising. Our AI platform was built from the ground up to mimic human behavior. So a customer that is speaking with a Voca agent is able to express him or herself in the most natural way possible, human conversation. Our AI listens to that customer, it feels its human intent and then speaks naturally.
Q. How are you different from other companies?
Other companies first take the speech from the user, then turn it into text and then they try to do the sentiment analysis or any other textual analysis on the text itself. This paradigm is very limited because a lot of the information is lost when you do speech to text first. When a human agent is speaking over the phone, the agent can tell when I’m more comfortable and when I’m less comfortable or when I’m more confident and when I’m less confident. But when you do these transcriptions to text you are losing all of this information and this information is very important when you are providing customer service. You need to be empathetic to the customer you need to feel the customer not just to understand the words.
So this is why we developed a very unique algorithm the “speech to intent” algorithm that basically uses the deep learning based neural network to identify the human intent directly from the customer’s speech. For example, if I am saying “I don’t know” (without emotion) is different from “aahhh, I don’t know” (with a lazy and skeptical emotional tone) are two completely different things. Identifying the pauses, emotions and the way a speaker delivers their speech becomes mandatory and that is a very important ability of our ‘speech to intent’ feature.
Another important feature of ‘speech to intent’ is the capability of improving the accuracy of the transcribed text because if you completely ignore the context it’s very hard to understand even the text itself. Considering the agent is asking a specific question, that question will be taken into account in the speech understanding. So the ‘speech to intent’ algorithm improves both the text understanding and the emotional understanding at the same time. This is very unique to us and it enables us to create the natural, real world experience that I’m talking about.
Q. What are the five reasons to use Voca?
Q. Can Voca.ai completely mimic a human?
One of Voca’s most appealing differentiators, is it’s the ability to generate human voice. During the past decade people have gotten used to speaking to machines (utilizing ‘text to speech’ functionality) but the end result, the machine’s voice sounds very robotic. Again, it’s a question of the intonation of the pauses, the fillers like the “ums” and “ahms” and even the laughter. A voice can sound very natural but at the same time it can also be very monotonous and you don’t want that. So at Voca, we have developed a technology that allows customers to adjust the prosody and information of the pitch to drive better business results. So for instance, when a customer is on the call and it is a sales call, the agent is trying to convince the customer to buy, renew or expand their usage of a specific insurance policy. In these types of cases, the information ‘visible’ to the agent is paramount as that information directly affects what the agent is saying. Our proprietary technology allows limitless personalization where you can choose between many different voices that are available. You can also design a dedicated voice or a dedicated set of voices that will represent your brand.
At the end of the day, we are capable of taking information from our customers and scientifically benchmark it. Eventually creating a custom tailored sales agent that is completely artificial and can be easily multiplied and replicated to form a sustainable, instantly scalable sales machine.
Q. What is the future roadmap of the company?
Voca’s mission is to deliver more success in different call center verticals. Our goal is to empower the human agents in the call center and not to replace them. So we want to make sure that humans and AI are working together to build the perfect experience for the end user. So this is our mission for the upcoming years - our goal is to eventually build the perfect experience that is both scalable and human.
Einav Itamar, CEO & Co-founder, and Dr. Alan Bekker, CTO & Co-founder
During the last 15 years, Einav has led several AI and Big Data tech startups. Before he co-founded Voca.ai, he was leading the adoption of deep learning technologies at eBay. Prior to that, Einav led the R&D of two AI startups, one of them, Corrigon, was acquired by eBay. Corrigon developed a deep-learning based technology for image recognition. Artificial intelligence was also the subject of his master research at the Technion University, one of the leading technology universities in the world. Following Corrigon’s acquisition, Einav met Dr. Alan Bekker – a leading deep learning and AI researcher that had the same passion to Conversational AI like himself. Together they co-founded Voca. Alan is an author of 10 papers in leading AI journals and conferences. Prior to starting his PhD, Alan has been working as a machine learning researcher in Intel. Recently (2019) he was named as one of Forbes 30 under 30.