Ravi N. Raj is the CEO and Co-founder of Passage.AI – He leads a team of AI and deep learning engineers who have created the industry-leading AI platform that businesses can use to create a conversational interface for their websites without a single line of code. Prior to co-founding Passage.AI, he led product for @WalmartLabs, Walmart’s hub for innovation around social, mobile and retail. Raj came to Walmart through the acquisition of Kosmix, where he served as the VP and GM of Kosmix’s sites, which included Kosmix.com, Tweetbeat.com and RightHealth.com, then the second largest health site on the Web. He has an undergraduate degree from the Indian Institute of Technology (IIT), Madras, and a graduate degree in Computer Engineering from the University of California, Irvine.
innomaniacs: Passage AI is the fastest growing AI chatbot platform coming from Silicon Valley. What pushes your growth?
Ravi Raj: Chatbots are widely talked about in the media and among companies. It all started with customer service chatbots, but now we see an increasingly growing number of external and internal use cases, e.g. in recruiting, e-commerce, IT and HR Helpdesks. Also, voice is further driving use cases that had never before been discussed, such voice-driven market research, user manual voice automation, hardware steering.
What are you most enthusiastic about when it comes to the capabilities of your platform?
We are using the latest deep learning architectures, but our clients really appreciate what we are able to achieve on top of that: our history in search and recommendation applications puts us in a position where we are able to deeply embed our system, in an automated way, into our customers’ organizations’ data landscapes. With a distinct combination of vector space and text space search approaches, we are able to generate very accurate information retrieval, which means users will get the right answers.
In the industrial application of chatbots, overall integration into the corporate processes and data landscapes are crucial. What are the key challenges today?
It is maybe comparable to the automobile: Most people think that a powerful engine is all it takes to have a lot of power on the road. And they think essentially the same thing holds true with chatbots, that a sophisticated NLU engine with some APIs addresses the problem.
But it is more complicated since we do not only need to extract the right intent but we must also provide the best possible answer. And the wider the use cases and the wider the search space, the more automated the approach of information retrieval needs to be handled.
We need to have an all-terrain vehicle as a starting point that is then adapted and scaled very easily to specific use cases and domains, while maintaining both high accuracy for the users and low complexity for the companies.
The more global our customers, the more they see the clear value of our solution that is easily scalable into other languages.
Only if the engine is very well connected and aligned to the steering and the chassis, then the car will drive efficiently, safely, smoothly – and without errors.
The more global our customers, the more they see the clear value of our solution that is easily scalable into other languages. Imagine our customers like Coursera and Udacity. They know AI and they could certainly build chatbots themselves. But the key to their success is instant scalability across a wide variety of languages. We have seen a 41% increase conversion rate through our chatbot on their English language website. The faster the chatbot is available in other languages, the faster you will increase the conversion rate in these languages.
What approaches does Passage AI take to solve these challenges?
We started our company with the idea that having to code a chatbot should be unnecessary. We are now seeing more positive effects than we could have anticipated. Not only is it very easy to set up fully integrated chatbots; it’s also very easy to scale them to new use cases or other languages by copying the structure and translating the content.
But due to the fact that there is no longer any coding required in the creation of chatbots, we are eliminated a potential source for errors that others might have when programming or putting together a chatbot from scratch. With our solution, we are certain that errors are related not to code but to content or information retrieval mechanisms. And that is quite easy to fix.
We also have a steep learning curve for creating a wide variety of different internal and external use cases for our 20+ customers around the world.
In the most common languages, we have also dedicated roll-out teams with native language skills. We are able to create text and voice chatbots in four to eight weeks, depending on the complexity of the respective use cases.
And we are innovative in enabling new use cases, so all our customers are benefiting from them. We have recently created a new approach and filed a patent, because we are seeing a strong increase of starting accuracies in wide text space information retrieval – something our customers should benefit from.
What is your personal goal for the next two years regarding innovation?
Chatbots can become a real digital companion for the users and employees of our customers.
We are seeing a broad range of voice-based use cases that have not yet been addressed. And we want to be the ones who end up seeing them realized in a way that is as accurate, cost sensitive and fast as possible for our customers.