VAX to VaaS-Journey in improving Customer Adoption of Voice Assistants

When we started Slang, we started with the primary motivation of democratizing the internet access for our moms (and dads). Our moms and dads representing the huge inflow of digital newbies (popularly knowns as the NBU or the Next Billion Users). Enabling them to use the apps that we (the blog readers) find so easy to use. And along the way, making it easier even for us “experts”. And we wanted to do it by enabling Voice as an interface layer on top of apps. Just as “touch” made the way we interacted with apps easier, we wanted “Voice” to be that new-age interface, enabling users to just talk to their apps and for the app to talk back to them. We called this VAX — Voice Augmented eXperience

VAX Platform

The way we did this was to first build out a horizontal platform that enabled developers of the apps to add a custom Voice layer on top of their existing apps.

‍

‍

Challenges in this approach

There were two primary components that made up the visible part of the platform. The Slang Console and the companion Slang SDK. The third component, Slang Cloud, is not exposed to developers directly and the SDK is the one that transparently talks to our backend, where a bunch of the magic happens.

‍

The Slang console is a web-based tool, that enables developers to configure their own custom use-cases. They can create their own intents and custom entity types, train intents with utterances, and also mark out entities that needed to be collected for that intent. It was a very visual tool and was built keeping developer flexibility and ease of use in mind. Users could also configure prompts and priority orders of entities to be auto collected.

‍

The accompanying SDK, called the VAX SDK, was the workhorse that interacted with the app and also the backend and together provided the VAX experience to the end-user.

While this was a very flexible model, it also had the following issues —

• Android and Web Developers had to learn basic NLP concepts like intents and entities
• They had to configure, train and maintain their own models
• Mapping the NLP concepts into the domain-specific use-cases that integrated into existing app flows, was a bigger challenge than expected
• Different apps would end up building similar use-cases in completely different ways causing non-standard end-user experiences, even for apps in the same domain

All this was causing unpredictable timelines and sometimes poor experiences for companies integrating the VAX experience into their apps.

‍

Enter VaaS — Voice Assistant as a Service

What if we could come up with a much simpler way for apps to integrate awesome VAX experiences but without them needing to understand any NLP concepts? Or need to construct their own complex conversation design? And enable them to do this in hours rather than weeks?

‍

This lead us to build the world's first VaaS platform.

‍

We now offer domain-optimized, out-of-the-box Voice Assistants that can be easily integrated into apps. The Voice Assistants are multilingual by default and highly accurate for the domain it's optimized for and provide the best VAX experience out of the box.

We now offer, out-of-the-box Voice Assistants for the following domains -

• eCommerce
• Travel
• … and lot more coming over time

All these Voice Assistants come with the following features —

• Highly accurate — We have trained these Assistants to be highly accurate for the domain they are serving and it can be easily augmented with any additional data that our customers can provide
• Constantly learning — Our models keep getting better over time as we manage them, without you needing to maintain it
• Multilingual — We make sure that these Assistants are multilingual by default, with us adding more language support over time.
• Great Assistant Experience — We make sure that these Voice Assistants are end-user friendly and come with all the features that typically make a great experience — The Assistant will talk to the user, help them get onboarded, patiently listen to their inputs (no more random timeouts), will respond back with questions where required and given them good confirmation messages.
• Analytics — The proof of the pudding is finally in the eating. We know you would like to understand how your users are using the Voice Assistants. So we provide them for you out of the box, so that you can keep tabs on what’s happening with the Assistants deployed with your apps.

Do you have apps in any of the domains that we already support? Or looking for a Voice Assistant in your own domain? Give us a shout here