Seen as India’s largest conversational AI platform, Haptik entered the market in 2013 when the term chatbot had not become commonplace yet. The startup’s unique model utilised the expertise of human agents in training chatbots to adequately handle various queries by users.
Haptik recently developed a chatbot for the Maharashtra government which is available on the
Aaple Sarkar RTS (Right to Services) platform. The bot provides easy, conversational access to information regarding 1,400 public services of the state government. CNBCTV18.com caught up with Aakrit Vaish, Co-founder and CEO of Haptik to understand the nuances of chatbots and the challenges they face in a digital era. Could you tell us about Haptik and how it started?
Haptik is one of the world’s largest Conversational AI companies, having reached over 100 million devices, and processed over 2 billion conversations till date. Our leading clients and enterprise partners include Samsung, Oyo Rooms, KFC, Coca-Cola, Tata Group and Club Mahindra, among others. Haptik is a part of the Reliance Group of companies, which acquired a majority stake in the company in a $100 million deal in April 2019. Swapan Rajdev and I founded the company in Mumbai in late 2013. The idea was that conversational interfaces would do a lot more than just peer-to-peer communication, and chat/voice bots would be used to complete tasks.
What were the initial challenges you encountered when entering the enterprise chatbot space? What was the evangelising process like?
When we started out, chatbots and AI were still words that evoked a sense of enigma than familiarity, as is the trend now. Particularly for the Indian market, incorporating chatbots in their day-to-day functioning was not something enterprises were prepared for. It took us considerable time to make them aware of the significant benefits AI assistants bring to the table including process optimisation, cost reduction and revenue enhancement. Our first product was the Haptik App, an Android and iOS compatible chat-based personal assistant which helps users manage their schedule and help them manage their tasks systematically. We used that product to showcase to enterprises the benefits of conversational AI, and how seamlessly users can get things done. Also, the initial fund raising process was tough as chatbots were a completely novel concept without any business precedent, hence making investors anxious.
What were the obstacles that came on the way of Haptik while working across sectors in the enterprise space? How did the company overcome it?
Every sector has its very own set of unique needs and demands that a bot needs to solve and keep in mind. Therefore, the initial challenge was to effectively customise our fundamental framework to effectively meet an individual enterprise’s needs. This was achieved through constant deliberations with the client. In fact, one of the most powerful aspects of Haptik’s operations which distinguishes us is our focus on developing an immaculate understanding of what the client wants. By developing a clear understanding of the same, we were able to develop chatbots that were perfectly suited to offer what was expected of them.
Users feel a bit frustrated when a support request fails to resolve via a chatbot than with a human at the other end. What are the processes in place that can help in mapping the unique customer flows of each enterprise customer and how do you go about it?
Just like any new emerging technology, it’s important to identify its gaps along with its strengths. For chatbots, we realised at a very early stage that they are very good at handling ‘level 1 queries’ -- things that form the bulk of the questions (80 percent) but don’t require very complex solutions. The remaining 20 percent queries usually require greater expertise, while some users may require more empathy. We have a proprietary routing algorithm which routes the chat seamlessly to a human agent whenever the bot cannot interpret it. Whatever the case may be, our bots are constantly learning and even when the conversation is transferred, the bot tries to observe the conversation and analyse how the human agent decoded the query, what approach was followed and how the query was finally resolved. This process ensures every subsequent conversation with the bot is more intelligent and satisfactory than the previous one.
Where do you see conversational AI at the moment and what is its future? What are the areas open to innovation?
If you think about the typical emerging cycle for any new technology, it goes through the peak of hype and then suddenly a ‘trough of disillusionment’ where the hype is not fulfilled. It is during this phase that people truly start understanding what the specific technology may actually work well for. Conversational AI is at that stage -- I like to call it the ‘slope of enlightenment’. Now, in 2019, enterprises and the world, in general, are recognising the use cases that bots can solve really well. Customer support across automation, across messaging and telephony, is starting to take off and conversational commerce is picking up thanks to the proliferation of voice assistants. Throughout sectors such as travel and hospitality, banking, retail and healthcare, users are increasingly preferring communicating with bots. The areas that still require research and innovation are across natural language processing to understanding complex statements and mixed languages. Speech recognition, particularly in the presence of ambient background noise, is another area where lots of work is being done.
What are Haptik’s plans in terms of working with Indian languages within chatbots?
Building vernacular conversational AI through multilingual chatbots is the next frontier we are working towards. Language essentially is nothing but an algorithm with contexts being assigned to certain variables. Furthermore, India is arguably the most linguistically diverse country in the world and a greater portion of the new internet user emerging in the country is more comfortable in their own language rather than English. As per a 2017 KPMG-Google report, the number of Indian language Internet users will be 536 million by 2021; far out-stripping the projections of 199 million English language internet users in India. And we have a belief that these users will interact with the internet potentially voice first, making this an exciting future for technology that Haptik provides.
We have a few unique aspects that differentiate us from other competitors:
How does Haptik intend to stand out in the increasingly crowded enterprise chatbot market? We are the only conversational AI platform that is built on real consumer chat data - our consumer virtual assistant Haptik has processed more than 100 million conversations, which has enabled the entire data set needed to train our models. So, instead of using ‘proxy’ data, our platform has been built on real conversations which makes our AI superior to others. We only focus on specific use cases for consumer brands. Again, because of our expertise of having worked directly with consumers, we understand those nuances better than most others. We don’t touch other use cases or verticals which do not fall within our expertise domain. We are a full-stack platform that acts as a long term partner -- we have our own technology, implementation team, analytics and success team. We work with each customer using a unique long term multi-year approach to eventually drive value for them. If you work with Haptik, you don’t need to work with anyone else. Haptik has an AI trainer team that works on the bots – How does the training happen? And what kind of skills does a ‘trainer’ possess?
We have a dashboard where AI trainers go in and enter the data. It is a simple DIY tool that anybody with general problem-solving ability can use. Our goal is that anyone should be able to build a basic bot using our platform in minutes. The training process involves entering the common ‘intents’ that the particular bot will solve. An intent is nothing but a topic, say ‘order status’ in the case of an e-commerce support bot.
Similarly, data associated with these intents and utterances is entered to create a fully customised model for the specific bot.
Could you also give us an insight into the tie-up with the Government of Maharashtra to launch their Aaple Sarkar bot?
Haptik recently developed a chatbot for the Maharashtra government which is available on the
Aaple Sarkar RTS (Right to Services) platform. The bot provides easy, conversational access to information regarding 1,400 public services of the state government. It has received a great positive response from the state’s residents, who have substantially benefitted from its instantaneous, accurate, transparent and complete information and tracking services.
For instance, users can simply search for services like ‘driving licence’ to see all the pre-requisites for the application process, and can then track the status of their application. Accessing relevant information, collating documents, initiating the application process, monitoring its progress to eventual completion and feedback submission, the RTS portal chatbot handholds the user through every stage and adds a great degree of convenience for users to easily avail the service they need from the state government’s massive listing.
Disclosure: Network18, the parent company of CNBCTV18.com, is controlled by Independent Media Trust, of which Reliance Industries is the sole beneficiary.