Deepak Gupta & K.V. Ravisekhar

Ever since Facebook expanded access to its Messenger in April’16 by giving businesses the ability to reach customers through APIs, “Chatbots” have been the buzz word in  developer communities across the globe.  Here is a short piece of advice on the essentials for investing resources behind chatbots for startups as well as established businesses.


What’s hot about it?

There is a certain wow when someone books a cab or orders a set of flowers through a conversation. It has a never-seen-before scenario of applying natural human language into computing intelligence and delivering simplified consumer experiences. Chatbots bring in the user advantage of accessing business services while on chat platforms like Facebook Messenger and Telegram. What this could mean is that search and discovery may shift from the Google browser and the dependency on Apps to access business services may also move away to chat platforms because of their pure simplicity and convenience.  This development truly has the potential to become the next big phenomenon when businesses invest sufficiently behind this and when more chat platforms like Whatsapp join the party in near future and hence the hype around the announcements of Facebook and Microsoft in the last few weeks is justified.

In India, we have already seen a few bots like Niki and MagicX trying to make their presence felt in travel and groceries categories in India.  There certainly are more developments expected out of them and a few more players like Haptik (funded by Times Internet) trying to solve more use cases assisting users on chat platforms through human supported bots.  One can posit 2 sets of chatbots targeting chat platforms emerging in the next few months: (1) stand alone bots that solve for conversations and (2) bots that are platform extensions from businesses like Flipkart and Amazon. The second scenario notably creates opportunities for the developers to offer bot as a service / customized product to businesses that cannot afford to develop it in-house especially in the functions like customer support that involve template based human conversations.

Why should a business invest in a bot?

In the west, human supported bots started surfacing in the customer support domain to reduce the dependency on man power and thereby the costs in the past decade or so by companies like Interactions successfully. The adoption of such technologies in conversations to reduce CS costs is imperative in the Indian market as well in the coming years.

As key chat apps are opening up for businesses, they now need to prepare to engage their customers on chat platforms across use cases – (1) service and product discovery (2) ordering and (3) customer support. As the supply and demand volumes go up, though it is very early to predict, one may soon come across ecommerce companies talking about the growing pie of the conversational commerce and content based online businesses about their content consumption on chat platforms. There are teams like Nanorep which are already in the market offering bot as a service to ecommerce firms.

The importance of AI

Right from Apple’s Siri days to the Cortana and Skype announcements of Microsoft in March’16, the ability to collect data and information from the assigned services and have a conversation has always been the talking point and it often sounds like conversations (chatbots) are something that standalone NLP Technology can deal but in reality it is the domain and expertise that the bot is used for. The AI, therefore, is fundamentally about your business domain, expertise and your consumer use cases which is not very different from what your App and Website are about and just that additionally the bot can interact with your consumers and have conversations playing the role of a sales rep, domain expert or a customer support agent without any human support.

What problem is your startup solving?

While there is a rush in startups in this space to take a head-start, it is important to solve specific consumer problems and bots banking purely on the conversation format, contextual understanding, NLU excitement and which are smart at calling other services would eventually have competition from global giants like Viv, Microsoft, Google and Facebook. In addition to that, large online businesses could soon follow Amazon which is estimated to have sold 3MN units of its Ecko so far making a strong case of expanding into conversational commerce through Artificial Intelligence (though not  specifically on chat platforms).

Building of a chatbot

While the actual Natural Language Processing (NLP) use cases were regarded complex demanding advanced Data Sciences capabilities, with players like the Facebook acquired trying to come up with APIs and offer simplified solutions to set up conversations, the key piece of the problem looks simpler than ever before. For one to interpret consumer queries, convert them to specific topics, run them through the conversation and knowledge engines to shoot back replies the skill sets needed are far simpler now and the technology needed is freely accessible. The need to feed up the knowledge engine with domain intelligence is where the business needs to focus on if the bot is an additional platform to their core along with web and apps. However for startups building their models purely on chat platforms, there could possibly be some merit in having their own libraries in components like knowledge engine as they scale in order to reduce the dependency on 3rd party tools.

Chatbot startups that are not extensions of websites and apps can essentially adopt any of the 3 approaches for training the bot:

(1) I shall build a smart bot that would learn from its user interactions – Consumers often expect your bot to be near-perfect in their first interaction and hence it is risky to assume that your bot can afford to learn from its user interactions. Its needs to be fully prepared before interacting with its users

(2) I shall have humans supporting my bot till it get trained sufficiently – Large players who are dealing with open ended use cases including Facebook’s M are reportedly taking this approach. While this brings in control, it can demand heavy investments in the initial years and also runs the risk of ending up as a semi-human application.

(3) I shall train my bot sufficiently with content and structured data – While eventful applications in AI like IBM Watson and Google’s AlphaGo were largely trained with a mix of structured and unstructured content, massive data of human actions and algorithms built on top of it over the years, it is very expensive and time consuming. Although this is arguably the best approach for building complex and scalable bots that are capable of dealing with even with qualitative and judgement based use cases, it has to be seen how the funding ecosystem in India would embrace this idea with patience and risk appetite.

With the biggies investing so heavily in conversational canvas and increasing traffic and engagement on chat platforms, it is almost certain that businesses would compete fiercely through chatbots sooner than later as they do on web and mobile platforms.  It would also be the time when the Indian startup ecosystem would build capabilities to grab a pie be it in (1) search and discovery (2) expertise, assistance and conversation dependent business cases and (3) reducing template based human interactions enabling businesses.


Deepak is CEO & Co-Founder of Equity Crest, a curated seed-funding platform and also an angel investor.

Ravi is an entrepreneur and startup advisor and previously led marketing tech,  digital marketing and CRM at ebay India.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s