For a couple of years now, there’s been a lot of buzz around chatbots. Earlier, the go-to term used to be conversational commerce, although today most companies you talk to will happily talk about artificial intelligence and machine learning as the power behind these chatbots. For companies that are looking to integrate the technology today, the likes of Google and Microsoft offer solutions that can be quickly activated. But there are also plenty of independent companies that are offering chat platforms that can power solutions for your business.
One such company is Senseforth, which is a «humanlike conversation platform built on NLP, AI, and machine learning,» according to Sridhar Marri, its CEO. Senseforth’s chatbots are used by companies in banking, healthcare, e-commerce, and more. Some clients include HDFC and ICICI, Club Mahindra, and Manipal Hospitals. We asked Marri about the need for companies like his, at a time when the big Internet companies are also releasing similar tools, which anyone can implement, often quite cheaply.
«What happens is that Google, Microsoft, Amazon, they can’t do as much for enterprises, they’re more consumer-centric companies,» Marri opined. «We work with enterprises, who don’t have the energy to hire for the skills required to build something like this in-house.»
Beyond that, he holds that Senseforth can also help companies because it’s able to customise its offerings to their specific needs. «If you’re an enterprise, you can use someone like us, or there are companies that use those [Google etc.] platforms,» said Marri. «We are using a proprietary algorithm, so we can be much more customised than people using the big platforms.»
Other Indian platforms offering these kinds of services are also continuing to grow. Most recently, Twitter co-founder Biz Stone invested in an Indian health chatbot called Visit. It is an AI-powered chatbot for health advice, learning from and assisting real doctors. Meanwhile, another Indian company, Haptik has been creating chatbots for a number of purposes, while it also has it’s own app that people can use to call for multiple services via chat.
One of the recent bots that it built was to help fight cyber harassment. Swapan Rajdev, co-founder and CTO of Haptik, believes that there’s a growing use case for these kinds of bots. «There are a lot of basic things that can be done really well with a GUI, and you should not try and replace those with a chatbot,» he said. «But at other times, you want to be more conversational, to have someone respond with the correct tone, and that’s what a chatbot can deliver.»
However, the accuracy and reliability of these bots can be a stumbling block. For example, Ixigo has a chatbot called Tara that can recommend travel plans and help you book tickets. But it’s not a complete solution yet, as Ixigo co-founder and CTO Rajnish Kumar says that the bot can be used to automate customer support to 70-80 percent.
That’s something Senseforth’s Marri agreed with, adding that the accuracy in understanding not just the words, but also the intent could be vary between 50 to 80 percent, depending upon the natural language processing (NLP) engine being used. He added that for Senseforth the number is around 85 percent.
«If you look at it, three very similar sentences — ‘How do I apply for a credit card?’, ‘Can I get a credit card?’, and ‘I want a credit card.’ — are all different,» explained Marri. «And parsing this perfectly is easier said than done. Even at over 80 percent accuracy, if you’re dealing with a financial transaction, like say opening an FD, you need to be at 100 percent. But we can use follow up questions and intelligent design to further cut down on the error rate.»
This kind of involvement, where companies like Senseforth and Haptik work closely with enterprises to help customise the bots, and work on design to reduce errors are the main reason why they still have an important role in the market, despite the big companies coming in with their own platforms, and this isn’t something that’s likely to change soon, Marri added.