Artificial Intelligence (AI), Natural Language Processing (NLP), Deep Learning, Machine Learning -- we’ve been overwhelmed by these terms over the past couple of years. Almost everyone in the ecosystem is announcing an AI project in some form or the other.
Machine Learning has been around for a long time. In fact, e-commerce as an industry has used it pretty successfully to deliver personalised recommendations to consumers. Financial businesses have been implementing complex algorithms to predict and automate financial models to support investments for clients.
With tech giants like Microsoft, Google, IBM and Facebook announcing platforms for everyone to leverage and build bots, we’ve seen a plethora of basic bots roll out, and roll off the radar at the same pace, leaving major doubts about their utility and scalability. Nevertheless, globally there are some serious players who have been making strides in the area of NLP AI. One such entity is Pune-based Light Information Systems Pvt. Ltd (LISPL).
Their journey began in early 2012, when three friends from engineering college met after a decade. They put together a team of 15 handpicked engineers with the aim to create a system that could process and understand text inputs semantically and respond to any user query across any topic in a natural language conversation. Fast forward to five years later, and the team is now 50 strong, still focused heavily on R&D, with their own Recurrent Neural Networks, 40+ proprietary algorithms and a product called NLPBOTS that can be highly customised to automate conversations across many enterprise functions. Today, Light is one of the only companies in India to have homegrown technology in NLP AI.
Starting with experimenting on Stanford NLP and other open source components that were available in 2012, the team quickly realised the limitations of the technology and decided to go indigenous.
And surprisingly for them, that decision paid off. As Sanjeev Menon, Co-founder & CEO who heads the technology function at Light, says, “When we were invited to Silicon Valley in 2013 to present our technology at the Pioneer Summit, we thought we would get to learn about the latest developments in the space. We were pleasantly surprised when the other startups at the event came up to us to know more about how we achieved many of the capabilities we demonstrated at the time. That’s when we realised that we were ahead in the game.”
Today, everything that powers NLPBOTS is built in-house, making it highly customisable to suit specific business use cases. The technology roadmap is closely linked with real world needs and hence the focus on user experience and related business goals is evident in the product.
NLPBOTS is the only solution in India (and among the very few globally) that offers realistic contextual conversations by virtue of Long Short Term Memory Models (LSTMs) with bi-directional Recurrent Neural Networks (RNNs). Simply put, if an employee asks the system a query: ‘how many sick leaves do I have?’, and then follows the bot’s response with ‘what about casual?’, the system understands the context of the second question, and answers with the number of balance casual leaves the employee has. Secondly, based on the user’s profile information and history of interactions, the system can engage in highly personalised conversations when servicing the user’s needs.
“Wouldn’t it be just awesome if the person taking your order knew just how you liked your pizza or coffee or your table setup, based on your past interactions or visits? Also, imagine the experiences one could deliver with well-informed proactive recommendations and offers that align with your customers’ specific preferences,” says Sanjeev Nair, Co-founder & COO, who strongly believes that it is the user experience that enterprises deliver with every interaction using the technology, that will define the overall adoption and future of AI NLP applications everywhere.
NLPBOTS is also one of the few systems that can be easily configured to handle handovers of the chat from the system/bot to a human agent. Human agents can intervene directly in an ongoing conversation with the relevant historical reference to better help the end user. Alternately, the system has also been setup to automatically escalate a user query to a ticketing system or a CRM tool like SalesForce and respond to the user with the relevant data for any particular open ticket. Add a UX element to this process, and you get a pre-emptive experience where the system can engage the user with a welcome message like this: “ Hi John, Hope you are doing well. Welcome to ACME Corp, I see that you have one open issue pertaining to ‘Reimbursement of travel expense’. I can see that it has been marked for second approval and should be updated later today. Is there anything else I can help you with?"
Every user’s interaction with the system can be analysed to drive the learning of the system. For example, in the case of a system that is integrated with a set of HR policies, the users can rate every answer the system gives, thereby allowing the system to improve from that feedback, in addition to the data qualification done by the admin. Every piece of information that is handled by the system can be tracked in real-time, thereby allowing enterprises to get a sense of the overall feedback or sentiment amongst the users. One can create quick surveys across different use groups to get focused polls, in addition to the automatic analytics generated, making it even more powerful for any enterprise.
Animesh Samuel, Co-founder and Chief Evangelist, says that the capability for enterprises to be able to get an understanding of how the data that they generate is being consumed, rated and reacted to, was always valuable. With NLPBOTS, that capability is now real-time and easy to access via natural language queries, making it extremely valuable.
“We’ve always had tons of data being generated by enterprises across various touch points. What we did not have was affordable processing power and the technology to make sense of it all, until now,” he adds.
The founders understand the opportunities and challenges when it comes to the fast changing landscape across technologies; Sanjeev Menon had founded NetYantra Inc., an early mover VOIP product company that he sold in 2006. He was Principal Engineer at Net Speak Corporation, Boca Raton, Florida and Head of Laboratory for Information Systems and Telecommunication, Gainesville, Florida before that, working on NLP way back in the late 90s, just after his MS at Univeristy of Florida.
Sanjeev Nair consulted for some of the leading brands in India including Sakaal Group, Raychem RPG, etc. on their digital marketing and design strategy in the late 90s. He then handled operations at Asianet Communications Ltd. as COO, co-ordinating the setup of mobile DSNG access points to feed into one of the first automated newsroom workflow systems set up at a time when high speed networks were still not prevalent in India, post which he switched back to his first love of design and UX. Animesh Samuel headed Compulink India as CEO and then founded KAPS International, a company that handled training programmes for global IT majors, which he took an exit from, to join Light.
The company’s focus is to enable the massive ITES ecosystem in the country to leverage their technology to build scalable solutions across diverse business functions such as HR, Customer Care, Tech Support, and Marketing Intelligence. Light has partnered with leading ITES companies to help them deliver focused AI NLP implementations across these functions. In fact, some of their clients have won global innovation awards for their enterprise applications powered by NLPBOTs, especially in the area of marketing intelligence.
“We’d like to become the de-facto NLP AI powered intelligence layer for every conversation that any enterprise anywhere on the globe automates. Be it customers, partners, resellers, employees, or business owner interactions, NLPBOTS will be powering the conversation at some point,” says Sanjeev Nair.
After being bootstrapped for nearly three years, Light raised $2 million from a private investor and NB Ventures in mid-2016. The current focus remains on building on the strong technology base and rollout a series of products for different enterprise functions, HR being the first.
The applications of the technology are widespread; NLPBOTS can be configured to automate call centre operations like customer support and tech support chats, delivering huge ROI in terms of cost savings and efficiency. The system can be set up to power natural language search interfaces for intranets, effortlessly making sense of structured and unstructured data sets for the domain it is built for. Custom live dashboards focused on marketing intelligence powered by natural language queries can make sales forecasting as well as lead discovery and engagement extremely focused and effective, saving millions in time and effort. HR-focused applications of NLPBOTS have already driven huge savings in terms of time and cost when it comes to hiring the right candidates across jobs and geographies. The HR teams are now more focused on planning and implementing strategies based on the insights NLPBOTS provides based on every user and data set it engages with, thus making their time more valuable.
So what’s next? Well, Light is currently rolling out their platform for partners to quickly build custom enterprise bots for their users and clients. The platform is the only one of its kind that offers auto intent identification, auto disambiguation, and auto context for free-flowing conversations, making it super easy for ITES companies to build and manage scalable AI NLP applications for their clients. The first version of the platform releases soon, followed by quick iterations over the next few months.
“Our partners in technology are an important cog in the overall AI NLP global story, and we have always wanted to offer cool technology that can be easily consumed and leveraged by the ecosystem, thereby ensuring high potential for innovation across industries,” adds Sanjeev.
For Light, this is just the beginning of the journey, with many milestones already mapped out. The road ahead is not easy, as AI in itself is still a grey area for most enterprises, with ROI not clearly defined in many scenarios. Even more challenging is the need for enterprises to look at the data that is available and the processes that can be optimized to derive more data and long-term value from their AI Srategy. To sum it up, if an enterprise wants to build tools and a vision for the future, LISPL is the company to watch out for.