One of the realities of an entrepreneur’s life is the high possibility (sometimes, inevitability) of failure. It was no different for Vishwanath Jha, who dons many hats - engineer, serial entrepreneur, AI thought leader and former analytics consultant. After dabbling with different ideas, Jha started a data analytics firm in February 2017, with the express purpose of building world-class voice and language technologies in Indian languages. But consulting wasn’t his calling. “Instead, me and my cofounders decided to develop a product platform for conversational AI. And this is how Saarthi.ai was born.”

Along with Sangram Sabat, Piyush Dangaich and Sameer Kumar Sinha, Vishwanath has developed Saarthi.ai to become one of India’s foremost platforms for omnichannel enterprise engagement in vernacular languages. Three out of four cofounders have been analytics and data science consultants, and have worked with several Fortune 500 companies on data-driven strategies. The collective experience helped them understand the value of building similar data-driven cognitive engagement models for enterprises that would deliver strategic, granular insights.

Having started out in 2017, Vishwanath is all too aware of how the market perception towards conversational AI has changed. “We had a tough time initially, especially while convincing enterprises to invest in chatbots. Ultimately, businesses look at cost savings and how a product is augmenting their bottom line. This led us to develop a focused strategy for building bots.” Today, developing bots has become simpler due to the availability of basic datasets, compute and cloud services and engineers. He says, “There are bots that cost as low as Rs. 5,000 and can go upwards of $100,000. It all comes down to what that bot can do for your business.”

While the logic seems fairly straightforward, the reality of achieving this goal was anything but. Vishwanath and his team spent months researching the specific needs of businesses, doing a 'deep-dive' into their datasets to identify areas where bots could deliver monumental impact for a business. However, a tangible positive change in perception towards bots started happening only after COVID19, believes Vishwanath. The risk of lapsed communication networks and inability to physically deploy agents in contact centres were among the first effects of COVID19, and this led to a frantic scramble for harnessing contactless communication services.

Despite the rise in demand, Vishwanath believes it is imperative to deliver customized value to clients. “Not many people can gauge two or more chatbots quantitatively. While decision makers are finally seeing the need to invest in bots, it is equally important to deploy and develop one that delivers value based on one’s vision for the business. With this goal in mind, it becomes easier to identify the right solution provider.”

Typically, an end-to-end cognitive engagement platform should be scalable (with a focus on call volume ability), should handle queries, service requests, process feedback and more. At Saarthi.ai, a thorough analysis into a business and its various metrics serve as critical data points for the platform, which can have a substantial impact on its efficacy. There are three kinds of cognitive engines that Saarthi.ai aims to deliver – social chatbots that continuously engage with the customer for maximum conversions; knowledge-based bots that aim to address customer needs in terms of information gaps; and goal oriented bots that possess a range of skills, designed to identify the nature of a query, its context, intent and so on. “This level of contextualization is the most challenging aspect of NLP engines today. Dialogue modelling isn’t one-dimensional – there are so many nuances and meanings to consider, especially when done in voice – so the machine has to work harder to identify context and make logical conclusions. Vishwanath says their platforms have become adept at personalizing services for clients, with custom-made templates and exhaustive domain ontologies. A short turnaround time of up to four weeks is a key offering to enterprises as it shortens their go-to-market time considerably. A data annotation platform called Pravid, built by Saarthi.ai, helps data annotators tag conversations, data and build fluid conversations on the platform.

In 2020, Saarthi.ai was included in the Microsoft for Startups programme to scale its multilingual conversational AI platform for enterprises. Markets&Markets hailed Saarthi.ai as a key market player in India. Since COVID19, Saarthi.ai has also been working on enhancing contact centre analytics – with every incoming call, the inbuilt AI engines passively learn from the data, while also flagging off key metrics and data points for human agents to act on. Other products that Saarthi.ai will be releasing include a call centre assistant, a multi-model assistant with voice and text interface for apps and websites, which will also have a ‘wake word’ to build brand identity, and an omnichannel voice assistant to help customers discover products, mainly for ecommerce, telecom and healthcare.