Chatbots have come a long way, from the elementary text-based chats to much advanced conversational platforms mimicking near-human interactions. Chatbots are changing how businesses interact with their customers, replacing humans in query management for various fields like banking, e-commerce and retail. As it stands, artificial intelligence, machine learning, and natural language processes are setting the trends for the chatbot industry in 2021.
Leading the way into the future of human-machine conversations is Ori — an end-to-end provider of conversational AI-powered chatbots. The bootstrapped startup has hit the ground running clocking a positive cash-flow in the first year of operations. Ori’s celebrity client list includes Vodafone, Idea, Tata Motor, DishTV, and Supreme Golf.
Microsoft has named Ori as the ‘Best Customer Engagement Chatbot’ in the Asia Pacific. Analytics India Magazine spoke to the co-founders, Maaz Ansari and Anurag Jain of Ori, to know how the company uses AI to automate the customer journey and bridge the gap in customer-client interactions.
The founders of Ori — Maaz Ansari and Anurag Jain — met in 2012 while working as analysts at Fractal Analytics — a multinational AI company, focusing on text analytics and NLP research. “This was a time when analytics was in its nascent stages and about to come to the mainstream. Terms like AI and ML were heard far and in-between,” said Ansari. The duo found the idea of deriving meaning from unstructured data and finding patterns from chaos too good to pass.
Engineers by training (from Bombay University and IIT-Kharagpur), the pair saw potential in driving effective conversations between enterprises and customers. “Conversations that form the crux of the relationship, conversations which, when done well and lead to extraordinary results,” clarifies Jain. They realised the existing market solutions did nothing to sustain the hype cycle. Bots weren’t up to the mark, and Ori was conceived to bridge the gap between expectations and delivery.
“With a hand on heart, 90% of folks using bots even today will tell you that they neither match their expectations nor do they fulfil the roadmap of expectations that they had. In other words, there is a sizable gap that is required to be filled,” Jain said.
Sharing the story of the company, Ansari said — “We had just shut down a startup, and were broke. Armed with two slides, we managed to convince the President of Strategy at DishTV to give us a break. He is a big believer in product innovation and jumped onto the chance of leading the way with breakthroughs. Consequently, the DishTV bot became the first AI-driven bot built for customer redressal by any Pay-TV company globally. It went on to win the ‘MSFT AI for All’ Award as the best bot for customer engagement.” Everything changed afterwards. With business steadily flowing in, Ori continued its innovation-driven approach and developed solutions for various use cases. The company features benchmark-defining NPS/CSAT on each conversation. “All of our bots are trained specifically for an organisation and imbibe the brand’s character, jargon, and customer communication norms with our patent-pending algorithm,” he added.
According to the founders, Ori’s platform truly mimics human conversations. Not only they are authentic, relevant, and empathetic, but are also being layered with multiple languages to create a solution that’s transformational in absolute terms. “When people talk about AI, they refer to natural language processing; in other words, the ability to understand humans. That’s where the buck stops,” said Jain.
Ori’s multilayered AI stack is customisable. The flagship product is a digital sales assistant that completely emulates one-to-one conversations between brand personnel and customers, and has been deployed for multiple clients across diverse segments, including automobile and consumer durables.
Being in the business of automated digital sales representatives using cognitive conversations, Ori has developed many AI and analytics modules. These models have been designed to synchronise with each other to create a world-class experience for users and maximise value for its clients. Natural language processing techniques are at the core of AI and analytics used at Ori.
Some of the AI techniques leveraged in different stages at Ori to enhance its automated sales representatives are:
AI to learn the basics of conversation: The aim here is to learn more about the brand and its domain using deep learning techniques. Ori’s automated sales reps start gauging the brand’s lingo and industry-specific ontology from tons of data gathered from brand vaults. Alongside, industry knowledge is gathered by web scrapers.
AI to gain business insight: Under the supervised contextual business logic training, the team taps supervised deep learning techniques to train the engine. It helps the engine to predict the next action based on business logic and context.
Learning from successes and failures in a live environment: Reinforcement learning is used in the products to reinforce best scenarios based on set goals for one-on-one conversations and readjust the model in real-time.
The full suite of analytics at play to optimise Ori’s automated sales reps behaviour includes:
Funnel Analysis: Ori created a funnel view of all user interactions, which helps the engine analyse progression down the funnel and effectiveness of different campaigns run by the company’s automated sales reps.
Session Flow Analysis: The team at Ori analyses the best conversion paths and gets user behavioural insights.
Anomaly Detection: Proprietary anomaly detection technique has been developed to identify and surface fracture points from thousands of conversations processed by Ori’s automated sales reps.
Free Text Clustering: Ori has also developed a novel clustering technique to visualise top unhandled queries in the form of clusters.
Talking about the core technology stack, Ansari said, “At Ori, we are following a microservices-based architecture and leverage the MERN stack for powering the platform. It enables us to handle volumes at scale.” The machine learning, AI, and NLP libraries are built on Python, TensorFlow, Keras, PyTorch, and Transformers.
Hiring Process At Ori
Ori focuses a lot on internal promotions for specific roles within the company, and has also created a talent community on social sites alongside strong employer branding. “We further ensure that we maintain complete transparency with applicants throughout the hiring process, reducing time to hire,” he added.
Ori even has a strong employee referral program that helps the company address its hiring needs. Passion and enthusiasm to work in a growing startup are the key traits Ori looks for in every candidate. It also expects the candidates to have skills like team compatibility, a proactive and can-do attitude, strong leadership qualities and good communication skills.
With its AI-powered chatbots, Ori can take care of the most complex queries with automated customer service response systems. According to the founders, it empowered the clients to service their customers at only 50% of customer service centre capacity, reducing costs by 60%.
In the future, the company aims to develop feature-intensive, AI-powered assistants that help brands get higher ROI and return on ad spend from their sales and marketing initiatives. “We are also increasingly focusing on voice-based chats and search technologies for our conversational platform,” concluded Ansari.