In the Indian generative AI ecosystem, only a few companies can claim to have a strong customer base, but CoRover.ai stands out as a clear exception. The company, known for its human-centric conversational AI platform for enterprises, claims to have 1.3 billion users today.
With a wide userbase, CoRover has an edge over its competitors Krutrim AI and Sarvam AI, the early-stage startups that have launched products for enterprises. Ola chief Bhavish Aggarwal who heads Krutrim, recently claimed that over 10 million people use their chat app. Sarvam AI’s userbase remains unknown.
Corover.ai offers a range of conversational AI platform-based virtual assistants, including chatbots, voicebots, and videobots, to hundreds of organisations. Its clientele includes notable names such as IRCTC, LIC, IGL, KSRTC, the Indian Navy (GRSE), Sneha AI, NPCI, and the Government of India, among many others.
The company, known for creating BharatGPT, its own indigenous LLM, recently secured $4 million in a Series A funding round led by Venture Catalysts. It was also accepted into NVIDIA Inception, a program designed to support startups, granting access to NVIDIA’s resources, including GPUs, unparalleled computing power, and robust software support.
With the latest funding round, the company is planning to expand its expert team. “We’re expanding our expert team with scientists advancing research, technologists developing innovative solutions, growth experts scaling impact and adoption, and psychologists enhancing customer experience through human-centric design and personalisation,” said CoRover chief Ankush Sabharwal in an exclusive interview with AIM.
“By merging technology and psychology, we truly want to understand human needs, driving business growth, operational efficiency and customer satisfaction, which would also help boost client product adoption through intuitive UX and personalised experiences,” he added.
Growing Customer Base
The company recently partnered with NPCI and IRCTC to introduce ‘Conversational Voice Payments’ for UPI at the Global Fintech Fest 2024.
This new feature, integrated with payment gateways and powered by BharatGPT, allows customers to complete transactions using their voice or by typing their UPI ID or mobile number. When a mobile number is provided, the system automatically retrieves the corresponding UPI ID and initiates the payment request via the user’s default UPI app.
Besides this, the company has built an assistant for banks and UPI that serves as a conversational alert system. “For example, if a bank like SBI is down, the monitoring staff receives an alert. They can also inquire about how many times it has been down and the top reasons for the outages,” explained Sabharwal, adding that it’s presented in a chat format.
The company has also built a grievance system for the J&K government, called eSevak. Sabharwal explained that it is available on WhatsApp, and complaints go directly to the government dashboard and the respective team responsible for resolving the issues.
He added that they are also building text-based virtual assistants for the Delhi Metro and the Karnataka State Road Transport Corporation (KSRTC). CoRover has created ‘Ask Vaani’, an intelligent virtual assistant that helps travellers with their queries in English and Kannada.
The company also created a video bot for Chennai Police. “This bot allows people to interact with a real policewoman. Currently, it is effective for students who can register complaints, but not FIRs. There is a feature called CSR, which enables users to report community-level issues virtually,” said Sabharwal. The company is also working with the Election commission of India for internal use cases.
Most recently, CoRover partnered with ChangeInkk to launch AskPurpleFlame, India’s first multilingual AI-driven help desk focused on learning disabilities. This chatbot supports over 200 million Indians with conditions such as dyslexia, dysgraphia, and dyscalculia by providing resources and promoting inclusiveness.
Earlier this year, the company launched an educational tablet called Milkyway for students.
Three-Layered Tech Stack
Sabharwal explained that their conversational chatbot tech stack consists of three layers. The first is classic NLP, the second is BharatGPT, and the third allows customers to integrate their own external LLMs from third-party vendors.
Regarding classic NLP, he said, “There is no GenAI, it is our own engine, which we have developed over the last seven years.”
He further explained that for BharatGPT, the base model is based on Pythia, which they fine-tuned using parameter-efficient techniques. “We trained it with a significant amount of Indian conversational data in 12 languages,” he said, adding that it supports 14 languages by voice and 22 by text.
He shared that BharatGPT is a small model that uses RAG based on the content provided by companies. Customers can upload their documents and link their websites to supply their own content.
“They then choose BharatGPT, after which the system creates embeddings and a vector database with typical chunks. Whenever a user asks a query, those chunks are dynamically retrieved. The data is owned by our clients, for which we generate the answers,” he said.
The third layer allows developers to use Gemini or ChatGPT and add any custom LLM. “If a user asks a question, such as “Who is the Prime Minister of the UK?” the query would be directed to Gemini or ChatGPT, depending on the configuration,” said Sabharwal.
Future Plans
Sabharwal said that the company next wants to enter video commerce. He explained, “Video commerce means creating a shopping experience similar to that of visiting a physical store, where customers can see hundreds of products at a glance.”
He contrasted this with online shopping platforms, saying that they often present a limited selection of products. “When you open a platform like Amazon or Flipkart, you might see only a handful of options at a time,” he said
Sabharwal believes that video commerce can replicate this engaging experience of physical stores by providing a dynamic online environment where users can interact with a broader range of products.
This will not be the company’s first foray into the video segment. Earlier this year, CoRover introduced AI Realtor, the world’s first video bot built exclusively for the real estate industry. It streamlines the home buying process, making it smarter, faster, and more efficient.
The post How CoRover.ai Reached 1.3 Billion Users with Conversational AI appeared first on AIM.
