Listen to this story
ChatGPT was a paradigm shift in conversational AI and Indian companies were quick to identify its potential. A vast majority of customers are approaching these companies to explore the possibility of replacing their current chatbots with ChatGPT.
“Almost every customer is engaging with us on having a GPT-powered bot for their website,” Praful Krishna, chief product officer at Jio Haptik told AIM.
Given that the technology is in a nascent stage, conversational AI companies are still preparing for the widespread implementation of LLM-powered chatbots. Even though the price remains an issue for now, it is expected to become cheaper in the days to come.
Notably, ever since ChatGPT’s launch, the cost has gone down 10 times. With more and more open-source LLMs coming up, the price will go further down, and it will make such bots easily accessible for most enterprises.
What LLMs bring to the table
Indian enterprises are eager to adopt LLM-powered chatbots, not taking in to the buzz they have generated, but mostly because of what LLMs bring to the table. LLMs have helped solve language understanding and generation in a big way. While earlier, chatbots were designed to solve a specific task, LLMs paved the way for more advanced and sophisticated conversational agents that are better at understanding and responding to natural language input.
Using LLMs helps save weeks of work in this new workflow. So, obviously it’s a much more efficient process and it’s faster to launch bots that have a natural human-like interface. “When it comes to language understanding, multilingual support, text-to-voice and voice-to-text conversion, things just sped up a lot using LLMs,” Gaurav Kachhawa, chief product officer, Gupshup.io, told AIM.
How LLMs are changing the industry
ChatGPT has forced conversational AI companies rethink their product offerings altogether. “We’ve been working with LLMs for quite some time now. However, the use-cases and precisely what LLMs can solve are things we did not explore before ChatGPT,” Jaya Kishore Reddy, co-founder and CTO at Yellow.ai, told AIM.
With ChatGPT coming into the picture, Yellow.ai also changed its roadmap pretty quickly. “We started using LLMs not just for content generation but as an intelligence layer. We have been working on smaller use case-specific LLMs as well as experimenting with different other LLMs, not just GPT-4. We work with Anthropic as well and have tested Meta’s LLaMA in house,” Reddy revealed.
Yellow.ai is also working on an in-house LLM, designed specifically for enterprise consumption. It is expected to be launched soon.
Similarly, Gupshup, a conversational AI company launched in 2004, was quick to jump on the bandwagon. Gupshup is not just building chatbots powered by GPT-3 and GPT-4, it is also experimenting with GPT3.5 Turbo and Anthropic’s Claude, LLMs developed by Google and others.
“So, what we want to do is create a ChatGPT-like experience, but more on the unique data for the enterprise. We have taken the language understanding of the LLMs and trained it, finetuned it to specific enterprise data,” Kachhawa said, in a previous interaction with AIM.
Haptik, the conversational AI company acquired by Reliance Jio, is also using LLMs for three different purposes. Firstly, it is using LLMs to make its existing offerings better. “Secondly, we are looking at re-engineering our entire AI stack and offer to our customers the option of a very advanced AI stack running on top of LLM,” Krishna said.
The Mumbai-based startup is expected to launch these solutions in a quarter or two. “Lastly, we are also exploring what kind of innovation we can do in the LLM itself. Can we take an existing LLM and rewrite for the conversational commerce use case,” Krishna asked.
Indian enterprises are hungry, but cautious
“There’s no customer conversation without GPT these days, because the expectations have raised significantly,” Reddy said. In fact, a few of its customers across the BFSI and automobile sectors are currently leveraging Yellow.ai’s generative AI-powered capabilities for their customer support and marketing use-cases.
Haptik, too, is witnessing something similar, Krishna revealed, especially from customers in retail, finance and the travel and hospitality industry. Almost all of Haptik’s conversations with customers involve ChatGPT or LLMs. “Look at our customer list, from Hero, Paytm to Byju’s, I can tell you that almost everyone is asking for it. In fact, we have seen approximately around 40% uplift in the sales numbers just because of this particular thing,” he pointed out.
Gupshup too has received enquiries from over 1,000 enterprises, with over 100 of them turning into leads. “We’re already seeing a lot of demand across different industries. I think everybody will start with a smaller volume to learn and understand how it’s helping them drive more sales, automate processes and things like that,” Kachhawa said.
However, enterprises are also wary of the risks LLMs pose, especially enterprises in the banking or insurance sector that handles sensitive data such as Personal Identifiable Information (PII). Further, LLMs also have the disrepute of getting hallucinatory. “Enterprises want to know how to keep the LLMs grounded. How do we make sure that it doesn’t blabber and only sticks to the context?” Reddy said.