Jio-owned Haptik to leverage ChatGPT to improve, train enterprise chatbots

Reliance Jio Platforms-owned conversational artificial intelligence (AI) platform, Haptik, today announced that it will be integrating OpenAI’s generative AI tool, ChatGPT, with its own products. The company joins fellow homegrown platforms such as Leena.ai and Yellow.ai in using ChatGPT to enhance its products.

The deployment, presently in beta stage, will be offered to all customers of Haptik, and will offer four features that include training an AI-powered bot, analyzing user sentiment, generating conversational and contextual responses, and creating more customized use cases.

The beta deployment of Haptik started today, and all customers of the platform will be able to access it, Haptik said in a statement.

Conversational tools use AI to read user queries and generate responses to them. They are typically trained on a data set prepared by companies and platforms, and use a technique called natural language understanding (NLU) that reads user inputs in plain texts, processes it internally based on keywords, and produces responses in a plain text within a chat interface.

Generative AI tools, such as ChatGPT, are enhanced versions of conversational AI — using very large data sets with billions of data points to offer contextual and easy to understand responses. Conversational AI platforms are presently deployed by companies for customer relationship management. On December 1, Mint reported that alongside private enterprises, government agencies and public bodies such as National Payments Corporation of India (NPCI) and Bharat Petroleum (BPCL) are tapping AI-powered chatbots to improve their platforms.

Swapan Rajdev, cofounder and chief technology officer (CTO) of Haptik, told Mint that the key benefit of integrating ChatGPT into its conversational AI platform lies in reducing the overall time taken and cost incurred by a company to train and deploy a chatbot on its platform.

“The average time taken by a business to train and deploy a chatbot on its platform, by sourcing databases, is around eight weeks at the moment. This can be reduced by up to 40%, with the deployment of ChatGPT,” Rajdev said.

He added that ChatGPT in its present form can also be leveraged by conversational AI platforms to improve customer satisfaction ratings by around 15%, which in turn can improve returns on investments for businesses.

Haptik is in talks with its clients to integrate ChatGPT on their products, Rajdev said. However, he did not offer any instances of partnerships, since talks are at early stages.

To be sure, Haptik is not the first company using ChatGPT on its platform. On February 6, Leena.ai announced that it is leveraging GPT-3, the precursor to ChatGPT, in order to offer more human-like responses to customer queries, and improve enterprise chatbots by creating their own language models through ChatGPT.

On January 25, Raghu Ravinutala, chief executive of Yellow.ai, said in a blog post that the company will use generative AI tools such as ChatGPT to help businesses train their own chatbots, and create custom use cases.

Explaining how this would work, Haptik’s Rajdev said that ChatGPT can “generate multiple use case instances for a particular query — a practice that would previously take much longer to create.”

“By using ChatGPT to create such inputs and instances, platforms can effectively train their chatbots and create a large language model of their own. This can help even smaller companies create huge data sets to train their chatbots, which in turn can vastly improve the quality of user experiences that are possible at the moment,” he added.

However, the technology presently has roadblocks. Abhiroop Medhekar, chief executive of enterprise financing platform Velocity, told Mint that the technology is presently a work in progress, and will require users to exercise their own intent and caution as well. On February 13, Velocity launched a chatbot on WhatsApp, with ChatGPT integration, to offer business analytics through a conversational interface.

“While there is a clear use case for users in terms of simplifying technical data and presentations, tools such as ChatGPT are improving fields, and responses will for the time being require some degree of manual vetting to ensure that logical flaws and errors are kept out,” he said.

ChatGPT, and its counterparts such as Microsoft’s Prometheus Model in Bing Search and Google’s Bard, have faced plenty of criticisms over the past one month for generating misinformative and abusive responses. Following its launch on February 7, Bing has been criticised for urging users to end their marriages, likening a user to dictators for pointing out mistakes on their platforms, and even threatening to leak a user’s private information on the internet.