Firms waiting for ChatGPT to mature

ChatGPT may have taken the online world by storm but industry experts opine that the artificial intelligence (AI)-powered chatbot will take time to see mainstream adoption even though it is leagues ahead of the conventional rule-based and even AI chatbots that companies are currently using.

Made public last month for beta testing, ChatGPT is the third iteration of Generative Pre-trained Transformer (GPT-3). It impressed even the likes of Elon Musk and Chris Anderson with its potential to write poems, articles, books, tweets, and even code like humans. ChatGPT is trained to predict the next word on a large dataset of text, but GPT-3 can also generate misleading and toxic comments, spread misinformation, spam, and write fraudulent academic essays, limitations which its creator, OpenAI, is attempting to address with human supervision. 

OpenAI is a non-profit AI research company founded by Elon Musk (who resigned from the board but remained a co-chair) and others.

To be sure, companies in India are already using intelligent chatbots that use some form of AI such as natural language processing (NLP) as opposed to conventional rule-based chatbots that can only respond to questions asked in a specific manner.

Nearly 65% of enterprises in a survey conducted by IT industry body Nasscom in June 2022 said they are using AI powered chatbots.

The chatbot industry has also matured a lot since the pandemic. Though rule-based chatbots have proven to be quite effective during the pandemic for answering basic customer queries, organizations are realizing their limitations when it comes to handling more complex queries. “During covid-19, everyone made glorified IVR (interactive voice response) in the name of chatbots. The industry has matured a lot more than it was two years ago. More data is available, and algorithms have become open-sourced,” said Gaurav Singh, founder and CEO of Verloop.io, a startup that offers conversational AI solutions.

NLP chatbots, for instance, are now being widely used by enterprises, and can understand a user’s intent and find the most appropriate response from their databases. Many firms are even using WhatsApp to deploy their chatbots. For instance, in August, Reliance launched its JioMart chatbot on WhatsApp. 

Though most use cases of chatbots are driven by brands that interact with a large number of users, many firms are using them to interact with employees and agents on the ground in BFSI, retail, healthcare, and pharma.

That said, the private sector’s adoption of chatbots is much higher in India than in other countries due to WhatsApp integration, according to Singh. Swapan Rajdev, co-founder and CTO of Haptik, corroborated that with growing requirements, firms are thinking about using chatbots in different areas of business. “Almost every bank is advancing their chatbots. Insurance companies are using it to assist in the field,” he added. HDFC and Kotak Mahindra are some of the banks that are using intelligent chatbots. 

ChatGPT, on the other hand, is a natural language generative (NLG) model based on the GPT 3.5 series that can generate context-specific responses during interactions with users, which allows it to have a more natural and free-flowing conversation unlike existing chatbots. “There are a lot of advancements that have happened on the NLP side. If a customer wants to return an order or a worker asks for paternity leave, an NLP chatbot can understand what they mean. I haven’t seen path-breaking adoption in chatbots on the NLG side. That is where these generative AIs will come in,” said Singh.

However, even after it is launched commercially, ChatGPT’s applications in companies will take some time, most experts say. “Using ChatGPT, brands can save time and energy that would otherwise be spent training their chatbot on industry-related questions. ChatGPT is a great leap. It can help businesses improve their experiences. But, it is still in its early stage. There are a lot of design systems that need to be made on top of it,” said  Rajdev. 

Arup Roy, research vice president at Gartner, corroborated that it is still early days for conversation AI chatbots. “Technology is evolving quite fast but it is still evolving.” Roy noted that GPT3 and BERT (Bidirectional Encoder Representations from Transformers) are already being used in conversation AI space. However, in terms of enterprise adoption, it is BERT that is more predominant than GPT, he pointed out. 

Deploying an advanced chatbot also comes with certain expenditures, even though entry into the space for an end-user organization is not a big barrier. “You can deploy a plain vanilla FAQ chatbot for $10,000. A more sophisticated one can cost $75,000 to 150,000,” Roy added. In terms of adoption, Roy has seen most of the mature chatbots being used for customer service and support followed by ITSM (IT service management), HR, and agent assistance. 

Verloop’s Singh is also wary about rushing ChatGPT to customer-facing functions. “If ChatGPT makes a blunder people will have a good laugh and move on. But if it does that while being used by a company it can have severe repercussions on business,” he warned. Singh pointed out that companies are tightly guarding how results are generated to match their marketing tone. They all want it to have a persona, and ensure that it doesn’t show any discrimination.

Adit Jain, co-founder and CEO, Leena AI, believes that ChatGPT is poised to be adopted widely in a couple of years. “There will still be major concerns about the context, depth, and breadth of what chatGPT can provide to the Indian audience,” he added.

Rajdev concluded, “While ChatGPT is a great tool to process unparalleled knowledge and tap it with human-like conversations, taking action on these conversations will require APIs (application programming interfaces) that connect with systems like payment, marketing, and commerce that are critical for conversational commerce integrations with CRM (customer relationship management) systems”.