These days, the internet is abuzz over ChatGPT. The technology is a text-based artificial intelligence (AI) tool developed by OpenAI and designed for use in chatbots and conversational systems. The power of this technology lies in its ability to understand and learn from conversations and deliver humanlike, personalized, and detailed responses. This conversational AI has great appeal in the contact center environment, which would benefit from advancements in intelligent automation; natural, personalized, and precise responses to customer inquiries; and workplace efficiency gains.
The technology’s promise to personalize responses, content, and offers; serve up detailed product descriptions; analyze customer reviews; and troubleshoot technical content will also give it prominence in the contact center. Scott Draeger, customer experience officer at Quadient, added that he is excited about ChatGPT’s ability to personalize to larger segments than some companies can currently reach: “This new tech could truly personalize communications, factoring in people’s individual communication and engagement styles, interests, and other factors.”
Don’t replace human interaction or search with generative AI
Much of the discussion around the potential use cases of ChatGPT centers on the possibility of the technology being positioned as a next-generation, conversation-based search engine. However, ChatGPT is a language model trained on large datasets to generate humanlike responses to questions, whereas search engines use algorithms to index and rank webpages based on queries. Therefore, ChatGPT is not designed to search the internet and provide information in the same way as search engines do. Draeger elaborates: “People say this will replace search. I disagree. It’s nothing without search. It’s an inflection point. It’s the inflection point bitcoin thought it was.”
Despite ChatGPT’s low-hanging-fruit cost benefits and efficiency gains, the technology succeeds most often at simulating human responses and responding to natural user input, and its extensive knowledge base allows it to skillfully navigate customer conversations. However, it falls short when the conversation becomes complex, and it does not have the necessary knowledge for a particular inquiry. Then, the technology instead fills in the gaps. “One of the big things with generative AI models is how you deploy that technology safely. It can generate amazing results that are perfectly accurate. But it can also generate results that look well considered, are articulate but completely wrong,” said Ed Challis, general manager at Re:infer, a UiPath company.
Additionally, many experts caution that ChatGPT should not be used as a replacement for human interaction. Some who have deployed it in their organizations have said that it creates new sentences for every inquiry and fails to follow a predefined script; therefore, the technology has been known to give incorrect answers. The failures could potentially stem from leveraging the technology as a standalone solution. Scott Jennings, head of retail and e-commerce industry GTM at Twilio, said ChatGPT works well when it is incorporated into an existing system. He said companies such as Twilio, with technology that wraps around systems, are well-positioned to take advantage of new offerings such as ChatGPT.
Working toward ChatGPT’s inflection point
Although ChatGPT may work well with repetitive inquiries such as “What is the status of my order?” most of the customer service market already largely automates such routine inquiries. In the short term, generative AI tools such as ChatGPT will show value in assisting or optimizing agents’ workloads and automating tasks rather than their jobs. Long-term opportunities for ChatGPT lie in the customer engagement communications that fall outside of repetitive service inquiries and require a powerful, conversational chatbot. Widespread adoption of ChatGPT will depend on the continued advancements of NLP and machine learning. As Jennings said, “Ultimately, the opportunities lie in the eye of the beholder.”
