Discussion of Pros and Cons
After reading about PulsePoint Solution’s dilemma regarding whether or not to deploy and a generative AI chatbot for their customer service and sales, they have quite a few reasons which push them in either direction. In terms of motivation to implement a customer-facing chatbot, ensuring that they are technologically competitive by keeping up with technological advancements is critical. Additionally, chatbots would greatly reduce headcount, even “scal[ing] [head count] back by as much as 30% over five years.” Additionally, utilizing a chatbot would enable the company to “anticipate prospective customers’ needs and respond to them faster and more precisely than before, potentially increasing [their] sales and customer satisfaction.” On the other hand, there are several reasons why deploying this chat bot might not be such a great idea. For one, one of PulsePoint’s largest clients expressed concerns regarding “data privacy and the potential mishandling of sensitive information.” Additionally, with chat bots, there are always concerns regarding hallucinations, reliability, and accuracy, which adds reputational concerns for the company.
My Advice
I do not think the negatives of deploying a customer-facing chat bot are inconsequential—they are legitimate concerns that can introduce long-lasting, detrimental impact. However, the idea of not utilizing a gen-AI chat bot and technologically falling behind compared to their competitors is also a serious concern. Therefore, I recommend deploying this chat bot internally at first. It can be utilized as an assistant to customer service and sales representatives, which would still increase company efficiency. As this chat bot becomes increasingly utilized, employees should provide feedback for the chat bot for each response, training it over time. At my previous internship, we utilized exactly this approach until the chat bot met a specific threshold of accuracy, at which we released it to our customers. This approach of initially internally deploying it, training it with response feedback, and then transitioning it to a customer-facing product greatly minimizes the risk of hallucinations and reputational risk towards customers.
