SkyNet? — The birth of intelligent customer service chatbots
When it comes to customer service, chatbots have become indispensable in recent years. They act as virtual assistants, working tirelessly around the clock to enhance the customer experience by handling inquiries and resolving issues promptly. However, not all chatbots offer the same level of service quality.
Conventional chatbots, such as those commonly found on platforms like Zendesk, operate using a decision tree-like structure. It’s like choosing which turn to take in a path where the end is obscured by a dense forest. If your quest is simple and predictable, this approach may lead to a satisfactory result. But more often than not it leads to a dead-end or worse, sending you into a circle looping you around and around, causing frustration and eventual anger.
It’s in this fundamental design where the limitations of traditional chatbots are laid bare. They operate under the assumption that the programmer has anticipated your question or problem and created an appropriate response. If your issue is common, this methodology may work efficiently. However, for more nuanced or unanticipated inquiries — such as a request for the company’s annual turnover — you may find yourself navigating through a labyrinth of irrelevant options, without a proper response at the end.
Note: When companies like Zendesk refer to their chatbot utilizing AI, they are essentially saying that they have employed AI or machine learning techniques to interpret user instructions out of a list of predefined options. While this may appear impressive, it should be noted that this process is not truly intelligent. The computer essentially presents three potential options and makes an educated guess as to which one best matches the user’s query (even a monkey will get it right 33% of the time). At no point is the computer actually comprehending the user’s request.
Nexoid’s chatbot is powered by GPT-3.5 Turbo. Unlike traditional chatbots that rely on decision trees, Nexoid’s chatbot leverages the advanced capabilities of GPT-3.5 Turbo to provide more intelligent and personalized assistance.
With GPT-3.5 Turbo, the chatbot can think, adapt, and make informed decisions better than the average helpdesk agent. It really is like having an intelligent human available to assist you. This new approach allows for a more seamless and efficient user experience, making problem-solving and communication easier than ever before.
Instead of relying on a rigid set of predefined rules, our approach to training the chatbot is more akin to teaching a human. We provide the chatbot with a clear language based instruction “You are an IT helpdesk agent, with access to various functions such as user lookup, ticket creation, ticket updates, and ticket search. Your job is to help the end user however you can.
The remarkable intelligence of GPT-3.5 Turbo allows it to not only comprehend the user’s request but also determine the most effective way to assist the user by utilizing these functions. For instance, when asked to create a ticket, GPT-3.5 Turbo automatically prompts the user for their name and email address in order to properly identify them. Once the user is identified via the user lookup function, the chatbot proceeds to create the ticket using the correct user ID. This ability to perform multi-stage logical problem-solving showcases the chatbot’s impressive capabilities. It also shows how this technology could be used in many other use cases not connected to helpdesk and traditional IT support.
You will also find that GPT-3.5 Turbo will be able to assist the user in unexpected ways. For example you could repeat the entire “create ticket” process in german and it will still work. Without any additional coding or instruction GPT-3.5 Turbo can interpret and understand that the user uses german and backend functions use english.
So the question for you is… if given the choice between a dumb chatbot following a decision tree or the chance to deal with an intelligent artificial assistant which would you choose?
We see the future of this technology as a lot more than simple customer support. We are already experimenting with making the GPT-3.5 Turbo part of the Nexoid user experience helping ITSM and ERP users navigate and streamline their day to day work.