ChatGPT's new features have caught the attention of customer relationship management and customer experience managers. In this article we discuss what CX leaders need to know about ChatGPT and explain how they can use it to improve their customer service.
ChatGPT is definitely one of the big revolutions of 2023. Although it was launched in late 2022, it was in 2023 that the world began to notice the potential of this conversational artificial intelligence chatbot.
Recently, ChatGPT has caught the attention of CX leaders due to its ability to understand and respond to questions in a human-like conversational dialogue format. While not intended to be used as a customer service agent, ChatGPT is still a chatbot, making it an incredibly useful tool for improving the customer experience.
Historically, the customer service industry has used conversational artificial intelligence resources, better known as chatbots, to automate the question and answer process between companies and their customers. Many companies use these technologies to provide faster and more efficient support to their customers.
Following the launch of GPT-4 and the introduction of new capabilities to the already revolutionary ChatGPT, the customer experience community is beginning to explore the potential of ChatGPT as a customer service tool.
To understand how ChatGPT could impact the future of CX, it is important to get familiar with its capabilities, limitations and use cases when it comes to customer care.
Let's start at the beginning.
Chat GPT (Generative Pre-Trained Transformer) is a large-scale conversational artificial intelligence model developed by OpenAI.
Built on a Large Language Model (LLM), ChatGPT interacts in a dialogue format, understanding a wide range of requests and generating human-like responses by predicting future words.
Large linguistic models are artificial intelligence tools —more specifically deep learning— that are trained on massive datasets to read, summarise and translate text, and then generate sentences by predicting future words. The more data used to train a large model, the more capable it is of learning, remembering and making connections between different words, concepts and things.
If the old ChatGPT model, GPT-3.5, already shocked the world with its high capabilities, Chat GPT-4 has the potential to change it.
ChatGPT-4 is an improved version of GPT-3.5, which was trained on a wide range of texts found on the Internet, and was already capable of performing Natural Language Processing (NLP) tasks with high efficiency.
According to OpenAI —the company behind ChatGPT— ChatGPT-4 has eight times the processing power of 3.5, being able to process up to 25,000 words at a time. On the other hand, its ability to understand questions, requests and instructions is also greater.
Another of ChatGPT's great capabilities is its ability to understand, write and classify text. GPT-4 introduces an important new feature in terms of text generation. Users can now send links to the chatbot, as it is able to interact with text on web pages. Thus, users no longer have to copy and paste texts into the chat interface.
Finally, Chat GPT-4 is also able to interact with images. That is, it does not only process texts, but can now also read content in image format.
Although its capabilities are incredible, we must also be aware of ChatGPT's limitations. The most remarkable one is that its results are not always accurate. Although OpenAI claims that GPT-4 makes fewer mistakes than GPT-3.5, we must consider that, even if they seem reasonable, it is possible that the answer given by ChatGPT is incorrect.
One of the great advantages of ChatGPT is that it has many use cases. Among many other things, it writes texts, composes music, solves mathematical problems, writes code, and so on.
Its wide range of uses also includes customer support. Although it was not created as an AI model for customer service, its ability to respond in a human-like manner and its generative chat interface have the potential to transform the ways in which businesses communicate with their customers.
Many companies already integrate chatbots into their websites to provide a fast way for customers to interact with them. Thus, one of the most logical uses of ChatGPT in terms of customer experience is as a chatbot.
However, for its use as a chatbot to be efficient, it would be necessary to train the model with large amounts of data from customer service interactions. ChatGPT is based on machine learning and deep learning. That is, it is an algorithm that has the ability to learn by exploring data. Therefore, if a company wants to use ChatGPT as a virtual customer service assistant, it should train the model and familiarise it with the most common customer service requests, as well as show it the answers to these questions so that the algorithm learns to answer other questions in a natural and human-like way.
Below we list some of the use cases that ChatGPT could have in the field of customer service and customer experience (CX).
We must also remember that, beyond customer service, ChatGPT is already capable of generating advertising slogans and creative copy. It is also useful as a content marketing tool, for identifying leads and as a marketing automation tool.
While it is true that ChatGPT still needs training to be used as a customer service chatbot, its capabilities already foreshadow a future revolution that could culminate in the automation of customer service.
For example, ChatGPT could go beyond answering questions and start performing other actions such as making changes to customer accounts and orders. If ChatGPT were to integrate these capabilities, artificial intelligence could solve a greater number of customer service tickets.
ChatGPT could become a comprehensive customer experience AI platform that treats consumers as individuals, using empathy to resolve each situation in the most responsive way and applying business policies and processes in line with each company's culture.
As ChatGPT progresses over time and becomes more sophisticated, it will impact customer service operations, support teams and the overall customer experience.
One of the main reasons why customer experience experts are taking advantage of the automation capabilities of artificial intelligence is to be able to offer a more productive and efficient experience to their customers. In this sense, the advantages of ChatGPT as a customer service tool go in that directon.
By quickly and efficiently handling large volumes of repetitive customer requests, ChatGPT can help support teams scale their tasks, even when there is a high volume of tickets. Not only would this be beneficial in terms of speed, but it would allow human agents to focus on resolving more complex tickets, which would also add to productivity and efficiency. With greater scalability and increased team productivity, brands could reduce hiring forecasts and ultimately reduce costs.
Below, we list some of the possible use cases for ChatGPT as a catalyst for customer experience optimization.
As mentioned above, although ChatGPT is already considered by many to be a customer service chatbot, it was not trained for it. Therefore, it still has limitations to be considered a comprehensive customer service and CX tool.
What changes should ChatGPT make in order to function as a complete customer care tool?
1. Has to be trained through instructional engineering.
ChatGPT responses vary depending on how the user formulates the question. Asking the same question in two different ways can lead to different results. Therefore, it is critical that the instructions given to ChatGPT are the exact right ones.
To address this challenge, experts train the model by "instruction engineering", which involves generating the right prompts for ChatGPT to understand and take the correct action.
2. Integration with enterprise systems
For ChatGPT to be an effective customer service tool, it must be able to integrate with other business tools as well as with external systems.
3. Corporate policies and processes
On the other hand, it is essential that ChatGPT knows and understands the company's corporate policies, as well as how internal processes work.
4. Integration with analytic software
Analytics and performance management are essential aspects of good customer service. Measuring performance over time is essential for all highly complex systems.
Currently, there is no industry standard test to measure the performance of generative models such as ChatGPT.
On the other hand, if we want to use ChatGPT as a customer service tool, we need to integrate the chatbot with other analytics software such as Google Analytics, for example.
5. Overcoming inconsistencies
As we have already noted, there is no such thing as a perfect chatbot, just as there is no such thing as a perfect human. If humans make mistakes on a regular basis, so do artificial intelligence tools.
ChatGPT does not always get its answers right. Although GPT-4 is less susceptible to mistakes, we must bear in mind that it is possible to get it wrong.
Moreover, if we want the model to act as a spokesperson for our company, we must train it to provide answers consistent with our values, processes and routines.
Conclusion
Although ChatGPT was not conceived as a customer service chatbot, its capabilities are being explored by many customer service and CX leaders. However, to become a comprehensive customer service tool, the model must overcome a number of challenges that can be easily addressed.