Here is another example, where I asked a medical question about epilepsy. As you can see in the screenshot, it gives me some generic information about epilepsy and then demurs further answers. When I asked this same question a week ago, it gave me very specific information about clinical studies, drugs in development, and treatment centers—proving it to be a capable conversationalist around medical themes. However, because of the risk inherent to offering medical advice and information, and due to the fact that ChatGPT can sometimes offer plausible sounding but made-up facts, OpenAI has been turning off many of these risk-prone areas.
ChatGPT is capable of much more than just chatting.
Adaptable to different writing styles.
Below, you can see it writing a poem about Napoleon in the style of Shakespeare. It is able to adapt to the prompt and write in a stylistic and expressive manner.
It can engage in creative content marketing and role-playing. In the role-playing scenario, it takes on the characteristics of a specific personality and engages in a fictional conversation.
Writes code like a pro.
- Write unit tests and debug code.
- Explain and understand complex concepts such as Regex (a regular expression—a sequence of characters that specifies a search pattern in text).
- Perform data entry tasks and format data.
Processes Natural Language.
ChatGPT can do translation, intent detection, and sentiment prediction.
These are some of the hundreds of possible use cases that range from playing chess to writing fanfiction about Star Wars.
Adapting ChatGPT and LLMs to the contact center.
ChatGPT is a static model that was trained until 2021 and does not have access to the internet or the ability to be retrained or fine-tuned. This means that, at the moment, it can’t be customized for specific purposes, such as providing customer service answers tailored to the needs of a business. So, making use of ChatGPT today for the contact center is not possible.
However, I expect that an API for ChatGPT will be available at some point. When it is, it would be possible to feed ChatGPT with agent transcripts, website data, and customer order history. This would enable it to provide personalized and detailed answers to customers. The impact of having a single LLM virtual agent with extensive knowledge of business and the ability to converse with customers on a variety of topics, with limited training to do so, is the game changer that all of the contact center pundits (including myself) are imagining.
Of course, it is important to consider the potential ethical implications of such a powerful tool, particularly with regards to bias, transparency, and privacy. Though these have always been important considerations for businesses as part of their duty of care towards customers.
My top predictions about the use of ChatGPT and LLMs in the contact center.
The advancement of ChatGPT means it’s possible to create virtual agents that can converse with customers in a natural and reasonable manner about a business, its policies, and customer specifics all without the 100s of hours it now requires to train bots to perform similar tasks.
This has been the goal of virtual agents for the past decade, but until now, it has been difficult to achieve. As a result, the impact of ChatGPT, and other LLMs, is likely to be significant to the customer service industry.
My first prediction.
Companies will replace many human agent tasks with LLM virtual agents, which are a more cost-effective option for roles where human empathy and judgment are not needed. Companies are likely to pursue the cost-saving benefits of using this type of virtual agent.
LLM virtual agents will enable companies to serve more customers simultaneously, increasing their efficiency and allowing them to handle a larger volume of inquiries. This could potentially lead to faster response times and improved customer satisfaction.
My second prediction.
Agents may start handling multiple phone calls simultaneously. This will be achieved through the use of LLM virtual agents that are able to handle most customer inquiries independently but can defer to the human agent for confirmation when their confidence is low. This will enable the human agent to serve multiple customers at the same time, while still providing a high-quality customer experience.
In this scenario, the human agent acts as a quality assurance layer, a human-in-the-loop, providing real-time feedback and corrections to the virtual agent, enabling it to learn and improve over time while still being able to handle most inquiries independently. It is likely that this model will be adopted by many companies, as it allows for a more efficient use of human agents and provides a superior customer experience.
My third prediction.
Due to the coding capabilities of ChatGPT, I can imagine a shift from “low code” to “prompt-based” user interfaces—in particular for contact center administrators. For instance, an administrator could use prompts to create a user interface for agents that connects to a support ticketing system and retrieves information about a customer’s previous ticket—all just by describing the interface—without any specific coding to do so. Prompt-based administrative tools would dramatically simplify the setup, configuration, and customization of the contact center.
I truly am blown away by the capabilities of ChatGPT and I can’t wait to see what the future holds. If you haven’t tried it out yet, I highly recommend giving it a go. It’s truly a game-changer!