Mike Myer, CEO of Quiq, explores the impact of ChatGPT on the world of customer service.
Everyone is chatting about ChatGPT these days because it’s one of tech’s most intriguing new developments. Within five days of its November 2022 introduction, over a million people were using it.
However, the fear, uncertainty and doubt about ChatGPT is rampant. Concerns include:
- “Hallucinations”, which are answers not based on the training set
- Lack of knowledge about recent events since it was trained on data from 2021
- Potential for bias because Internet data contains human biases such as racism and sexism
- Its blackbox nature because ChatGPT does not explain how it reached its answer, can inspire fears that AI has become sentient
- Plagiarism, because ChatGPT uses material available from the Internet, some of which may be copyrighted
- Privacy, who is using the information entered into ChatGPT?
These concerns, along with the fact that ChatGPT does not know a company’s proprietary information, have left many business leaders with the conclusion that, while ChatGPT is fascinating, it isn’t fit for use in their business. Right now, some may be perplexed about the potential business use cases because all they see is a text box on the OpenAI ChatGPT landing page. To be useful in a business context, the AI underlying ChatGPT must be embedded in the applications and workflows business people use every day, not a remote text box on a third-party website. When this happens, the potential efficiency gains will be massive.
The latest AI isn’t a tech fad that will pass. It’s going to be a revolutionary game-changer possibly as big as the internet, affecting all aspects of society in profound ways.
One of the first areas recent AI advances will impact greatly is customer service and support – the exchange of information between companies and their customers. In one survey of businesses, customer service was the highest area of AI deployment with 61% of companies focused on using AI for customer service.
Why prior generation customer service chatbots must evolve
Chatbots are a modern website staple, but to the disappointment of brands, customers often find them frustrating. Chatbots were first adopted to provide customers with faster issue resolution, but their lack of intelligence can make them a barrier to issue resolution. Chatbots were adopted with the goal of lowering costs through automation, but there have been some common problems.
For one thing, many chatbots aren’t really “intelligent,” because they don’t utilize AI. Instead, they’re deterministically programmed, to ask a fixed set of questions and produce a given output. Developers are provided a list of frequently occurring problems and the appropriate troubleshooting steps to program into the chatbot. However, since humans aren’t always predictable, they may ask questions outside the chatbot’s programming, which requires more programming. Customers may also state queries in ways that have not been anticipated, which also requires more programming. While this endless cycle of continuous improvement is admirable, it may not come soon enough for customers who are frantically trying to solve a problem.
Enter the latest AI. A large language model (LLM) like the AI underlying ChatGPT has been trained on an enormous volume of data, basically much of the information on the internet. Because it was trained on such a massive corpus of language, LLMs have the ability to understand and generate language unlike anything we have seen before.
Traditional chatbots attempt to understand the intent of a customer’s question and then return a static answer that the chatbot developer matched to the intent. No matter how a customer asks a question, once it is mapped to an intent, all customers get the same answer.
For example, if my daughter lost her credit card and I sought the advice of a chatbot, the best the chatbot would do was to provide the standard answer about how to replace a credit card. It would not provide the answer personalized to my situation with my daughter.
A next generation chatbot (we call it an Assistant) is different. It can tell me that my daughter should contact the credit card company and report the card stolen.
That’s a huge difference in capabilities. In fact, the delta between traditional chatbots and ChatGPT is so great, that the businesses which refuse to experiment with it now will find themselves at a competitive disadvantage later.
The truth is that no one understands all the ways that the AI underlying ChatGPT will be used just yet. What we do know is that the capabilities will be integrated into the everyday systems we use as consumers and professionals. In other words, people won’t be flocking to OpenAI’s webpage to see what ChatGPT can do. Instead, they will start reimagining jobs, workflows, applications and processes based upon LLM powered AI.
For customer service, this means delivering on the promise of automating more of the customer service functions and requiring human involvement in a decreasing number of instances. Already, some organizations are automating 95% of customer service inquiries while providing a better customer experience.
What ChatGPT means for your business
Customer service representatives are accustomed to referencing an operations playbook that tells them how to resolve customer issues, step-by-step. LLM powered AI will replace that by handling most cases automatically and then routing the outliers to agents.
Right now, people irrespective of their department or industry are trying to understand how ChatGPT will impact their jobs and their company, just based on the queries they’ve posted on the OpenAI website. They’re having trouble connecting the capabilities with the services they’re providing because ChatGPT basically looks like Google. One types a query into a text box and receives an answer.
However, within the context of a workflow or application, a LLM will be able to make humans more efficient at what they do. What companies need is the infrastructure to take advantage of the latest AI. This already exists today, at least in a customer service context.
Already applications like Quiq have incorporated LLMs into both customer facing and agent facing AI. From a customer perspective, LLMs provide an unmatched ability to understand a customer’s query and provide a personalized response. The best solution vendors have developed innovative ways to ensure that the answers the LLM provides are accurate and on brand. By restricting the LLM to use only a company’s internal knowledge and data the risks of using ChatGPT directly are overcome. Anecdotally, the quality of LLM powered assistants are at least twice as good, perhaps several times better than the chatbots we are used to.
The latest AI can also be used to make agents more efficient. The AI can understand the context of the conversation and provide the agent with a suggested response that is learned from similar conversations in the past. LLMs can also be used to turn tasks that previously required the development of advanced AI models into simple off the shelf features. For instance, editing responses to improve the quality or summarizing conversation transcripts are now trivial with a LLM.
Finally, customers will have the type of conversational experience they’ve been craving that helps them resolve issues faster. And from a customer service leader’s perspective, they now have AI tools at their disposal that are more powerful and easier to use than ever before..
Bottom line
ChatGPT isn’t fit for use in customer service, but the AI that makes ChatGPT work is going to have a transformative impact. Consumers are fed up with traditional chatbots that don’t help them problem-solve efficiently, and they’ll measure one company’s user experience with another’s – specifically, the one using the smarter technology – the one that is LLM based.
Now is the time to start piloting this technology because it could mean the difference between long-term competitiveness and being seen as “behind the times.” The infrastructure already exists to use LLMs in a customer service context, which is an easy way to get started and provides automated resolution capabilities unlike anything previously seen.
About the Author
Mike Myer is CEO and co-founder of Quiq. Quiq combines AI and human agents to increase efficiency, drive revenue, and improve customer satisfaction.
