Three Ways LLMs Will Revolutionize Customer Support

There has been a flurry of predictions and insights into how ChatGPT can be best applied to customer service. But so far, the vast majority of what we've seen from customer service tech vendors about how they plan on leveraging the technology has been vague and nebulous. When applied the right way, it will generally improve the delivery of customer experience across the board, from first response time to customer satisfaction and more, but we need more discussion around the practical applications of large language models (LLMs) in customer service. Below are three such use cases that showcase how ChatGPT can be leveraged to generate real value for both the company and the end customer.

One of the biggest areas of impact in CX will be handling complex inquiries, such as technical troubleshooting. Sometimes known as tier 2 support, technical troubleshooting has historically been the most difficult and expensive type of customer support inquiry. They can often take a long time to resolve (thus decreasing efficiency) and require hand holding by more technically experienced and expensive human agents.

However, recent innovations by LLMs such as ChatGPT have made it possible to automate a significant percentage of technical inquiries. An excellent example of this is in the consumer technology space. Say, for example, a customer calls in saying that his WiFi router is not working. Through vastly superior conversational intelligence compared to what was available even a year ago, the LLM can quickly identify the customer's technical expertise and adjust the conversation accordingly. It can automatically whittle down to the potential issue by scanning through knowledge base content and then provide the steps for the customer to resolve the issue in a conversational, friendly manner. All of this can be done in a fraction of the time compared to even the best tier 2 customer agents.

On top of this, LLMs will play a huge role in augmenting human agents for even more complex or troublesome inquiries, helping to automatically locate the relevant information in a large knowledge base and formulating a context-dependent and thoughtful response.

Another area where LLMs will have a profound impact on customer service is conversational commerce. Historically, customer service has been seen as a cost center because many organizations haven't been able to prove how great customer service directly and positively impacts the bottom line. This has been changing recently, as some of the more forward-looking and innovative customer service programs have found ways to leverage machine learning to perform acts such as recommending other relevant products that might interest a customer right within the context of a customer service chat. LLMs are bound to accelerate this trend.

New use cases will arise, such as shopping assistants, product recommendations, and helping companies upsell/cross-sell. The bots can ask relevant clarifying questions to refine customer needs and identify the most appropriate products. The bot will also be able to leverage existing information about each customer, such as prior purchase history, customer service interactions, and location, to generate more human-like rapport with each customer, which has also been proven to increase repeat purchase rate and customer lifetime value.

Finally, no conversation around the impact of AI and large language models on customer service is complete without discussing analytics. LLMs are excellent at summarizing, organizing, and prioritizing topics. When combined with other machine learning capabilities, AI will redefine what it means to generate valuable insights through data analysis.

Existing machine learning models within customer service can analyze every customer inquiry to make sense of the enormous number of inquiries and identify emerging topics based on content, source of friction, and customer dissatisfaction. When you add LLMs to this, the result will be not only the accumulation and analysis of such data, but then recommended next steps about what to do with the information, tailored for each internal audience. For example, say you're a fast food chain running a national coupon promotion. AI can process all customer reactions, identify trends, and understand potential emerging complaints as soon as they arise and can recommend actions to mitigate impact. Ultimately, LLMs dramatically accelerate the discovery of topics and tighten our feedback loop.

LLMs represent a step function improvement in how AI can positively impact the overall customer experience. However, as is often the case with ground-breaking technologies, it all depends on how it is applied. Technical support, conversational commerce, and analytics are three areas where immediate, profound impact can be generated, and I can't wait to see where else we can use AI to facilitate exceptional interactions between company and customer, every single time out.

Damien Thioulouse leads artificial intelligence and machine learning across all domains at Simplr.