Call Summarization and Infrequently Asked Questions bring real value today
There is so much pushback and hullabaloo these days related to generative AI and how it does not appear to be paying off yet. It hasn’t been two years since the launch of ChatGPT and already people are asking, “what have you done for me lately?” In the contact center, the answer is plenty. There are already two significant use cases where generative AI is making a big difference.
Use case #1: Automated call summarization:
Historically, agents had to summarize each interaction and provide a disposition code for its outcome. Yuck! Writing notes is thankless, and there is always pressure on agents to get to the next customer. The result is exactly what you would expect – horrible notes that help no one.
Deterministic AI-driven solutions for call summarization helped a bit, but overall, the effort was high, and benefits marginal. Enter generative AI: easy to install and able to deliver clear summaries that are better than what the agents produced. AI does the donkey work, and everybody wins. Call summarization is the hot ticket add-on for contact centers in 2024: interest is high, and sales are significant. This is a great first step into the genAI waters for any customer service team – especially as it doesn’t require significant operational changes in the contact center.
Granted, this is just a little point solution, but allowing a computer to easily and effectively handle tasks that millions of agents dislike will save brands billions of dollars annually and result in more useful call summaries.
Use case #2: Infrequently asked questions:
We’re all used to frequently asked questions applications. Someone anticipates what people will ask and provides answers either online or via a chatbot. Sometimes these applications are helpful tools for customers and save brands money.
With generative AI you can use something called retrieval augmented generation (RAG) to tell a system that it can ONLY use answers from a specific data set such as pile of pdfs, webpages, or maybe a knowledge base. With genAI, you no longer need to anticipate all the questions that customers may have. You can just let genAI learn your documents and share what it finds. This is what I mean by infrequently asked questions. If you don’t have to anticipate questions, the door opens to a world where you can answer many more of them. If you use Google, you are familiar with the user experience for this application. When you ask a question online these days Google often provides a highlighted paragraph from a document that the software thinks is the answer. Quite often the software is right.
Infrequently asked questions will provide a customer experience very similar to this search capability. Imagine asking how to use a feature on your microwave oven and having generative AI summarize the answer straight from a product manual and then pointing to that doc. This means that brands don’t need to limit their applications to answer only pre-trained frequently asked questions. If the answer is in a set of documents, all the brand needs to do is point its bot at the documents. Generative AI will find and share the answer. If the right information is there, the customer will likely get their answer.
The conversational AI vendors from my recent have built frameworks on top of RAG to provide additional guardrails against hallucinations. These conversations are typically anonymous, assuaging privacy concerns, helping infrequently asked questions gain traction. Back in January, many of the vendors from this Wave shared reference customers who had put infrequently asked questions into production, and since then I am seeing many more brands getting in on the action. This is the next use case that can save billions of dollars.
Please note, the risk of putting a generative AI-powered chatbot in front of your customers is still real. Hallucinations happen, and data needs to be secured. Be sure you fully understand the risks inherent in generative AI for customer-facing uses and look for a quality conversational AI vendor.
I can’t speak to the use of generative AI writ large, but within two years of the launch of ChatGPT in November of 2022 we have two separate use cases that show us where generative AI will provide better customer service experiences for customers and will save brands billions of dollars annually.
