How Generative AI Succeeds Where Chatbots Failed

The buzz about generative AI is everywhere right now. For those in the contact center space, it might sound a lot like the hype surrounding chatbots several years ago, when customer experience programs were rushing to implement the latest time- and money-saving tech. But where chatbots never truly lived up to their promise, generative AI is set to usher in a major business revolution. Gartner predicts that by 2026, 10% of current live contact center interactions will be automated using artificial intelligence—the holy grail that chatbots never found. But what makes generative AI different from traditional bots—different enough to create a fundamental change in the way contact centers operate?

The potential lies in pairing the natural language capabilities of a generative AI model like ChatGPT with an AI solution that can also learn and perform actions. Compared to chatbots, which simply present information available elsewhere before transferring more complex cases to an agent, a generative AI solution could facilitate an entire interaction with a customer—even a highly complex problem or an issue that is not well defined by the consumer.

Why GenAI Alone Isn’t Enough

Many businesses rushed into adding chatbots on their sites with the goal of cutting costs. They assumed bots would deflect calls from coming into the contact center, thereby decreasing labor expenses. But chatbots can’t actually do anything other than present basic information a customer could likely find elsewhere, so these companies ended up paying for chat responses that didn’t fully solve customer inquiries—the bots still had to pass customers to the agents to complete any necessary actions.

ChatGPT now offers powerful natural language processing and can comprehend even highly detailed plain-language questions, but even those well written responses are likely offering information that can be found elsewhere by customers—and it may even be completely wrong. ChatGPT is designed to write very compelling text, but accuracy is not guaranteed or even a priority of the model. Brands need to ensure that any information coming directly from their representatives is a source of truth. This is why you can’t just open up a ChatGPT window on a contact center agent’s desktop and leave them to it.

Without properly training the AI, putting in guardrails to prevent incorrect information and off-brand or inappropriate responses, and adding the ability to take action, ChatGPT doesn’t add much more value that the traditional chatbot—and can, in fact, present significant risk. Still, it’s an incredible breakthrough and companies are very quickly learning how to use it for its advantages, adding technology and process guardrails to harness the power of its natural language processing.

Creating the Contact Center Digital Worker with GenAI

When contact centers implemented chatbots, many imagined deflecting a significant amount of live interactions—say, 20% fewer calls answered by agents. However, even if they achieved that level of deflection, chatbot results would inevitably plateau. Now imagine, instead of deflection, making each live agent 20% more efficient. And imagine that efficiency improves over time, reaching 25% more efficient, 30%, 50%, and beyond. That’s what generative AI has the potential to do in the contact center.

The biggest limitation in machine learning and AI is data. GPT represents such a major breakthrough because of the sheer volume of annotated data used to train it. When you transfer that technology to the contact center space, where it can observe and learn from human agents, and pair it with intelligent automation, you get a digital worker that can deliver higher concurrency, shorter handle times, and a higher level of precision.

It’s not tomorrow’s technology. It is here today.

Guest blog post written by Laivly. To learn more about this topic and others, visit the events page to check out all of our upcoming events.