Why Conversational AI Is The Future Of Business And Customer Success

And it's all built on a knowledge graph.

By Christian J. Ward

Jan 26, 2023

6 min

Opportunities to leverage AI have recently skyrocketed, thanks to the sensational abilities of the recently-launched ChatGPT from OpenAI. For many of us in the data space, it's been exciting to see the promise of conversational AI — which we first glimpsed back in 2018 when Google demonstrated its Duplex system — accelerate into the mainstream.

But this isn't just a phenomenon of the last few weeks. AI has been affecting our lives more and more for years — but the difference is that people have not generally witnessed those changes directly. Instead, many of the benefits of AI and machine learning have existed behind the scenes.

For example, AI has been used in the healthcare industry for image analysis to improve diagnosis outcomes. In financial services, AI has long assisted in the identification of potentially fraudulent transactions. And most people know that AI helps improve their search results and personalized experiences online while also improving data privacy. But the launch of generative content through AI has now captured people's imagination in a way these prior developments have not.

Humans are built for conversation. Conversational AI is opening up new possibilities.

Over the last several weeks, huge numbers of people have tried out OpenAI's ChatGPT interface. From drafting love letters to helping pen a television commercial for Ryan Reynolds, the platform is being tested from all angles — and people have a lot of thoughts about it. That's not exactly surprising: ever since the first depiction of AI in a movie occurred in 1927 in Metropolis, the world has created a fictional relationship between humans and AI. It has been the fuel for both fantastic myth and brutal horror, but imaginings about the AI/human relationship have always shared a common theme: conversations.

Humans are wired for conversation. From as early as seven months, we begin our lifetime of conversation, with a journal of Science article in 2007 estimating that the average human speaks roughly 16,000 words per day. But what happens when the number of potential conversation partners moves beyond just humans? That's the AI tipping point that has people so animated. And it has big implications for where we go from here.

First: it means that there is no digital experience today that will not eventually be converted to or augmented by conversational AI. But, as we will see, the path to meaningful digital experiences that leverage conversational AI will be fraught with challenges and epic failures. Conversational AI needs more than a testing ground to be truly embedded into every digital experience — and companies need to plan now to maximize these capabilities in their customer journey.

How Can AI Create Meaningful Conversations? And How Can Businesses Take Advantage?

Conversational AI is only as valuable as the knowledge it possesses. As with all machine learning applications, there exists significant concern around bias, misinformation, and the "garbage-in, garbage-out" problem. To a large extent, initial findings from experiments with ChatGPT and other large language models (LLMs) demonstrate that the output provided by these models can be an uncanny valley reflection back of the query itself or the data it was trained upon.

For this reason, companies should take a step back. They shouldn't view a conversational AI interface as the solution to all types of questions from their employees or customers. But if companies segment the different types of questions asked by humans into four categories, they can immediately see where the opportunities exist for meaningful, AI-driven conversations while limiting the potential errors or missteps in their strategy.

There are two groupings of question types that are asked by humans in common conversations or searches. The first grouping is branded or unbranded questions, and the second is objective or subjective questions. These two types of questions, when combined, create four different categories of questions: unbranded objective, unbranded subjective, branded subjective, and lastly, branded objective.

While the first three categories are more common with broad searches or people seeking opinions or rating information, the last category, the branded-objective questions, offer the most likely place where conversational AI will be successful. Conversational AI will have problems with subjective questions and conversations, particularly because humans will question how and why a particular opinion was provided about any given topic. Additionally, most companies are at the mercy of third-party review sites for opinionated questions.

Now, consider the four types of questions below:

The top right quadrant — the branded objective question — offers the clearest opportunity for discussion through a conversational AI for most businesses. Objective questions directed to a business or company allow for the best experience — but they also require a different approach to data. The knowledge necessary to answer questions with direct, authoritative answers must be gathered from different parts of the enterprise and stored in a way that the conversational AI can easily access the information and any changes made.

Conversational AI Meets Yext Content

At Yext, we use a knowledge graph to gather, organize, and connect a business's objective facts to hundreds of platforms that need access to correct information. While we currently use it to feed data to Google, Bing, Siri, and Alexa (all, collectively, AI platforms themselves), it's easy to see how that same approach could work to ensure that conversational AI systems like ChatGPT will also have the correct data to use to when answering branded, objective questions.

Starting the Conversation (AI)

The battle for the best conversational AI technology has only just begun. Google, Bing, Amazon, and even Netflix are all likely to have you speaking in a fluid conversation with their platforms shortly. That said, the issue of ethical AI and data integrity will rise up as the most important issue conversational AI will face, similar to massive technological platforms of the past. For businesses to leverage the promise of conversational AI, they must take a proactive approach to managing their data — and their corporate knowledge — in a manner that is controlled, transparent, and compliant.

That's something we're pretty well versed in here at Yext. For over a decade, we've worked with businesses to ensure that every fact and piece of knowledge that a customer or employee might need to know is both accurate and present in platforms, no matter where that person looks. With conversational AI, the opportunity for brands to speak to consumers or their employees will ultimately move beyond the indirect world of updating third-party platforms like search engines, to a more direct interaction, powered by their own knowledge graph that is combined with a conversational AI interface.

Beginning with branded objective questions, every business needs to develop a strategy of centralizing and organizing critical knowledge in a platform designed to feed AI systems answers while maintaining human control and oversight over exactly where each answer originated. By leveraging knowledge graphs, your business can prepare for this incredible new user experience, powering every customer conversation with the correct answers.

Learn more about AI and the future of digital experiences with Yext's new virtual event series and join the waitlist for Yext Chat at yext.com/platform/chat.

Introducing Yext Chat: Conversational AI for the Enterprise

Spend your limited time on strategic initiatives, not conversations that could be automated.

Share this Article

Read Next

loading icon