The possible applications of AI have recently multiplied thanks to the capabilities of ChatGPT launched by OpenAI. It is with enthusiasm that players in the world of data welcome the popularization of conversational AI, first appeared in 2018 with the launch of Google Duplex, explains Anne-Laure Blondeau, Senior Director Client Success, Southern Europe, at Yext, one of the leaders in online marketing and search.
The phenomenon is not new and AI has occupied an increasingly important place in our lives for years.. Simply, the general public generally does not observe these changes directly. Indeed, Most of the benefits of AI and machine learning only happen behind the scenes. Today, the rise of generative AI stimulates everyone's imagination more than ever and conversational AI opens new horizons. So, in recent months, many people have tried OpenAI's ChatGPT interface. From writing essays to writing a commercial for Ryan Reynolds, the platform is tested from all angles and elicits many reactions. Nothing surprising about that. But whether it inspired fantastic myths or horrific stories, this relationship with AI has always shared a common point : la conversation. Language is part of human DNA. From the age of seven months until the end of his life, human being speaks. What if our interlocutors were no longer just humans ? It is precisely this aspect of AI that excites so much. And the consequences are enormous.
L'IA : its potential and its limits
All digital experiences can potentially be transformed or enhanced by conversational AI. However, designing relevant digital experiences with conversational AI is a journey fraught with pitfalls. To be truly integrated into every digital experience, conversational AI must go beyond simple testing, and businesses must plan now how to maximize results on the customer journey. How AI can create relevant conversations ? How Businesses Can Take Advantage ? Conversational AI is only as valuable as the knowledge it possesses, as is the case with all machine learning applications. Here, poor quality of source information potentially leads to poor quality of responses provided. Any incorrect data, incomplete or obsolete input into the process can only produce distorted results at the end of the chain. In many ways, Early results from experiments with ChatGPT and other large language models show that the results provided by these models strangely reflect the query itself or the data on which the model is built.
Four categories of questions
This is why businesses should take a step back. Above all, you should not consider the conversational AI interface as the only solution to answer all the questions asked by your employees or customers.. There are two groups of questions asked in common conversations or searches. The first group consists of applications with or without mention of a brand, and the second, objective or subjective queries. When we combine them, these two groups produce four distinct categories : subjective questions that do not mention a brand, objective questions that do not mention a brand, subjective questions that mention a brand and objective questions that mention a brand. By segmenting the different types of questions into these four categories, businesses can quickly identify the opportunities offered by relevant, AI-powered conversations while limiting potential missteps.
Conversational AI can excel at objectivity
Whether the first three categories are common in broad searches or when people search for reviews or ratings, the last category (objective questions that mention a brand) is where conversational AI can excel. Conversational AI is limited in subjective questions and conversations, especially when humans ask how and why a particular opinion was expressed on a given subject. otherwise, most businesses rely on third-party review sites for subjective questions. Conversely, objective questions that mention a brand offer the best opportunities for businesses to maximize the impact of using conversational AI. These questions help achieve the best customer experience, but also require taking a different approach to data. The data needed to respond accurately and reliably must be collected from different parts of the business and stored so that conversational AI can easily access up-to-date information.
The impact of conversational AI merged with data
To gather and organize information, brands can use a content manager (knowledge graph) which powers hundreds of platforms such as Google, Bing, Siri and Alexa – these platforms are themselves already based on AI. This same approach could allow conversational AI systems like Chat-GTP to access the same data to answer objective questions that mention these same brands.
Start the AI conversation
The battle for the best conversational AI technology has only just begun.
Google, Bing, Amazon and even Netflix will all tomorrow be likely to integrate a conversational interface into their platform. That said, the question of AI ethics and data integrity will become a major issue in conversational AI. Companies that want to exploit conversational AI to improve their customer experience must now ask themselves the question of the management and organization of their data so that they comply with expectations in terms of protection of private data.. With conversational AI, brands will have the opportunity to interact with their consumers or collaborators directly rather than going through third-party platforms such as search engines. The interaction will then become more direct and will be powered by their own data associated with a conversational AI interface..
Powering AI systems
By first looking at the objective questions that mention a brand, businesses must develop a strategy to centralize and organize relevant data into a platform designed to power AI systems, while maintaining human control and supervision as to the exact origin of each response. Thanks to this approach, businesses will be able to prepare for this new user experience, by feeding each customer conversation with the right answers.
