Meet our special speaker and AI specialist Dawn Procopio. Dawn has been working as an expert on interaction and AI for numerous notable clients such as Google, Meta and Twitter (now X) and explores user experience and AI ethics. This is a summary of our latest webinar addressing how companies can use AI to their advantage and what we should think about when choosing to work with an AI conversational model in our sales operations.
What role can AI play in an omnichannel customer journey context?
Understanding the persona factor
Dawn Procopio, AI specialist, believes that given that there are so many different types of shopping personas, one of the keys that a store associate has is the the agility and the sense of context to know which persona themselves to take on based on who they’re talking to. When we are dealing with a customer online, the Maestro AI™ tool works in a similar sense. It will use Large Language Models along with its compiled data to find which persona it is interacting with and which language to adopt for that particular context of interaction. It is able to do so, because it has been trained using 10 years of empirical data, allowing it to know how to communicate with shoppers and how to ask them the right questions for a specific personae. To summarize, it allows it to contextualize. She explains that the personas come from, the various motivations and pain points that happen in each context.
“The usual issue with Large Language Models is that we actually designed it to have many, many personas and many, many personalities all at once. And so there’s a little bit of a problem where we’ve only been talking, for instance, to one person at a time. Typically that’s the first mental model. And then maybe you have some experience talking to many people at the same time. But, you don’t have the experience of talking to people, in different contexts at the same time. Right? You don’t typically have the ability to talk to several different customers at several different sites, all at the same time, and updating your own training or learning mechanisms, at the same time.”
Using AI as a sales oriented communication tool
Dawn explains that the new challenge for people wanting to use AI as a communication tool in their retail operations, is that it is a model that has been trained often to speak very 1 to 1. It needs to understand the nuances specific to the brand that maybe it hasn’t been trained on.
She suggests that retailers looking to implement AI in their sales process are going to have to think about how they can optimize their tool from using a generalized Large Language Model into the tuning phase. In the case of Salesfloor, this critical tuning phase has been trained on extensive data specific to the apparel industry, it is that critical model training that really makes it perfectly tailored for selling retail.
Building trust and authenticity
In this intricate system, how do we reconcile the challenge of catering to a targeted personae with a specifically tuned Language model without compromising trust and authenticity in our interactions with customers?
When it comes to delivering a successful experience to customers using an AI tool, you have to test it first in numerous realistic scenarios. That is an unfortunately very expensive step, because in order to deliver an authentic experience, it requires extensive testing of all the specific possible interactions you might encounter, in a very realistic manner. This also means being able to get people who interact with the model to try to behave naturally.
To understand this learning curve concept, Dawn believes we have to think of the human-like mechanisms:
“First, we have to look at the human or the base reality version of what you’re trying to innovate on. Let’s forget the algorithms for a minute. When we are talking about the social experience, how would a normal sales associate with an earpiece, learning about different things happening on other sales floors at the same time, talk to their own sales associate in the moment? Now imagine if that sales associate is also a person who’s never seen before and they’ve only had access to hearing, they speak the language, but they don’t have an intrinsic sense of the visual experience.”
Dawn explains that when you’re trying to simulate these conversations, you would have to bring it down to a very specific thing. This would be called the scenario design or mechanism design in game theory. She compares it to playing the role of a director in a sense, where you have to create a scene. She argues, that we cannot think about those mechanism building in a theoretical standpoint. “ It would be like standing from your 100 foot high helicopter and looking at the traffic patterns and wondering, why are people crashing here?”
To summarize, authenticity will emerge from the amount and the quality of the scenarios created in order to allow a successful learning curve for the model, because the personas will then emerge from this experiment.
