Human-Assisted AI in E-Commerce over Mobile Messaging

ReplyYes is an e-commerce over mobile messaging platform for brands that want to engage and transact directly with their customers. ReplyYes sends personalized product and content recommendations. Our platform learns from customer feedback to make the experience unique to each person and brand. It has been architected to support endless numbers of stores, all utilizing lightweight artificial intelligent (AI), product recommendation algorithms, inventory and fulfilment management systems, and payment processing. As a strategy, we launched our own stores in categories where our AI and recommendation algorithm learnings can be leveraged into platform partner opportunities.

We started with a constrained AI and user education around the specific commands that can be interpreted by it. In the event that a phrase cannot be understood by the AI, we pass the customer message and information to a human who will both respond to the customer and train the AI on new keywords and phrases used by the customer; this input into the AI helps improve its ability to automatically respond to customer requests over time.

To that end, we view AI as a means to emulate and scale human interaction. Traditional brick and mortar stores restrict even the best humans to handling a single customer at a time. A huge advantage that the ReplyYes platform offers is the idea of empowering humans to talk to many more customers at once utilizing AI. In other words, we want our AI to be able to converse with customers over messaging as if they were talking with humans.

Content generation is a difficult process and prone to errors that can ultimately lead to a poor user experience. To improve our likelihood of success, we believe that a necessary first step towards an AI supported conversation are tools that suggest content that can be updated and approved by humans. The automatic generation of a first draft that is then checked and updated by a customer support agent or sales rep can improve their efficiency by an order of magnitude by enabling them to handle more customers at the same time  while minimal context switching. Thus, in a way, we are scaling people using AI.

Humans are essential to the ReplyYes platform experience. They create an atmosphere that is enjoyable to customers. As such, our intention is not to remove them from the store but to empower them to talk to more and more people while not decreasing their quality of service. It is our view that there will always be humans involved in our platform customer stores as it is not reasonable to have the AI handle every single type of conversation; the focus of our AI will be on the most frequent of these types. Like the recommendation algorithm, improvements and gains made to it are not coupled to specific product types. Any store powered by ReplyYes should be able to leverage advances that we make with the AI.