Natural Language Processing (Shopping and Messaging)

Conversational shopping, or conversational commerce, seems to have emerged as a concept in 2015. The AI bots behind conversational commerce interfaces act as agents on the other side of a messenger conversation with a consumer. By asking questions to the bot, a consumer can receive personalized recommendations, product care instructions, customer service, and even purchase products in one click.

The term conversational commerce has been attributed to technologist

Chris Messina who described it as follows:

Conversational Commerce pertains to utilizing chat, messag-ing, or other natural language interfaces (i.e., voice) to inter-act with people, brands, or services and bots that heretofore have had no real place in the bidirectional, asynchronous messaging context —Chris Messina, founder, Molly.

A conversational interface does not necessarily include only messaging. It also encompasses buttons, web views, images, and other simplified graphical user interface (GUI) components. These components can help guide the conversation between the human user and the machine by providing possible outcomes to the specific context.

To mimic the experience of talking with an in-store sales associate, companies like Levi Strauss & Co. have partnered with AI companies like In late 2017, Levi’s and released a conversational commerce bot that helps consumers discover their perfect jeans. This particular example uses a mixed UI relying not only on the user’s improvised inputs, but allowing the user to select common options by clicking preexisting buttons.

Natural Language Processing

Personalized Shopping Experiences

Part of what makes conversational commerce so enticing is the ability to create customized experiences for users. Particularly for younger generations, personalized experiences are predicted to dramatically shift online purchasing. According to McKinsey, “Personalization can deliver five to eight times the ROI on marketing spend, and can lift sales by 10% or more.”

AI agents learn as a consumer interacts with them, making suggestions for products and actions based on products they’ve liked. At times, these agents even adapt the GUI based on click-rate. For example, UI components like buttons may be revealed. As users interact with the bot, data is collected about engagement with each UI component. These components can be adjusted in real time, depending on their success with customers.

Bot-to-Bot Interaction


For brands, the proposition of hosting bots that interact with other bots makes revenue sharing easier. As Karen Ouk from has pointed out, a T-shirt brand bot might recommend pants from a pants brand bot, and this cross-recommendation can help both brands expand their reach and grow.

The idea of bots having business relationships with one another might sound absurd now. As these bots begin generating real revenue for each other, it may be difficult to imagine a time when brand partnerships were marked by sweepstakes and pop-up shops.

Context-Based Decision Making

An AI-based agent has the opportunity to not only provide product-­ based recommendations but also tailor recommendations to the user’s context. Examples of the user’s context include location, language, and demographic data. An agent should make different recommendations for winter clothes in New York than for winter clothes in San Francisco.

In the future, the context for these AI-based conversations will become more correlated with the user. It could sync with user calendars, suggesting outfits for the holiday party they’re going to attend tomorrow or the board meeting they have on Friday, providing a full-blown comprehensive fashion assistant.

Live Chat

In 2017, the National Retail Federation’s Omnichannel Retail Index called live chat one of the fastest growing areas of omnichannel retail. According to the report, 54% of retailers have implemented live chat on their web sites. Many businesses incorporate live chat on their web sites in order to answer in-the-moment customer service questions, send order information, and more.

For consumers, talking to a sales associate in a shop is a way of getting a second opinion. The ability to interact with customers while they’re shopping online has been historically limited. The moments in which the consumer is making a purchase decision are critical to conversion rates.

Live chat, or sales chat, is just one way that retailers are using conversational interfaces to drive consumer decision-making. Live chat interfaces connect consumers with a human, an AI bot, or a hybrid solution. Historically, live chat meant talking to a human through a chat interface embedded on a brand’s web site. By using AI bots, these one-to-one services are much more scalable. The distinction between live chat and conversational commerce bots will likely disappear as the implementations of these services become more streamlined.

If you’re a smaller brand like Betty & Ruth, you might think that AI bots are out of reach. It’s normal in the fashion industry that we don’t have things like an application programming interface (API) that other software companies can use to access our product data.

Fortunately, the companies working on conversational commerce have recognized that fashion brands don’t always have tech resources. As a solution, for brands that agree to participate, they’ve built site crawlers that will pull product information from the brand web site for use in their AI-bots. Preparing a web site for basic integration doesn’t require brands to implement any new technology.

The downside is that most of the out-of-the-box solutions can’t address all of the needs we have as a brand at Betty & Ruth. With the current state of things, using third-party apps, we would need more than one specialized chatbot to address product discovery and customer support.

There are other bots we are looking into integrating that are more general-­ purpose bots and focusing on chat features like cart recovery, which addresses a pain point we have: abandoned carts.