Conversational Commerce

In normal online shopping, one clicks through categories, filters and product pages. Conversational Commerce works differently: you simply write down what you’re looking for. This happens in a chat window, in a messaging app like WhatsApp, or by voice. The answer comes either from an employee or from a program that responds automatically. Advice, ordering and payment also take place within the same conversation. So the purchase is no longer a form, but a dialogue.

Why retailers are betting on the chat window

Many purchases fail not because of price, but because of an open question. Will these trousers really fit in size 32? Is the device compatible with my laptop? Anyone standing in a store simply asks. In a classic online shop, this is often where people abandon the process. A chat that responds immediately keeps these customers in the buying process.

Then there’s a cost argument. Customer service by phone is expensive because every inquiry ties up a person. An automated chat can carry on thousands of conversations at once. Companies therefore regularly report that they answer a large share of standard questions without staff. For follow-up questions, a human still remains reachable in the background.

Economically, there’s a third point of interest: reach. Messenger apps have billions of users worldwide, and many people open them daily. A company doesn’t need to push its own app there. It goes to where customers are already writing anyway. This is precisely why corporations like Meta are investing heavily in business features for WhatsApp.

From decision tree to language model

The early systems were simple rule-based machines. They recognized individual keywords and output a pre-written response accordingly. If the question deviated from the expected pattern, all that came back was: “I didn’t understand that.” These chatbots therefore had a rather poor reputation.

Today, large language models are usually behind them. These are programs that have learned from vast amounts of text how language is structured. They also understand unusually phrased questions and respond in full sentences. So that they don’t make things up freely, they’re linked to the retailer’s real data: product range, stock levels, prices, order status. The model formulates the answer, the facts come from the database.

Such a system therefore consists of several parts. A channel receives the message, such as WhatsApp or the chat window in the shop. The language model interprets the intent behind it. Interfaces to the inventory management system supply the concrete information and ultimately trigger the order and payment. In sensitive cases such as complaints, the system hands off to a human.

Between WhatsApp orders and voice assistants

In everyday life, one encounters Conversational Commerce more often than the unwieldy name suggests. The chat in the bottom right corner of a shop page is part of it. So is ordering pizza via WhatsApp, booking a hair salon appointment through messenger, or tracking a shipment by asking via message. Voice assistants like Alexa also fall under this category when you use them to reorder something.

In business news, the term mostly appears in connection with retail corporations and messenger platforms. In Asia, development is further along: in China, many people handle chat, payment and shopping entirely within one app like WeChat. European providers have been trying for years to build something similar.

A common misconception is that Conversational Commerce is simply a customer service chat. The difference lies in the conclusion: here, the conversation ends with a purchase, not with a link to the shop. Nevertheless, the topic is viewed critically. A sales conversation with a machine is hard to verify, and chat histories contain very personal data.