Margus Eha: AI Can Put Your Online Store on the List – But That Alone Is Already Yesterday’s News

In recent years, online retailers have focused on making sure their products can be found in ChatGPT and other AI services. But according to Margus Eha, CEO of Opus, a developer of large-scale e-commerce platforms, the next phase has already begun: online stores must increasingly think about how to provide AI with information on exactly which products are in stock, when a specific item could reach a particular customer, and, if necessary, even which shelf it is located on in a physical store.

“People’s expectations for AI-assisted shopping have already changed so significantly that it is no longer enough for ChatGPT simply to say that Muri Market sells a yellow dog leash. To make a purchasing decision, the customer needs to know whether the right model is currently in stock, what it costs, when it will be delivered, and whether that specific customer is eligible for a loyalty discount – or whether another product might suit them better. And if one online store provides that information directly in the chat window, while another requires the customer to browse a product page, it is obvious which store the customer is more likely to buy from,” he said.

This means that, in order to gain a competitive advantage today, online retailers need to take the next step. At the same time, Eha says, taking that step has become easier than it may first appear: solutions already exist that allow AI systems to query information from an online store and display it in conversations. The most common of these is MCP.

AI Is Starting to Act Like a Clone

Behind the technical name lies a simple principle: AI can ask an online store for the information it needs at that moment to serve the customer. For example, instead of relying on a price it previously found online, the AI can request the current price and stock availability. It can also check delivery times, find suitable alternatives, or, in the case of a physical store, even determine where the product is located – all based on the pricing and delivery rules that apply to that specific customer’s contract.

According to Eha, the difference becomes especially clear with more complex purchases. For example, a customer of a home improvement store could give AI the dimensions of their terrace and ask it to select suitable decking boards and fasteners, calculate the required quantities, and check whether everything is immediately available. In a B2B online store, a customer could ask AI to repeat a previous order and automatically find replacements for products that are no longer in stock.

“For the customer, this means they no longer have to browse dozens of products, set filters, and piece together information from different pages themselves. For the online store, it means an opportunity to reach the purchase decision faster. AI can bring the customer to the shopping cart before the person has even opened the store’s homepage,” Eha explained. “In essence, this allows AI to behave like a human: to browse information, ask for more details, and sort out what matters most.”

New Online Stores Should Be Built with AI in Mind from the Start

Eha believes that large online retailers should already be actively assessing the opportunities AI can offer. This is especially important for new, custom-built e-commerce platforms, which may remain in use for years and are often connected to several of the company’s other systems.

According to him, product search, pricing, stock availability, and other business logic can be built in a way that allows them to be used later outside the online store’s own user interface. If this is only considered after a new sales channel has emerged, making the necessary information available may require a substantial number of integrations to be rebuilt.

The expert also notes that companies should understand that AI can act as a rather unforgiving form of quality control. In a traditional online store, incomplete product information may remain a problem limited to a single product page. But AI will begin using the same data to answer customer questions, compare products, and make recommendations. If a competitor’s data is more accurate and their systems can provide a definitive answer, it will be easier for AI to recommend their product.

“The first wave of AI made online stores ask whether ChatGPT could find their products. Now they need to ask whether AI can get a good enough answer from their store to recommend their product to a customer. In a few years, that may be as basic a question for an online store as asking today whether Google can find its products,” Eha said.