At the start of this year, Fevad, a French organization representing e-commerce, emphasized the growing role of agent-based commerce. This refers to the use of generative artificial intelligence (AI) tools, such as chatbots and virtual assistants, to help consumers during their online shopping journeys. According to Fevad, around 31% of online shoppers in France are already using these AI tools to assist with their purchases. However, the French Competition Authority is closely watching this trend, as it is largely controlled by a small number of major players. Measuring the impact of these tools, however, remains challenging due to the complexity of the data involved. Fevad revisited the topic at NRF Europe, an international retail conference, where it discussed the return on investment (ROI) of conversational AI tools in online sales. Some data points appear contradictory. For example, at Boulanger, a French electronics retailer, traffic generated by large AI models was capped at less than 1% of total traffic. However, the conversion rates—meaning the percentage of visitors who made a purchase—on these AI-driven visits were significantly higher, sometimes doubling, tripling, or even quadrupling. Fevad notes that while these figures are not representative of the entire market, they highlight the importance of focusing on the quality of traffic, not just its volume. The organization also admits that measuring the true impact of AI-driven commerce is still imperfect and evolving. Agent-based commerce includes a variety of AI tools, such as virtual assistants and automated agents, which differ in how much they can operate independently. KPMG, a global consulting firm, has identified three stages of maturity in these tools: the assistant, which helps users by comparing options and reducing confusion; the preparer, which helps configure a shopping basket by making decisions on behalf of the user; and the executor, which can take autonomous action after being given instructions. However, consumer behavior varies widely. While many use AI to save time—such as comparing products or gathering information—many remain hesitant when AI tools handle sensitive data or payment processes. Fevad emphasizes that any "black box" automation, where the process is unclear or uncontrolled, is not acceptable for brands. Trust remains crucial, and consumers expect humans to have final approval at key decision points. Looking ahead, Fevad suggests that by 2026, companies must ensure their product data is accessible to AI systems, allowing them to be read and understood by machine algorithms. As AI agents become the first point of contact for consumers, e-commerce platforms need to evolve to interact seamlessly with third-party AI models. This means shifting away from traditional search engine optimization (SEO) tactics focused on keywords and instead structuring data in a way that allows large language models to understand not just products but also the context, guarantees, and availability of offers. Fevad stresses that the strategic goal for 2026 is to map how brands are exposed to AI agents and structure their assets so they can be used by third-party systems without losing control.