The session opened with a striking figure: according to the NVIDIA 2026 State of AI in Retail and CPG study, the share of retailers using AI at scale—rather than just testing it—has risen from 42% to 58% in one year. Around the table were Marlies Hersbach, Chief Online & Customer Officer at Mango; Dorian Candavoine, eCommerce Director for EMEA at L'Occitane en Provence; and Bastian Sehnert, Head of Data & AI Initiative Office at MediaMarktSaturn. All three emphasized that the future of retail lies in adapting to the customer, not the other way around.
The first shared observation was that traditional keyword-based search is becoming outdated. Dorian Candavoine from L'Occitane put it plainly: "We ask the customer to understand our catalog, when it is the catalog that should understand the customer's needs." The brand is currently testing Google Cloud Vertex AI in A/B testing to simulate the experience of a beauty advisor helping a customer with dry skin. Meanwhile, at MediaMarkt, conversational search is already in use in Germany, where customers describe a problem instead of typing in product codes. Bastian Sehnert views this as part of a larger shift—from human-driven processes to agent-based systems that can handle customer interactions more efficiently.
The participants agreed that the era of AI proof-of-concept trials is ending. Bastian Sehnert framed this in financial terms: "A pilot without a scaling trajectory is a dead end. Creating a technically perfect agent that no one uses is an unnecessary cost." At Mango, this philosophy has already been implemented: AI agents now handle tasks such as styling advice and order tracking, allowing store managers to focus more on customer interaction and less on analyzing reports. At L'Occitane, AI has improved product stream quality, boosting impressions, click-through rates, and return on ad spend on Google Shopping. On the creative side, Marlies Hersbach noted a significant improvement in video production, which used to be too expensive to scale. Now, the creative team is integrated directly into operational workflows.
As an advisor on the AI Act, Bastian Sehnert sees regulation not as a hindrance, but as an opportunity for European companies to gain a competitive edge. MediaMarkt has created "fast lanes" for low-risk AI applications, giving teams autonomy while maintaining control over high-risk uses. This approach helps avoid costly compliance issues later in the development process. Looking ahead, all three participants pointed to conversational and agent-based commerce as the next major frontier. Dorian Candavoine hopes to test agent-based systems in real conditions soon, while Marlies Hersbach aims to launch a unified conversational commerce platform. Several French startups, such as Alpic, are already exploring how to connect product catalogs and payment systems directly with tools like ChatGPT and Claude. This signals a shift in the retail landscape: the next battleground will be within conversations, not on search engines.
Retail Leaders Discuss AI Integration and Future of Conversational Commerce
AI-rewritten from original reportingHow it works
ai-in-retailconversational-commerceagent-based-systemsai-actcustomer-centric-aiscaling-ai



