Agentic Commerce: how AI is transforming commerce?
Chaired by Grégoire TOUSSAINT (Edgar Dunn & Company), this session brought together Nicolas MORDAUNT-CROOK (Bird & Bird), Benjamin DÉSERT (Stripe) and Charlotte PEUTIN (Chronodrive).
The speakers shared their vision of agentic commerce, a model in which artificial intelligence agents assist, or even carry out, certain stages of the purchasing journey, while raising new technological, operational and regulatory challenges.
Artificial Intelligence and Payments: AI is Already Widely Integrated into Secure Payments
The discussion first highlighted a shared observation: artificial intelligence is already deeply integrated into payments, even if it often remains invisible to users.
Benjamin DÉSERT recalled that Stripe has been using AI for several years to improve transaction performance and payment security: “AI is used across the entire lifecycle of a payment transaction,” he explained. The company published a guide to agentic commerce.
In practical terms, AI notably helps to combat fraud, reduce unjustified payment refusals and simplify onboarding for merchants onboarding.
Stripe recently deployed a language model dedicated to payments, capable of analysing a transaction as a whole, taking context and behavior into account. This approach has proved particularly effective against card testing, a type of fraud consisting of automatically testing thousands of stolen card numbers with very small amounts. “We have moved from 60% to 97% of attempts blocked,” specified Benjamin DÉSERT.
This means that the vast majority of these fraudulent attempts are now detected and stopped before a payment is even authorised. These technical advances form an essential foundation for the development of agentic commerce, in which online payments must be seamless, secure and largely automated.
Agentic Commerce from the Retailers' Perspective
For retailers, agentic commerce represents a major evolution in the customer relationship.
Charlotte PEUTIN explained that this topic has become strategic for distributors, particularly in light of rapidly evolving user habits such as online shopping: “AI is already in our customers’ pockets, and younger generations use it on a daily basis,” she emphasised.
At Chronodrive, reflection is focused on several scenarios: developing its own agents integrated into the brand’s ecosystem and anticipating the arrival of third-party agents used directly by consumers. The agent can compare, recommend and purchase on behalf of the customer, according to their preferences.
The first use cases focused on customer service, particularly after delivery. Chronodrive has therefore deployed an agent capable of answering simple customer queries.
In the long term, the ambition is to go further, towards more natural and personalised interactions. “Customers no longer want to search using keywords; they want to have a dialogue,” summarised Charlotte PEUTIN.
This development is pushing retailers to rethink their role. They no longer fully control the journey. They must make their offers understandable and accessible to the agents used by customers.
Commerce and Artificial Intelligence: Framing Agentic Commerce, a Legal and Strategic Challenge
In response to these new practices, the regulatory framework remains partly to be built.
Nicolas MORDAUNT-CROOK recalled that current rules were designed for human interactions, not for autonomous agents: “Our legal frameworks are still based on human individuals, particularly in matters of consent and liability,” he explained.
Texts such as the GDPR, PSD2, the forthcoming PSD3 and the AI Act provide elements of response, but areas of uncertainty remain. One of the major challenges concerns the way in which a user authorises an agent to act on their behalf, particularly to make a payment.
These regulatory issues are closely linked to the technical challenges mentioned by Benjamin DÉSERT. In agentic commerce, payment is only a final step. Agents must first be able to correctly understand product catalogues and access reliable data: “Agents do not browse a website like a human; they query data directly,” he recalled.
For European players, the lead taken by the United States represents both a challenge and an opportunity. The opportunity to structure a secure, compliant model of agentic commerce adapted to the specific characteristics of the European market.
Agentic commerce illustrates a new phase of maturity in digital commerce, in which artificial intelligence becomes an active intermediary between the consumer, the merchant and the payment.
The discussions showed that this transformation is already underway, driven by concrete use cases and significant technological advances. However, its widespread adoption will require a balance between innovation, security and a clear regulatory framework.
As AI agents establish themselves as new entry points to commerce, the ability of stakeholders to make their offers clear, reliable and responsible will become a key differentiating factor.