E-commerce is undergoing a silent yet profound transformation. For several months, I have observed how customers interact differently with shopping platforms. They are no longer just looking for products; they are asking AI assistants for advice, to compare offers, and even to finalize their purchases. According to the Fevad and KPMG study from September 2026, 31% of French online shoppers are already using generative AI in their online shopping journey. This figure rises to 73% among regular AI users. However, despite this growing adoption, a gap persists: consumers trust AI for advice but are still hesitant to entrust it with payment. Agentic commerce accounts for less than 1% of e-commerce transactions in the United States, where it is the most advanced. This situation reveals a fascinating paradox. AI can analyze thousands of products in seconds, identify the best options according to our preferences, and reassure us about our choices. But when it comes to spending money, we revert to our old human instincts. This article explores how agentic commerce is redefining e-commerce, why trust remains the main obstacle, and how e-merchants must prepare for this inevitable revolution.
đź“‹ Summary
Generative AI: A Trustworthy Advisor Above All
When I analyze the data from the Fevad and KPMG study, a clear pattern emerges: consumers primarily use generative AI to obtain advice before making a purchase. Among AI users in the buying process, 58% use it before the transaction for simple advice and product comparisons, with a relative trust of 47%. Customers ask questions, seek recommendations, and let AI guide them to the best options. This is a role that AI masters perfectly. 🤖 It can process hundreds of criteria simultaneously, analyze customer reviews, compare prices in real-time, and propose personalized solutions in seconds.
What is fascinating is that 57% of online shoppers who still refuse generative AI justify their choice with concerns about commercial neutrality. They wonder if AI truly recommends the best product or if it favors certain brands. This concern is legitimate. Recommendation algorithms can be biased, intentionally or not. But for those who accept AI, this advisory phase represents undeniable added value. AI becomes a personal assistant that understands our needs better than we do.
In France, e-commerce continues to grow by 7% in 2025, with an average basket of 62 euros. In this context, generative AI offers e-merchants a unique opportunity: to enhance customer experience without increasing support costs. AI-powered chatbots can answer thousands of questions simultaneously, 24/7. This is an economic revolution for small e-merchants who could not afford to maintain a large customer support team.
Agentic Payment: Where Trust Stops
Here lies the breaking point of agentic commerce: during the purchase, only 27% of AI users solicit it. And trust dramatically drops to 30%. This decline reveals something deep in our psychology. We accept that AI advises us, but we want to maintain control at the critical moment: the moment money leaves our account. Agentic payment accounts for less than 1% of e-commerce activity in the United States, according to Bernstein. Even in the country where this technology is most advanced, it remains marginal. Why? Because only 23% of consumers trust an agent to pay on their behalf, according to a Visa survey.
I understand this reluctance. Entrusting an AI with the decision to purchase a product is one thing. Giving it access to our payment methods is another. đź’ł The perceived risks are enormous: fraud, input errors, unintentional purchases, or worse, malicious use of our banking data. Payment companies are aware of this. They are building technological safeguards to secure these transactions. But technology alone is not enough. There must also be transparency, traceability, and clear accountability in case of issues.
In Europe, platforms remain cautious regarding the market. They test, measure, but do not deploy massively. This is a smart strategy. Better to move slowly with customer trust than to rush things and create a crisis of confidence. The European regulatory framework, particularly the Payment Services Directive (PSD2), imposes strict requirements for authentication and security. These constraints slow down innovation, but they also protect consumers.
The Three Steps of the Purchase Journey: Where AI Creates Value
The Fevad and KPMG study details precisely where AI intervenes in the customer journey. Before the purchase, it is advice (58% of users, trust 47%). During the purchase, it is assistance (27% of users, trust 30%). After the purchase, it is customer service (35% of users, trust 43%). This progression reveals an important truth: AI creates the most value when it helps without deciding. It excels at providing information, answering questions, and solving problems. It is less comfortable when it has to make engaging decisions.
The post-purchase phase is particularly interesting. đź“ž Customers use AI for customer service questions, after-sales service, or product usage. This is an area where AI can truly shine. It can access purchase history, understand context, and propose quick solutions. A trust level of 43% in this phase shows that customers are beginning to accept AI as a reliable service partner. This is a huge opportunity for e-merchants. Reducing customer support costs while improving customer satisfaction? That’s every CFO’s dream.
But there is an important nuance. AI works best when it is transparent about its limitations. If a customer asks a complex question, AI should acknowledge that it cannot answer and propose escalation to a human. This technological humility reinforces trust rather than undermining it. Customers accept the limits of AI. What they do not accept is when AI pretends to know when it does not.
The Roadmap 2026-2030: Industrializing Agentic Commerce
According to Marc Lolivier, General Delegate of Fevad, nearly one in two French people could resort to agentic commerce by 2030. This is a bold projection, but it reflects a real trend. E-merchants cannot ignore this evolution. They must prepare now. The study recommends a three-phase approach: first, map exposure to major AI agents (ChatGPT, Claude, Gemini, etc.). Next, measure journeys and potential risks for three months. Finally, move to selective industrialization through enriched catalogs over the next 6 to 12 months.
I see this approach as very pragmatic. Too many companies dive into AI without preparation. They create chatbots that give silly answers or worse, dangerous ones. Then they give up, convinced that AI is not ready. But AI is never “ready” on its own. It is the organization that must prepare. 🏗️ This means training teams, cleaning data, defining processes, and putting safeguards in place. It’s work, but it’s necessary. You can explore how AI and data science tools can accelerate this transformation.
Once in production, eight key performance indicators (KPIs) will determine the success or failure of the strategy: the share of agentic traffic, product visibility to agents, assisted or unassisted conversions, cost per automated contact, human recovery rate, agentic error rate, security incident rate, and customer satisfaction. These KPIs are not gadgets. They reflect operational reality. A high agentic error rate? That’s a red flag. A high human recovery rate? That means AI is not mature enough. These governance metrics allow e-merchants to steer their transformation with rigor.
Why E-Merchants Must Act Now
Agentic commerce is not a trend that will disappear. It is a structural transformation of the market. Consumers are getting used to AI. They use it to seek information, make decisions, and solve problems. Tomorrow, they will use it to buy. E-merchants who wait for the market to be “ready” risk falling behind. The competitive advantage will go to those who invest early, learn quickly, and adapt their model.
But there is another reason to act now: data. The sooner you start collecting data on how AI agents interact with your products, the sooner you can optimize your catalog. AI agents do not see products as humans do. They analyze them differently. They look for precise descriptions, complete specifications, and structured reviews. 📊 If your catalog is not optimized for agents, you will be invisible. It’s like having a non-responsive website in 2024. It’s commercial suicide. Consider implementing a e-commerce strategy that integrates AI now.
Finally, there is the issue of trust. Consumers trust brands that are transparent about their use of AI. If you use AI to enhance customer experience, say so. If you use AI to reduce costs, be honest. Customers respect honesty. What they hate is manipulation. Building a trust relationship with your customers around AI is an investment that will pay off for years.
Conclusion
Agentic commerce is inevitable. It is not a question of if, but when. The data shows it: 31% of French online shoppers are already using generative AI for their purchases. This figure will likely double by 2030. E-merchants who ignore this trend risk losing market share. But those who embrace it intelligently, with a clear strategy and a gradual approach, can turn this disruption into an opportunity. I am convinced that the coming years will see a consolidation of the market. Small e-merchants who do not invest in AI will be absorbed by the larger ones. The large ones who invest poorly will waste resources. But those who find the right balance between innovation and caution, between automation and humanity, will thrive.
AI is not a threat to online commerce. It is a tool that amplifies what already works: the ability to understand customers, to offer them the right products at the right time, and to serve them excellently. The e-merchants who succeed will be those who use AI to strengthen their relationship with customers, not to replace them. It is a subtle distinction, but it makes all the difference.
đź“ť In Brief
- 31% of French online shoppers are already using generative AI for their online purchases, but only 23% trust an agent to pay on their behalf
- AI creates the most value in the advisory phases (58% of users) and customer service (35%), less in payment (27%)
- Agentic commerce represents less than 1% of e-commerce transactions in the United States but could reach 50% in France by 2030
- E-merchants must map their exposure to AI agents, measure risks, and then gradually industrialize with 8 key tracking KPIs
