The new era of the customer journey with agents

The era of intelligent agents has already begun to transform digital commerce: conversational algorithms capable of representing our preferences are now becoming the true "shoppers" while we go about our business. This delegation of tasks signals the end of a customer journey based on clicking and scrolling; it ushers in a continuous purchasing funnel, driven from end to end by artificial intelligence that identifies desire, negotiates terms, executes the transaction, and then verifies delivery compliance. Whereas traditional e-commerce relied on promotional frenzy to capture attention, the agentic approach promises a silent, proactive commerce centered on perceived value.

In the short term, the most dramatic impact for consumers is the reduction in "decision fatigue." A personal assistant can, for example, detect a weekend offer that meets its user's budget and ecological requirements, discreetly compare available places, book the highest-rated room, and then confirm the choice by notification. According to the projections presented on page 1, nearly half of the users surveyed already say they are ready to entrust their search to an AI rather than a traditional search engine, and a third still prefer the traditional method; the inflection point is approaching.

For retailers, the upheaval promises to be even more far-reaching. Value will no longer be based solely on product differentiation, but rather on the ability to make offerings "discoverable" by third-party agents, to describe them according to standardized ontologies, and to propose machine-to-machine negotiable contractual terms. Current SEO strategies will need to evolve toward semantic visibility protocols: the challenge is no longer to please the human eye, but to appear in the API calls made by a digital advisor looking for the best solution for their virtual wallet holder.

This transition relies on an emerging technological infrastructure built around two major protocols: the Model Context Protocol (MCP), which standardizes how an AI model exchanges its objectives, context, and input data with the platform; and the Agent-to-Agent Protocol (A2A), designed to allow agents to negotiate, collaborate, and execute complex tasks without human intervention. Standardizing these exchanges opens the way to much more fluid cross-platform value chains, capable of managing, for example, the assembly of a trip including transportation, accommodation, and event ticketing—each component being optimized in real time.

A new ecosystem of players is therefore emerging around three categories. The "cores" include agent platforms, payment infrastructure providers, and conversational search engines. Then come the "adapters": marketplaces, identification solutions, compliance and security tools, not to mention logistics and fulfillment APIs. Finally, the "enablers" provide the analytical, data, and compliance building blocks essential to making everything reliable. A detailed illustration on page 10 maps these roles and highlights the critical mass that will need to be reached to survive.

This new situation also reshuffles the cards in monetization. Where once people relied on direct gross margin or display advertising, agent designers now benefit from a broader range: micro-segmented commissions, dynamic brokerage fees, value-added services (insurance, guarantees, cashback), and even arbitrage between loyalty programs. Some payment giants are already experimenting with invisible "wallets" where biometric authentication, coupled with advanced tokenization, makes the checkout process entirely transparent. However, this fluidity opens the door to new risks: automated fraud, loss of brand identity, or even the disappearance of direct customer contact—all challenges that require robust algorithmic governance.

Trust then becomes the cornerstone of agentic commerce. A five-dimensional matrix, presented on page 18, reminds us that perceived security is not limited to data encryption: it includes clarity of intentions, model governance, system resilience, and the ability for the user to regain control at any time. There are many potential problems: systemic bias, runaway autonomous loops, or decisions made based on erroneous information. Companies that succeed in establishing explainable transparency—and communicating it—will be best equipped to convert curiosity into mass adoption.

The size of the opportunity justifies this effort. Forecasts indicate that in the United States, personal agents could orchestrate up to a trillion dollars of transactions by the end of the decade, redirecting a substantial portion of retail commerce to circuits where decisions are made by algorithms. For brands, the cost of inaction will quickly outweigh the cost of experimentation: not being able to integrate into the protocol economy means leaving the field open to more agile competitors who will make their catalogs readable and actionable by digital advisors.

Concretely, the roadmap for a company wanting to embrace this wave begins with an inventory of its product data. Technical attributes, stock conditions, dynamic pricing parameters, and logistics performance indicators must be aligned with a structured schema, readable by public or semi-public APIs. Next comes the modularization of architectures: decoupling the front and back offices, containerizing pricing functions, and exposing microservices for personalization or fraud detection, in order to be able to interface with requests from external agents. Finally, the company must invest in an explicit ethical framework: model audit charters, governance committees, robustness tests, and systematic training on automation risks.

The benefits aren't limited to e-commerce. In physical stores, agents embedded in mobile devices can reduce wait times, orchestrate click-and-collect, or optimize back-of-house inventory management. In warehouses, collaborative AI adjusts picking flows based on orders anticipated by digital advisors, smoothing out peaks in activity and reducing the need for buffer storage space. The agent thus becomes the conductor of a predictive supply chain where humans focus on high-value exceptions.

In terms of customer relations, the shift from push marketing to agent-to-agent dialogue is changing the very notion of loyalty. Users are no longer focused on a brand but on the relevance of the recommendation provided by their assistant. New metrics are emerging: "agent selection" rates, basket share captured without human intervention, and even machine-to-machine "average negotiation time." Initial field feedback shows that agent-optimized SEO can boost conversion, particularly for repeat purchases (supplies, everyday consumer goods), where algorithmic logic excels at balancing quality and price.

The flip side of the coin affects historical ad revenue. As AI filters out upstream signals, paid impressions risk eroding. Media companies must imagine new formats—structured content, contextual free trials, dynamic bundles—that agents can analyze and select as viable options. Creativity will need to shift to information engineering, combining rich semantics and transparent financial incentives so that the assistant judges the offer attractive without the user feeling manipulated.

On the regulatory front, the cross-border circulation of purchasing data organized by agents already raises questions of sovereignty: who is responsible for a selection error leading to an inappropriate medical decision? How can the portability of histories be guaranteed when several intelligences succeed one another in the service of the same individual? Legislators are moving towards safeguards: traceability standards, shared accountability gateways, and unalterable logging requirements. Companies that anticipate these obligations—by adopting “Safe-Audit APIs” and testing their emergency shutdown protocols—will build a decisive defensive advantage.

Beyond respecting these frameworks, the major challenge is aligning AI with the user's actual preferences. The first deployments of vector profiles, capable of representing tastes, constraints, and life goals, bring us closer to truly hyper-personalized commerce. Ultimately, we can imagine an agent optimizing not only a purchase, but the entire use of a product: a vehicle maintenance plan, a preventive approach to a suitable diet, or the selection of complementary services, all orchestrated from a holistic perspective.

For startups and investors, now is the perfect time to position themselves in underserved segments: semantic catalog enrichment, algorithmic trust certification tools, trading API marketplaces, or user preference aggregation services. Demand is growing rapidly in these niches, and barriers to entry are shifting toward obtaining partner data rather than physical capital. It's therefore important to build alliances now, before standardization locks in dominant positions.

In conclusion, the rise of intelligent agents is reinventing the grammar of commerce. More than a technical advance, it represents a massive redistribution of value: from interface mastery to protocol mastery, from human persuasion to algorithmic negotiation. Organizations that embrace transparency, interoperability, and software modularity will not only gain a competitive advantage, but will also actively participate in shaping a future in which the fluidity of transactions frees up time, reduces decision-making waste, and redefines the relationship between the customer, the brand, and the machine.

To go further: the passage from individual productivity to the orchestration of agents, how The rise of AI agents is revolutionizing business and how unlocking business value through agents.