Key insights for brands in the agentic era

In Europe, a growing share of purchasing decisions is already being filtered by AI agents before users even click on a site or open an app. Brands are no longer competing solely for human attention, but to be retained -- or discarded -- by systems that compare, synthesize, and explain options on behalf of the consumer.

This shift toward agentic commerce is no longer a matter of foresight: it is measurable in the daily habits of European consumers and is beginning to reshape the growth levers available to brands.

1. The first customer in the funnel is an agent, not a human

Before a European consumer buys a product, an AI tool may have already narrowed the selection down to three options and explained why certain references were eliminated. AI is thus becoming the primary interface for discovery, comparison, and recommendation, at the exact moment when preferences are formed and critical trade-offs are made.

McKinsey estimates that by 2030, agentic commerce could orchestrate between $3,000 and $5,000 billion in transactions globally, influencing discovery, decision-making, and progressively, purchase execution. In Europe, even a moderate adoption scenario would have a material impact on retail and consumer goods sectors.

For marketing and e-commerce leaders, the strategic question is shifting: it is no longer just about "converting" a visitor, but about remaining visible and convincing when the first "customer" in the funnel is an AI agent. In this context, visibility is no longer won solely through placement, advertising, or UX, but through an offering's ability to be found, compared, and defended within the agent's logic.

2. AI is already the default evaluation layer in Europe

The latest McKinsey data shows that 84% of consumers in France, Germany, and the United Kingdom report using AI tools in their daily lives, well beyond simple experimentation. 38% already use them to search for products or services and decide what to buy.

In the purchase journey, AI shows up primarily "upstream" of the transaction, at the stage where preferences are formed and options are narrowed. The most common uses are: comparing brands, models, prices, and reviews (63%), learning about a category or product (55%), and discovering new products or finding inspiration (46%).

In other words, AI already functions as a generic evaluation layer, not as a gimmick confined to a few "tech" verticals. Usage remains moderately differentiated across categories: most fall within a narrow band of 30 to 36% adoption for decision support, with a slight over-representation in more complex areas (fashion, health, electronics, travel).

One case stands out: eyewear, which shows an adoption rate of approximately 13%, significantly lower than the rest. Here, perceived value relies heavily on physical try-on, in-store evaluation, and infrequent decisions -- areas where current tools offer little marginal value compared to the offline experience.

For brands, the lesson is clear: near-term opportunity depends less on the category label than on AI's ability to create value in the evaluation process without requiring physical validation.

3. Winning the battle for "machine legibility": Agent Engine Optimization

As discovery and comparison migrate toward conversational interfaces, competitive advantage increasingly depends on brands' ability to be accessible and interpretable by agents, and not just visible to human consumers. McKinsey refers to "machine legibility": structured credibility is progressively replacing simple presence in search engine results. This is what we call SEO & GEO optimization.

We are seeing the emergence of the equivalent of a new SEO: Agent Engine Optimization.

Three workstreams are becoming priorities:

• Strengthening the quality and structure of product data: rich metadata, standardized attributes, stable taxonomy, explicit descriptions of uses, limitations, and trade-offs.
• Aligning the brand promise with objectifiable evidence: reviews, studies, independent tests, expert mentions, which serve as fuel for the syntheses produced by agents.
• Maintaining strong technical foundations: robust PIM, quality feeds, consistent naming conventions, API exposure, to feed agents in real time as they query the brand's ecosystem. Busony supports e-commerce merchants in this upgrade.

This work is not a "nice to have": retailers that do not invest in their agentic surface risk becoming invisible upstream, before the consumer ever reaches their proprietary interfaces. A SEO & GEO 360° audit helps identify these blind spots.

4. Optimize for explainability, not just conversion

European data draws a clear boundary in how consumers trust AI throughout the journey. Trust is highest when AI supports judgment (synthesizing reviews, highlighting trade-offs, providing reasoned recommendations) and drops as soon as it approaches action (pre-filling carts, autonomous checkout, automatic replenishment).

This is not a rejection of AI's competence, but a resistance to unlimited authority. Comfort is high as long as the agent's actions are reversible, explicitly authorized, and easy to audit, and it decreases as soon as actions become persistent, opaque, or difficult to undo.

For brands, this means designing agentic experiences that:

• Make reasoning visible: explain why a given reference is proposed or discarded, on what criteria, and with what concessions.
• Make authority explicit: clearly define what the agent is mandated to do (suggest, pre-fill, execute under conditions, etc.).
• Make regaining control trivial: confirm, adjust, or cancel in a single gesture, with an understandable history.

In a world of agents, differentiation is no longer just about persuasive copywriting or a polished front-end, but about the ability to express a value proposition in structured "reasons why," usable by both a human and a machine. Brands that cannot articulate who they serve, for which use cases, with what trade-offs, and with what evidence, will be relegated outside the shortlists generated by agents.

5. Think in terms of an agent web, not a single super-assistant

The emerging ecosystem is not one of a monopolistic mega-assistant, but one of an agent web: personal agents, retailer agents, "broker" agents that negotiate and orchestrate on behalf of the consumer. This architecture is being structured around open standards for agentic commerce and payments (Agentic Commerce Protocol -- ACP, Agent2Agent -- A2A, Agent Payments Protocol -- AP2, Universal Commerce Protocol -- UCP).

In this world, owning your own agent is neither necessary nor sufficient to "own" the customer relationship. What remains decisive is the brand's ability to:

• Be reachable and negotiable by any duly authorized agent, via well-defined APIs and protocols.
• Remain merchant of record: identity, payment authorization, commercial policy, fulfillment, and after-sales service, even if discovery and negotiation are delegated to agents.
• Differentiate trusted agents from malicious bots, by finely instrumenting access to data and critical functions.

Building a proprietary consumer-facing agent can make sense where the brand has strong category authority and can offer a genuinely superior guided exploration. But lasting competitive advantage will shift toward the ability to be evaluated, recommended, and transacted reliably in a multi-agent landscape, regardless of which interface the customer chooses.

6. Design a trust-centered delegation trajectory

McKinsey's data shows that delegation will not progress uniformly toward total autonomy. European consumers are currently more inclined to delegate one-off cognitive tasks (receiving suggestions, comparing options, validating a pre-filled cart) than to entrust continuous or implicit authority (subscription management, automatic replenishment without confirmation).

This caution can be read as conservatism, but also as a set of requirements for acceptable delegation: reversibility, clear accountability in case of problems, and explicit consent over the scope of action. For brands, the right approach is not to slow down investment in autonomy, but to structure it as a "delegation roadmap":

1. Phase 1 -- Assistance: the agent helps choose, synthesizes options, suggests bundles, pre-fills a cart, but leaves the final decision and validation to the customer. 2. Phase 2 -- Conditional autonomy: the agent can execute certain actions (replenishment, renewal) within a clearly defined scope, with spending limits, systematic alerts, and easy cancellation. 3. Phase 3 -- Advanced orchestration: for certain complex use cases (travel, health, moving...), the agent orchestrates end-to-end, but within a transparent contractual framework, with standardized identity, authorization, and payment rails (AP2, etc.).

If this pattern holds, the adoption of autonomy will not be a leap toward "fully automated commerce," but a selective and conditional progression, modulated by users' ability to define, monitor, and adjust the scope of their agents. Here again, Europe could offer a blueprint for trust-conscious agentic commerce, rather than representing a structural lag.

7. New marketing and business model trade-offs

In an environment where "agents speak first," part of the traditional media funnel is being short-circuited. Budget trade-offs between branding, performance, retail media, and CRM will need to incorporate a new line item: the ability to feed and influence the AI layers that structure decisions upstream.

Several value creation paths are emerging:

• Monetization of multi-brand bundles optimized by agents, with value-sharing models between partners.
• New forms of real-time negotiation on price, warranties, and services, conducted agent-to-agent within defined protocol frameworks (ACP, A2A, UCP).
• Monetization of insights derived from agentic interactions (within privacy compliance frameworks), to inform product development and segmentation.

In this landscape, customer acquisition cost is no longer reducible to buying impressions, but extends to the ability to be detected as a relevant solution within the flow of signals interpreted by agents (intents, context, constraints). The winning brands will be those that invested early in their legibility to agents, while remaining transaction-ready as soon as consumer trust allows for increasing the level of delegation.