Why the next brand battleground will be fought in front of agents, not humans.
In just a few years, generative AI has moved from novelty to default front door. Large language models (LLMs) already attract close to 45 billion visits per month, the equivalent of 56% of traditional search volume, which has plateaued for years. But the figure that should capture every executive’s attention is not the size of the audience: it is the behavior behind it. The average LLM session now lasts six to eight minutes, up 34% year-on-year. Where a Google search was a transaction, a conversation with an AI agent builds a relationship.
Customers no longer buy a product; they commission an outcome
This shift transforms the very nature of the purchase act. Nobody buys a drill; they buy a hole in the wall. Yesterday, a customer asked, “Which drill should I buy?” Tomorrow, they will say, “I need to hang a frame before Saturday, what should I do?” and the agent will choose the drill, the seller, and the price.
For twenty years, the intermediary between brand and customer was the sales associate in store, then search engine rankings (SEO/SEA). Tomorrow, that intermediary will be an AI agent. The question then becomes: do you want it to choose you? The stakes are not marginal, and AI is no longer a side channel: in a single year, AI-driven traffic to retail sites has surged by almost 400% and has gone from trailing every acquisition channel to outperforming them all.
Your visibility is already slipping away
The risk is not speculative; it is measurable. On the web, 57% of requests are now automated, rather than human. Your website is already read more by machines than by people. More unsettling still, 85% of brand mentions in AI responses come from third-party content (forums, encyclopaedias, review platforms…) that you do not control. A brand’s visibility no longer depends on its own website, but on how the rest of the world talks about it. For the first time, your digital exposure is largely out of your direct control.
The transaction is the last domino: that is your window
Here is the paradox every leader should understand. Discovery has already migrated to AI, the surge in AI-driven retail traffic is proof enough. Yet the purchase itself remains, overwhelmingly, a manual and human-led act. Industry observers note that while shoppers increasingly turn to chatbots for recommendations, platforms are still seeing a low volume of transactions actually being completed through AI tools. For now, even OpenAI has refocused ChatGPT on product discovery, pushing the act of buying back out to merchants’ connected apps.
Part of the reason is that the rails are still being laid, amid a very public battle over competing standards. OpenAI and Stripe are pushing the Agentic Commerce Protocol; Google is advancing its Universal Commerce Protocol; Amazon is renting out its own agentic shopping engine to other retailers. No single standard has won. Being “machine-ready” is therefore a bet on rails that have no clear winner yet. Which is precisely why, today, foundations beat features. The transaction is the last domino to fall, not the first. The brands that win will not be those that rushed a checkout button into a chatbot, but those that quietly made their data, APIs and processes ready while the standards settle. The window is open exactly because no one has won the transaction layer.
Three workstreams, three timeframes
Faced with this tipping point, we identify three complementary workstreams, each corresponding to a distinct action horizon:
1) Be visible and readable: GEO. Generative Engine Optimization is not a replacement for SEO; it builds on top of it. You must appear (SEO), then be extracted into AI answers (AEO), then be cited and recommended by name (GEO). The counter-intuitive truth: you cannot skip the first rung. Roughly 92% of AI-answer citations come from pages already ranking in the top ten. An LLM draws on two sources: its training memory, a black box refreshed every three to eighteen months, over which a brand has little influence; and real-time web search, where the true leverage lies. Making content “machine-readable” (semantic HTML, JSON-LD structured data, precise and recent product information) is the fastest way to exist there. And recency matters, as content less than 90 days old is about three times more likely to be cited. The expert takeaway: you are no longer writing for humans who scan, but for models that extract.
2) Be operable: interoperability. This is where the most costly trap awaits. The temptation is to deploy a visible, polished chatbot. But 80% of the work is invisible: it lives in the foundations: system APIfication, machine-readable data, an agentic platform with supervision and guardrails. Exposing an agent inside a portal like ChatGPT takes a few days; making a business process genuinely actionable by a machine takes months. A strategic fork is already here: “agent-ready” is becoming a product that platforms will sell to you. Amazon now offers an embeddable agentic shopping assistant for third-party retailers, while Alexa adds price tracking and auto-buy. They’re convenient, but renting someone else’s agentic layer can mean ceding the customer relationship itself. This make-versus-buy decision is among the most consequential a brand will make in the next eighteen months.
3) Stay preferred: Brand Equity. When an agent compares dozens of equally “machine-ready” offers, the ultimate risk is commoditization: being selected on price alone. The only defense is brand preference: being the trusted reference in the customer’s mind before they ever delegate the decision. Consumer appetite is already there: shopping-related use of generative AI grew 35% in under a year, with buyers increasingly treating AI as a direct, objective and personalized expert. Trust is shifting; it is not disappearing.
A conviction we are already putting into practice
These principles are not theoretical. At Artefact, we are already building agentic conversational journeys for major brands. Tomorrow, your customers will no longer complete fields, they will talk to your brand, or to an agent speaking on its behalf.
Agentic commerce is not an evolution of e-commerce; it is a displacement of the decision point. The question is no longer “How do my customers find me?” but “Am I chosen when I am not in the room?” The brands that lay the foundations now – visibility, interoperability, preference – will write the rules everyone else will have to follow.
