Agentic Commerce: Threat and Opportunity for Businesses

What is agentic commerce?

Agentic commerce is a model in which AI agents would actively participate in purchasing processes: from searching and comparing products to autonomously deciding which brands and products to buy. Unlike chatbots or traditional recommendation engines, these agents would not just suggest options. They would research, compare, negotiate, and in some cases execute the entire purchase on behalf of the user.

Today, this model is more of a vision than an operational reality. Truly autonomous purchasing agents do not yet exist at scale. However, there are concrete signals that the infrastructure to make it possible is being built at remarkable speed.

Shopify enabled checkout within ChatGPT for over one million merchants. Walmart and Target announced integrations with AI platforms. Google launched the Universal Commerce Protocol (UCP) to connect AI agents directly with ecommerce systems. Amazon is developing agentic capabilities within Alexa+, though it simultaneously sued Perplexity for attempting to automate purchases on its marketplace, revealing the tension between adopting and controlling this new channel.

It is important to distinguish between what already exists and what does not yet. What is happening today is that consumers are increasingly using AI tools to research and compare products before buying. According to Adobe, AI-driven traffic to US retail sites grew 760% between November and December 2025. Salesforce reported that AI assisted 17% of orders during the 2025 Thanksgiving weekend, equivalent to $13.5 billion. And Morgan Stanley projects that by 2030, nearly half of online shoppers will use AI shopping agents, accounting for approximately 25% of their spending.

These figures do not describe agents buying autonomously. They describe a prior step: AI integrating into the purchase decision process. But it is precisely this prior step that lays the groundwork for agentic commerce to materialize. The question is not whether agents will participate in purchasing decisions, but when and to what extent.

An example to ground the concept

To understand how this could work in practice, consider a concrete case.

A person wants to optimize their diet to reach a specific weight or body composition goal. AI tools can already generate personalized nutrition plans, suggest recipes, and adjust recommendations using data from devices like smart scales.

But implementing those diets requires something very concrete: buying the food. Today, that means going to the grocery store or placing orders across multiple platforms: the supermarket for base ingredients, Amazon for specific supplements not available in a single store. A fragmented and repetitive process.

A purchasing agent would take the AI-generated list and execute the purchases automatically, optimizing for price, availability, or any user-configured preference. The entire process, from the nutrition plan to product delivery, could be fully automated.

The key point: in this flow, for a significant subset of products, the brand would be selected by the agent, not by the consumer. And this pattern would replicate across virtually any consumer category: from a family’s weekly meals to the recurring purchase of cleaning or personal care products.

Why agentic commerce could reshape the competitive landscape

AI shopping agents, seeking to deliver the best results for their users, would likely give more weight to objective attributes (composition, certifications, ratings, price-to-value ratio) and less to brand positioning. Unless, of course, the user has specified a preference for specific brands.

This does not mean traditional branding will disappear. But it likely will not be able to fully defend the market share that purchasing agents capture. What portion of buying decisions will migrate to agents, and how quickly, remain open questions. But the direction seems clear.

For established brands, this would represent a threat. For new and emerging brands, an opportunity. Three factors explain why.

First, the historical advantage of large brands rests largely on achieving strong brand positioning: familiarity, shelf presence, advertising recall. An agent optimizing for results for its user would have little reason to replicate those biases.

Second, unlike the traditional search model where a single brand tends to dominate the top result, the probabilistic nature of AI models suggests that multiple brands could be selected across different purchasing processes. In the marketing exercises for LLMs we have conducted at Julius, particularly in B2B, we have found that models tend to recommend diverse brands rather than concentrating brand visibility on a single one. This creates scenarios very different from Google, where the top result captures most of the traffic. The playing field would become more distributed.

Third, positioning a brand for AI agents could require significantly less budget than traditional marketing aimed at millions of consumers. Barriers to entry would shrink. This does not mean it would be trivial, but the investment required to be “visible” to an algorithm is fundamentally different from building brand recall in the minds of millions of people.

The ecosystem being built

Agentic commerce is not developing in a vacuum. There is active competition among major technology companies to build the infrastructure that would make it possible, though real-world use cases remain minimal.

OpenAI launched “Buy it in ChatGPT” with direct integration to Shopify and Etsy, enabling purchases without leaving the conversation. Google responded with the Universal Commerce Protocol and agentic checkout capabilities within Gemini. Amazon is developing Alexa+ with autonomous purchasing features. Perplexity launched “Buy with Pro” with integrated checkout via PayPal. And on the payments front, Mastercard, Stripe, and PayPal are building the infrastructure for agents to execute transactions securely.

That these companies are investing simultaneously and aggressively does not guarantee the model will work as they envision. But it does indicate the bet is serious and that, if it materializes, there would likely not be a single dominant agent but multiple platforms competing to deliver better results to their users. For brands, this would mean that a marketing strategy for AI agents could not be limited to optimizing for a single platform.

Potential implications for Latin America

Agentic commerce is advancing faster in the United States, where ecommerce and digital payments infrastructure is more developed. If this model gains traction, the implications for Latin America would be direct.

Latin American brands that export or sell to US consumers would face this shift immediately as agents gain adoption in that market. And in local markets, platforms like Mercado Libre, Amazon Mexico, and Rappi have the incentives and base infrastructure to integrate agentic capabilities into their ecosystems as the technology matures.

Additionally, the Latin American context presents a particular dynamic: markets where brand loyalty tends to be lower than in more mature economies, where price sensitivity is high, and where emerging brands can compete with greater agility. If agentic commerce lowers barriers to entry and redistributes visibility, Latin American brands that prepare early could capture market share previously reserved for competitors with much larger budgets.

How to prepare: marketing for AI agents

Although agentic commerce does not yet operate at scale, there are actions brands can take today that generate value regardless of when and how this model materializes. The three lines of action proposed below improve a brand’s competitiveness both in the current environment and in a future agentic scenario.

Objective product attributes

In an environment where agents prioritize measurable data, brands with clear, well-documented, and easily processable attributes would have an advantage over those that rely primarily on aspirational positioning. This means investing in the quality and structure of product data: detailed composition, verifiable certifications, performance metrics, user ratings. It is not about abandoning branding, but complementing it with information an algorithm can interpret and compare objectively.

This investment also has immediate benefits: it improves the purchasing experience in current digital channels, facilitates marketplace integration, and strengthens traditional SEO.

AEO: Agentic Engine Optimization

AEO is the practice of optimizing a brand’s presence to be correctly interpreted and recommended by AI models. The term encompasses both optimization for answer engines (Answer Engine Optimization) and optimization specifically for autonomous agents.

Where this is already a tangible reality is in B2B: more and more buyers research solutions directly through LLMs like ChatGPT, Perplexity, and Gemini, rather than traditional Google searches. This is not agentic commerce per se (there is no agent buying autonomously), but it is the same principle applied to brand and product discovery. Gartner projects that by 2026, 25% of organic search traffic will migrate to AI chatbots and virtual assistants.

If agentic commerce materializes in B2C, AEO would become equally critical for consumer brands: if an agent does not “know” your product or does not interpret it correctly, it simply would not recommend or select it.

The good news is that AEO does not replace traditional SEO but builds on it. The foundations of good SEO (structured content, domain authority, accurate data) are also the foundations of AEO. At Julius, we are currently offering these services to a growing number of clients, in both B2B and early B2C engagements.

Multichannel availability

Purchasing agents would likely seek to minimize the number of retailers involved in a transaction to optimize costs and logistics. A product available in only one channel would be more likely to be substituted by a comparable alternative the agent can include in a consolidated purchase.

This is particularly relevant in categories where products are highly interchangeable. Multichannel availability, which has always been important, would acquire a new dimension when the “buyer” is an algorithm that can evaluate options across multiple platforms simultaneously and in real time.

The window to act

Agentic commerce is still more of an idea than an operational reality at scale. But the pace of AI model advancement, combined with massive investment from major technology platforms, suggests the window to prepare could be shorter than it seems.

McKinsey projects that by 2030, agentic commerce could represent up to $1 trillion in orchestrated revenue in the US B2C retail market alone, with global projections reaching $3 to $5 trillion. Even if these figures prove optimistic, a fraction of that volume would represent a significant shift in how market share is distributed.

There is legitimate uncertainty about the pace and magnitude of this transition. But the actions needed to prepare (improving product data, optimizing for AI, strengthening multichannel presence) generate value today, regardless of when agentic commerce arrives at scale. That is perhaps the most pragmatic argument for starting now: the cost of preparing is low, the cost of not doing so could be high, and the benefits are immediate.

The topic will continue to evolve. What is not optional is starting to pay attention.