Agentic AI meets e-commerce: discovery, loyalty, and channel strategy

AI agents are starting to sit between consumers and merchants. That changes who controls product discovery, how brands get surfaced, and what loyalty means when a model picks the product instead of a person.

This article distills a recent LinkedIn Live event hosted by Marvin Labs on how agentic AI is reshaping e-commerce from the customer journey through to channel economics. The session featured Alex Hoffmann (Founder and CEO, Marvin Labs), Zvonimir Filjak (AI Strategy and Innovation Senior Manager), and moderator James Yerkess (Former Global Head of Transaction Banking and FX, HSBC Wealth Management).

Four layers, not one trend

Much of the discourse around "AI in commerce" conflates several distinct shifts. A more useful way to think about it: break the end-to-end journey into four separate layers.

1. Discovery. Can your brand appear in an AI-generated answer? When a consumer asks ChatGPT or Gemini for the best running shoes under $200, the response draws on web-accessible content. Brands that are included in those answers get surfaced. Others do not.

2. Transaction execution. Can an AI agent complete a purchase on behalf of the customer? This requires API connectivity, inventory data, and payment integration between the agent and the merchant.

3. Business agents. Companies deploying their own AI agents to represent the business, answering customer questions, checking inventory, and handling bookings directly.

4. Infrastructure. How do these systems talk to each other? Protocols, structured data, and product feeds determine which platforms can participate and which cannot.

Once you start separating those pieces, the whole ecosystem becomes much easier to understand. Discovery, transaction, infrastructure, business agents. Four separate conversations.

For investors, this means asking which layer a given retail or consumer name is actually exposed to. A business investing in answer engine optimization faces a different cost structure and competitive dynamic than one building API-enabled transaction infrastructure.

Discovery is the new storefront

Product discovery is increasingly expanding from keyword search into prompt-driven AI responses. For a decade, SEO meant ranking on page one of Google. In an AI-mediated world, the target is smaller and the stakes are higher.

In SEO, your target was about 10 spots on the first page. The difference between number three and number nine was small. In an AI world, the space is much smaller. The AI may give you a much smaller shortlist of options.

That shrinks the competitive set considerably. Brands with strong SEO and discoverability practices will benefit disproportionately. Those built primarily on social media presence face a harder path, because social platforms are largely closed gardens that most AI systems do not index.

For a brand to show up in an AI response, three things need to be in place. First, accessibility: can AI systems access your website and product pages at all? If your content is not machine-readable, you're much less likely to be included in the model's response. Second, structure: schema markup, structured product data, and content aligned to how customers phrase questions in natural language. Third, credibility: even if AI can find and parse your content, it still needs to trust you enough to recommend you. Strong reviews, thought leadership pieces, and mentions in industry publications all feed that trust signal.

Marketers call this answer engine optimization (AEO). AEO is not a replacement for SEO but a second layer on the same foundation. SEO gets your content indexed. AEO gets it included in the AI-generated answer.

For analysts covering retail and consumer names, the question is practical: which companies in your coverage universe have the content infrastructure and data quality to be included in AI-mediated discovery? And which are invisible?

Customer relationships move inside the agent

In traditional e-commerce, the business owns the customer relationship. A consumer visits a website, browses products, compares options, and completes a purchase. The merchant sees the full journey and builds a behavioral profile over time.

When an AI agent handles that journey, the dynamics shift. The conversation between customer and agent captures intent data, comparison criteria, and purchase rationale. That data sits with the agent, not the merchant. The merchant may only see the final transaction without context on how or why the customer arrived.

The AI agent might be sitting on that golden piece of information, which might never actually get passed on to you at the execution and transaction layer.

That feeds directly into customer acquisition cost modeling and lifetime value calculations. If the agent becomes the primary discovery and comparison channel, merchants compete for the agent's recommendation rather than for the consumer's direct attention. The data advantage shifts to whoever runs the agents.

This also changes what loyalty means. Traditionally, customers stayed loyal because they liked the brand, trusted the experience, or built a habit over time. In an agent-mediated world, the AI assesses businesses based on data quality, pricing, reviews, and delivery speed. Those signals determine which businesses get recommended. Call it algorithmic loyalty: the AI increasingly shapes which brands get shortlisted and recommended.

The question is no longer just whether customers love our brand. It's also: does AI understand our business well enough to recommend us?

Worth monitoring: how retail and consumer companies report on data infrastructure, API readiness, and content strategy alongside the usual metrics like same-store sales and customer satisfaction scores.

AI as a channel decision, not a pilot project

The mobile and omnichannel transitions offer a useful parallel. Some brands adapted early and grew with the new channel. Others treated it as an experiment, fell behind, and never caught up.

In 12 months, from a brand's perspective, the agent making the purchase decision should look the same as somebody buying on Amazon. Right data, right APIs, right inventory. You don't care if they buy on Amazon or through ChatGPT. Just be ready.

There are two ends of the spectrum. On one side, businesses focus on driving customers to their own apps and websites, environments they fully control. On the other, they pursue a full distribution approach, actively participating in AI-mediated channels and making their products available for agents to recommend and transact directly.

Most businesses will land somewhere in between. But the decision needs to be strategic, not a side experiment. Companies that treat AI commerce as a pilot project risk finding themselves on someone else's platform without understanding what they gave up.

The question for coverage models is how a company's AI readiness fits within its broader channel strategy. Companies already selling through aggregators and marketplaces (Amazon, comparison sites, multi-brand retailers) may find the transition more natural than pure direct-to-consumer brands.

Friction points for analysts

Several areas of friction between AI agents and traditional commerce models are worth tracking.

Product presentation. In traditional e-commerce, brands design every element of the customer experience, from product pages to checkout flows to how comparisons appear. When an AI agent summarizes all of that in a single response, the brand loses control over how the product gets presented to the buyer.

Transaction ownership. If a consumer purchases through an AI agent, who owns the transaction legally? Who handles returns, disputes, customer service? Those frameworks are still forming.

Comparison markets. For complex products like car insurance, where multi-million dollar comparison industries already exist, AI-mediated comparison could restructure entire market segments. The AI aggregates what comparison sites do today, but in a single interaction.

Marketing accountability. AI-driven discovery may surface problems that glossy marketing currently covers up. An insurer with poor claims ratios and high complaint numbers can mask those issues with advertising spend today. AI systems that pull from reviews, complaints data, and regulatory filings will surface that information directly. The discovery layer starts prioritizing substance over presentation.

What to track in the next 12 months

Three areas deserve attention over the coming year.

Algorithmic loyalty signals. How AI systems assess and recommend businesses. The signals AI uses to form recommendations (data quality, pricing, reviews, delivery performance) will matter as much as brand perception among consumers.

Execution readiness. Whether companies' systems can handle transactions from any channel. Right data, right APIs, right inventory. The technical bar is lower than mobile and omnichannel were, but the brands that are not ready will get left out again.

Disclosure signals in earnings commentary. How retail and e-commerce companies discuss AI in the context of commerce and customer acquisition. Where transactions are happening, how channel strategies are evolving, and which companies are investing in infrastructure versus talking about it.

For the full conversation, including audience questions on platform data moats, monetization model shifts, and the comparison between AI channels and social commerce, watch the video above.