For years, brands have invested billions to perfect their online storefronts. Every pixel calibrated, every checkout flow streamlined, every funnel obsessively measured. But something fundamental has changed. Shoppers aren’t arriving through the front door anymore. First, they used the mall entrance (Amazon, Walmart) because parking was easier and there was so much selection. Going forward, AI agents are going to start letting themselves in through the side door and make themselves at home.
AI agents now take action on behalf of shoppers. Tools like Perplexity and ChatGPT are already discovering products, comparing prices, building carts and executing transactions on behalf of users.
The shift is fast, quiet and sweeping. And it’s about to catch merchants flat-footed.
Here’s why: AI agents don’t navigate homepages or click through categories like people do. They parse structured data, interpret user intent and transact programmatically. Checkout becomes a backend function rather than a shopping experience.
Advertisement
In this environment, traditional merchandising breaks down. That means product bundles may be split apart. Regional availability may be ignored. Promo codes may be applied incorrectly or inconsistently. And yet, when an agent gets it wrong, the customer will still blame the brand.
In this model, the storefront loses both traffic and relevance. So what’s the next best move? Like many things in tech, we can learn by studying patterns in the past.
Shifts in Discovery and Checkout
We’ve seen this kind of shift before. In the early internet days, brands depended on homepage traffic and organic links. Then came search engines, where product discovery moved to Google and SEO-focused deep content. More recently, traditional web search has been disintermediated by social channels like Instagram and TikTok, and marketplaces like Amazon and Walmart.
Agentic commerce is poised to threaten both ad-centric business models (like search, social and retail media) and marketplaces by jumping straight to the transaction. But the wild-wild-West approach that’s happening now isn’t going to work.
When search crawlers first arrived, there were no rules. So robots.txt was born — a lightweight protocol to declare what bots could or couldn’t access. Today’s agents need something similar, potentially llms.txt, but far more powerful, with both visibility controls and transaction-level permissions.
And as AI agents evolve, checkout is being reduced to an API call. Platforms like Visa and Shopify already are introducing infrastructure to support agent-based purchases. Instead of optimizing a marketing funnel, it’s about exposing the right data and validating that the agent requesting a transaction is authorized to do so, while capturing the corresponding take rate on the transaction.
This mirrors the evolution of payments in the early 2000s. Online commerce was chaotic until PCI DSS created a security standard for handling credit card data. PCI both made payments safer and made online checkout scalable. Agentic commerce needs a similar trust layer. Otherwise, brands are left exposed and unable to verify who’s buying, from where, or under what conditions.
Are Merchants Prepared for the Next Phase?
Most brands aren’t ready for this shift because their data isn’t structured. Their inventory and pricing aren’t real-time. Their systems were built to block autonomous agents, rather than transact with them. As a result, they’re invisible because they aren’t operating storefronts that agents can find, interpret, or trust.
The Domain Name System (DNS) solved this problem on the web by mapping readable names to digital identities. Without it, brands were unfindable. Today, MCP (Model Context Protocol) and agent-accessible endpoints will serve the same purpose. They’ll be the lookup layer for commerce. If you’re not there, you don’t exist.
As for the storefront, what used to be a visual experience is now an architectural one. The storefront is being replaced by APIs, permissions and policy layers that define how agents engage with your business. Brands will need to adopt the equivalent of OAuth for commerce, or token-based scopes that determine what an agent can do: browse, price or purchase.
These changes will enable trust. Just as OAuth transformed third-party app security by making access delegated, limited and auditable, commerce will need the same scaffolding to prevent fraud and errors. Whether that’s orchestrated by payment providers like Visa or commerce platforms themselves has yet to be seen.
The Most Disruptive Technologies are Quiet
The most disruptive technologies aren’t always the loudest. AI agents may not crash your site, but they’ll silently bypass it. If brands want to compete in an agent-mediated world, they need to shift their focus from building the best storefront to orchestrating their infrastructure.
The question is no longer “How good is your storefront?” It’s “Can an agent understand you, trust you and transact with you accurately?”
For many, the answer today is no.
Bryan House is the CEO at Elastic Path, where he leads the GTM, customer success, global services and product teams.
