When e-commerce first took off in the 1990s, many established retailers faced an existential crisis: resist or embrace? Those that prioritized brick-and-mortar identities only over an online presence eventually fell behind in both sales and consumer mindshare. Conversely, retailers that aggressively invested in their own e-commerce platforms with clear differentiators and strong omnichannel presence were better positioned to flourish over time.

Written in collaboration with

Written in collaboration with

---

Now retailers are facing a new source of disruption: agentic AI. AI agents, which can autonomously complete tasks on a user’s behalf, are poised to significantly transform how consumers discover, research, compare, and purchase products by completing some or all of these steps for them. Full agent-to-agent (A2A) commerce, in which the buyer’s and retailer’s AI agents transact directly, could eventually shortcut or bypass classic retail websites altogether as consumers rely on agents to determine the best value based on predetermined goals.

For consumers, the potential upside is clear. Buyers can expect to find the best offers suited to their preferences, be it lower prices, attractive rewards, better services, faster shipping, or unique products, through one interface.

The outcomes are more mixed for retailers. The ability to cross-list more items in more places could lead to greater discovery and higher conversions for offerings that are clearly superior. But there is a very real risk that AI intermediaries will weaken unit economics while extracting value over the long term.

To be sure, A2A commerce is still in its infancy. Around half of consumers say they are not ready to let AI handle an end-to-end transaction without their oversight. The number of consumers who will behave cautiously is likely meaningfully higher.

However, generative AI platforms like ChatGPT, Microsoft Copilot, and Gemini are already reshaping consumer behavior. According to Bain’s Consumer Lab Generative AI Survey, 30% to 45% of US consumers use generative AI to research and compare products. Early adopters are also using ChatGPT and Copilot to shop and check out directly from the platforms.

As AI becomes more embedded in everyday tasks and consumer trust in AI agents strengthens, adoption of A2A commerce will accelerate. This gives retailers a narrow window to apply lessons from the past to their strategy for the future.

Three ways to play in an A2A ecosystem

Just as they did with the arrival of e-commerce, retailers will need to rethink operations, marketing, technology, data, and pricing to stay relevant in an agent-first world. First, they must determine how open or closed they will be to third-party agents. Three overarching categories are emerging along a spectrum of possibilities:

Embrace the agents. One option for retailers is to allow third-party agents to “crawl” their sites, list their products, and often close transactions. This may be the smarter play for retailers that do not have the brand awareness or consideration in their target audience to drive traffic at scale or that lack the resources to build their own agentic platform.

Monobrand retailers will have some advantages in this scenario. Third-party agents can expose their products to net-new buyers and become a meaningful additional revenue stream. However, they will need to contend with the risk of commoditization as well as a fraying of brand loyalty and direct customer connections.

Multibrand retailers and marketplaces will likely face a greater disintermediation threat. Third-party AI agents increase market transparency, giving big, low-cost, high-speed players an edge. Retailers that cannot compete based on price, unique assortment, or shipping speed alone will need to develop a competitive differentiator that is attractive to agents. Online marketplaces will also need to prepare for the possibility that agents will bypass their storefronts and route consumers directly to sellers. Trust measures, such as buyer guarantees, product ratings, and consumer reviews, will provide important defense mechanisms to gain consumer confidence.

Build the agents. Retailers that enjoy a strong brand following and capture significant traffic at scale could opt to build their own agentic ecosystem. By creating a front door for searches using proprietary agents, marketplace retailers could make their brands synonymous with convenience and value. Retailers that pursue this strategy may also opt to attract other retailers to their platform and monetize the traffic they bring. Right now, consumer sentiment is on their side. Shoppers indicate they trust retailers’ on-site agents three times more than third-party agents.

With the release of its “Buy for Me” agent, Amazon has staked out a strong offensive position in this category. The retailer’s AI agent can shop other brands’ sites if a product is not available on Amazon. All sales are processed through Amazon, giving the retailer access to customer data. However, the company has prohibited external agents from interacting with their site directly. While this model enables purchases from retailers outside the Amazon ecosystem, it reinforces Amazon’s position as a launching point for the consumer journey.

The strategy is not without its risks. Retailers need to consider how to attract other retailers to allow their agents to shop them, how to securely facilitate purchases using a trusted payment processing option, how to manage customer service for third-party retailers, and how to protect their own brand from fraud or customer complaints lodged against the retailers they partner with.

Fortify the home-site value proposition. A third option for retailers is to provide limited exposure to third-party agents while investing in initiatives to maintain and even build brand loyalty.

In this scenario, retailers can open some products or categories to crawling by AI agents while making some experiences exclusive to their home site or app. A relatively simple approach is to gate certain products and services—for example, allowing the purchase of appliances through an AI agent but not the installation services. Retailers can do the same thing with limited-time offerings, warranty support, free delivery, and other added services.

Although it is more involved, retailers can also fortify their home sites by offering unique, native agentic capabilities that are difficult for third-party agents to replicate. Home Depot’s AI companion, Magic Apron, is only available on their website and provides specialized customer support. With access to customer and purchase data that third-party agents lack, Magic Apron can offer more personalized service that taps into existing consumer trust with the brand.

This strategy provides some measure of protection against commoditization without requiring the resources and brand standing needed to build a new ecosystem. Negotiating agreements to share customer data on transactions that close off-site will still be essential.

Retooling the e-commerce foundation

Retailers can keep control of customer data and loyalty only if they control where and how customers complete their transactions. It won’t be as simple as using existing e-commerce platforms. Many current e-commerce fraud prevention technologies are built to detect and block bots, as bots have historically been the quickest way to drain large pools of stolen credit card numbers by fraud rings.

Tokenization, an established payments technique that replaces traditional payment account data (e.g., a plaintext 16-digit account number, expiry, and CVV) with non-sensitive equivalents (the token), are one method of validation for protecting against bad bot behaviors. Tokens can only be redeemed for payment by the intended recipient and often include additional proof that the account holder authorized the payment. However, the seamless, token-based checkout APIs that are needed for agent-completed purchases are largely absent from existing retailer checkout systems. Without these capabilities, retailers are at risk of being locked out.

Fortunately, several payment processors and credit card companies provide integrated value-added services through embedded payments. Payment processors such as Stripe and PayPal offer AI-forward solutions and layered security that can support A2A commerce. For example, Stripe currently powers purchases made through Instant Checkout in ChatGPT by passing shared payment tokens (SPTs) from agent to seller. SPTs allow agents to securely handle credit card information; they also include fraud risk scores that sellers can use when deciding whether to fulfill a purchase request.

Preparing for the agentic wave

Unfortunately, there is no “one size fits all” approach to A2A commerce. Retailers may find the answers to the questions they face are not a simple yes/no binary, but rather a yes/no/and/or. What’s clear is that “do nothing” is not an option in this iteration of retail economics. No matter which path forward retailers make, they should prioritize the following:

• Own the checkout process. When possible, retailers should drive consumers to close transactions on their own platforms using their preferred payment processor. Otherwise, customer retention strategies, such as support and refunds, could be ceded to the third-party AI platform, potentially exposing the brand to greater risk.
• Retain access to customer data. If closing the transaction on your own platform is not an option, partnerships will be needed to retain access to customer data and consumer behavior. This will be key for retailers to enable critical marketing strategies, such as retargeting and encouraging repeat purchase behavior. Watermarking, tiered access to critical software, and tracking can all be deployed to protect first-party data. Without this, retailers will become little more than a fulfillment pipe.
• Optimize data for AI engines. Unless you stay fully closed to A2A commerce, you will need to ensure your products or offerings can be ingested by agents. Marketplace retailers can better position themselves as a preferred choice by adapting their architecture to support agent accessibility, offering preferred pricing for agentic channels and structuring promotions and bundles in ways that appeal to agentic logic.

Retailers who lead with strategy, operational discipline, and collaborative spirit will not only survive the AI revolution, but set new standards for what seamless, client-centric commerce can be. Success will be defined not by chasing the latest hype, but by methodically retooling for an A2A-mediated marketplace.

About Stripe

Stripe is a technology company that builds economic infrastructure for the internet. Businesses of every size—from new startups to public companies—use our software to accept payments and manage their businesses online.