OPINION: The Real Threat of Agentic Commerce (And How Merchants Can Keep Control)

This is an opinion piece from one of our experts: views expressed in this article are the author’s own.

Many in the payments industry are focused on the shiny new technology right now. True, we are seeing chatbots negotiating prices, digital wallets acting autonomously, and AI agents executing purchases in milliseconds (even though a good amount of this is happening in test environments or by individuals in the industry themselves).

The innovation is undoubtedly impressive.

However, tech giants and payment service providers – whether intentionally or not – appear to be actively laying the groundwork to step between you and your buyers. We need to look at the long-term impact of Agentic Commerce and its potential to strip merchants of their customer relationships, data, and margins.

The primary risk right now is surrendering control of your infrastructure too early. Here is the strategic reality of AI-driven commerce and a playbook for retaining your independence.

The Three Silent Threats of Agentic Commerce

1. Third parties are hijacking your customers’ intent.

When a customer delegates their shopping to an AI agent, the entity controlling that agent controls the purchase intent. If a third-party platform sets the spending parameters, defines the limits, and holds the proof of purchase, they own the customer. In that scenario, your brand may fade into the background. You become a commodity supplier competing entirely on price and fulfillment speed.

Protocols are being developed to reestablish the visibility merchants get into this intent but still require more merchant input in order to offer a better channel experience than existing ecommerce, rather than a return to the days of early internet selling.

To that point, take the payment performance of a very prominent AI Platform. This is real data from CMSPI’s Merchant Payments Intelligence, drawing from across our client base.

For card-based payments initiated by Agents (for which I’m determined to make AIT a thing), the weighted average net authorization rates hovered only marginally above 50%. For a customer-initiated transaction (CIT) or even a merchant-initiated one (MIT), this is objectively poor. Compound that with the fact that this approach costs merchants 3 times the U.S. average cost of acceptance, never mind the associated fraud and dispute challenges, and you can see that the channel is simply not ready for wider adoption*.

*Which is okay, by the way: we are still at the very early stages and the pace of progress is unprecedented. 

2. Your product catalog is completely invisible to AI agents.

AI agents do not browse beautifully designed websites. They read raw data. Your product feeds, pricing, availability, and return policies must be structured into perfectly compliant, machine-readable metadata. Without that, your products are invisible in the agentic ecosystem.  

Again, much of the work happening on Agentic Commerce today is focused on the Discovery aspect (how products are surfaced to consumers) and improving access to structured product data on merchant websites. You may find that many of your favorite sites as a consumer now have a version of the website fully set up to be machine readable. The best example of this in the wild is actually Stripe and OpenAI’s Agentic Commerce Protocol developer site, which looks different depending on who is viewing it – as shown below. 

3. Proprietary payment ecosystems will trap your data.

As AI fragments the shopping journey across various interfaces and bots, traditional customer recognition tools like cookies and logins will break down. Payment data will become the only connective tissue linking you to your buyer. If your PSP locks your payment tokens inside their proprietary ecosystem, you lose the ability to recognize your own customers across different AI platforms. You may also be surrendering your ability to leverage optimization strategies: debit routing rights being one of them.

The Merchant’s Defense Playbook

You cannot wait for the dust to settle, and you need to avoid locking your business into closed ecosystems. Building an agnostic foundation is essential.

• Treat your product data as core infrastructure: Overhaul your product catalogs today. Make all pricing, eligibility, and policy data completely machine-readable. You need to be ready to feed multiple AI discovery surfaces instead of relying on a single search engine.
• Demand absolute portability for your payment tokens and data: Demand payment portability. Implement network or independent tokenization strategies. This allows you to recognize your customers and route transactions efficiently, regardless of which AI agent initiated the checkout.
• You must dictate what AI agents are allowed to do: Third parties should never define what an AI agent is allowed to do on your behalf. You must explicitly own the mandate. Set the exact spending limits, context, and permissions yourself, and keep an auditable, merchant-owned record of that authorization. You may wish to consider building your own Merchant Agent…
• Align your internal teams to handle automated mistakes: Machine-speed purchasing leads to machine-speed mistakes. Fraud, Treasury, and Product teams must align immediately on how to distinguish authorized agents from malicious bots and apply that to transactional data. You also need to agree on who holds the liability when automation fails.

Where CMSPI Stands

The industry is rushing to build the infrastructure of Agentic Commerce. Our focus at CMSPI is ensuring you maintain the leverage to navigate this shift on your own terms. We help translate the complex protocols of Big Tech and the card networks into actionable strategies specifically aimed at maximizing payments performance. Our goal is to ensure your payments stack remains interoperable, your costs remain optimized, transaction success remains high, and your customer relationships remain yours.

Agentic commerce is coming. You need to decide today who will own your customer tomorrow.