Guardrails Needed as AI Agents Transform Commerce

McKinsey says AI agents will create an “agentic commerce” era that simplifies buying and selling. The promise is real: less search friction, more tailored offers, fewer clicks. But the pitch that agents will be neutral matchmakers for consumers and merchants is incomplete—frankly, it’s optimistic in a way that hides who gains power and who gets squeezed.

Agent convenience vs. data feudalism
McKinsey’s piece correctly frames agents as a new retail layer that can orchestrate discovery, comparison, and purchase. The part it soft-pedals is that these agents will sit on top of data pipelines owned by a small set of platforms, payment providers, and dominant retailers. Whoever controls those pipes controls what gets seen, at what price, and on what terms.

Right now, you still see the “storefront”: search pages, product grids, filters. With agents, many of those choices move behind an interface that simply says, “Here’s the best option for you.” Once the decision logic is encapsulated in opaque models, you’ve effectively outsourced your preferences to a system whose incentives you can’t inspect. The math doesn't lie: when choice is filtered through an agent’s embedded agenda, “consumer empowerment” can quietly turn into vendor selection engineered by the platforms that own the rails.

Platforms already behave like rent-collecting intermediaries. Agents just move the toll booth from a visible marketplace to an invisible decision engine. Merchants don’t only fight for shelf space anymore; they fight to be ranked favorably in an agent’s internal scoring. That’s not some neutral upgrade to market efficiency—it’s a redistribution of bargaining power toward whoever designs and hosts the agents.

Who benefits: merchants versus marketplaces
McKinsey is right that agents can help merchants find customers more efficiently. But that upside is heavily skewed. The sellers best positioned to win are those that can feed agents pristine structured data, run continuous experiments to tune their presence, and negotiate direct integrations and commercial terms.

Small merchants will not magically gain access to this layer on equal footing. When access depends on API integrations, clean metadata, and participation in evolving agent ecosystems, you tilt the field toward players with engineers, legal teams, and the cash to test business models that may take years to pay off. McKinsey hints at this, but treats it as execution detail; it’s a structural filter determining who even gets to participate.

We’ve seen this movie before with search and marketplaces. Early Google rewarded relevance; later Google rewarded whoever could afford entire SEO teams. Amazon’s marketplace started as a lifeline for small sellers, then evolved into an environment where advertising and paid placement dominate visibility. Agentic commerce is poised to follow that path, just with more automation and less transparency around why one merchant gets surfaced and another doesn’t.

A blind spot: the cost of coordination
McKinsey is enthusiastic about the new business models agents unlock—personal shopping agents, automated procurement, dynamic bundling. What it mostly skips is the cost and politics of coordination. For agents to interact across platforms—say, a consumer’s agent talking to multiple merchant or logistics agents—there have to be agreed rules: identity, messaging formats, dispute procedures, failover when agents disagree.

Who sets those? An industry consortium will skew toward incumbents. A regulator-led process will be slower but potentially more neutral. A dominant cloud or commerce platform taking the lead will accelerate adoption—at the price of cementing that platform as the default gatekeeper. Each path bakes a different distribution of power into the infrastructure. The choice of coordination model is not plumbing; it’s policy by other means.

Privacy, trust, and missing guardrails
McKinsey sketches compelling scenarios: agents that compare offers, negotiate on your behalf, maybe even execute purchases once you set constraints. None of that works without deep access to your data—purchases, browsing, financial profiles, even context like health or household composition.

The risk isn’t just misuse of data; it’s misaligned incentives. If an agent is funded by referral fees or bounties from merchants, its notion of “best for you” may quietly include “also best for my commission.” Without clear consent mechanisms, audit trails for agent behavior, and visibility into compensation flows between agent providers and merchants, trust will erode quickly. Treating regulation and standards as afterthoughts—as the article mostly does—is a convenient way to ignore how brittle this ecosystem becomes once real money and real data are involved.

Counter-argument and reply
There’s a popular rebuttal: agents will make discovery so efficient that niche and small sellers finally get a fair shot. I buy a limited version of that. If you tell an agent you want a very specific product or ethical profile, it can theoretically scan far beyond page one of a search result.

But discovery is a funnel, not a finish line. If the cost or complexity of integrating with major agent ecosystems is too high, many smaller merchants will exist only at the margins, surfaced occasionally but never fully plugged into automated reordering, bundling, or personalized promotions. Agents without low-friction, low-cost on-ramps become very polished discovery engines that keep routing demand back to the same advantaged suppliers.

Practical stakes merchants should demand
McKinsey sells agentic commerce as an opportunity. Merchants, especially smaller ones, should read it as a warning label too. There are concrete things they should push for: plain-language rules describing how agents route and rank offers; integration paths that don’t require enterprise-scale engineering; and clear separation between “paid influence” and “merit-based” recommendation inside agent systems.

Call this efficiency, progress, or just another cycle of platform consolidation. Agents will reorder how demand and supply meet; the interesting part will be watching who ends up writing, and quietly revising, the rulebook that McKinsey only sketches at the edges.