Agentic Commerce: Where Adoption Actually Stands

What's happening

A year ago, agentic commerce was mostly announcements: assistants that could theoretically browse, compare, and buy on a user's behalf, checkout protocols being drafted, and a lot of speculative "this changes retail forever" commentary. The picture now is more mixed and more useful — real transaction volume flowing through assistant-mediated purchases in a handful of categories, alongside plenty of infrastructure that's live but barely used because the user habit hasn't caught up to the capability.

Why now

The gap between "technically possible" and "actually adopted" is closing unevenly by category. Straightforward, low-consideration, replenishment-style purchases — the kind where a user already knows exactly what they want and just needs it reordered or price-compared — are where assistant-mediated buying has gained the most real traction. High-consideration purchases, anything involving fit, taste, or brand preference, are moving much slower, because users still don't fully trust an agent's judgment on things they'd normally research themselves, and merchants remain wary of losing the storefront experience entirely.

What it means for marketers

The honest read for most marketing teams: agentic commerce is real enough to prepare for, not urgent enough to reorganize around. A few things are worth doing now regardless of category:

Structured product data has become a prerequisite, not a nice-to-have. Agents making purchase decisions rely on clean, machine-readable product feeds — accurate specs, current pricing, real availability — far more than they rely on persuasive on-page copy. Product data hygiene, historically a low-priority ops task, now has a direct line to a new purchase channel.

The influence layer matters more than the transaction layer, for now. Even where an agent doesn't complete the purchase, it increasingly shapes the shortlist a user considers — comparing options, summarizing reviews, recommending a category leader. Showing up in that comparison set, through the same structured-data and citation-worthy-content practices that matter for AI search visibility generally, is the higher-leverage near-term play compared to building for checkout-completion agents directly.

Attribution is still broken here, and that's not getting fixed soon. When an assistant researches, compares, and initiates a purchase across sessions and sometimes across platforms, the referral trail that classic attribution models depend on frequently disappears. Teams should expect a growing "dark" share of purchases with no clean attribution path, and treat that as a modeling problem rather than a data-collection failure to be engineered away.

Category matters enormously. Commodity, subscription, and replenishment products should be actively testing agent-readiness now. Considered-purchase and brand-driven categories can reasonably wait another cycle or two — the infrastructure will still be there, and the user behavior is what's actually missing, not the technology.

Watch signals

• Checkout and payment protocols for agent-initiated purchases moving from draft standards to actual merchant adoption at scale
• The share of a category's purchases that show measurable agent involvement, even short of full checkout
• Merchants publishing agent-specific product feeds as a distinct workstream from their standard catalog feed
• Attribution and analytics vendors shipping features specifically for agent-referred traffic, rather than lumping it into "direct" or "unknown"

Treat this as a slow-building infrastructure shift, not a channel to chase this quarter — the prep work (clean data, comparison-set visibility) pays off regardless of how fast the transaction volume actually arrives.