A measurement gap hiding inside a growth forecast
The headline finding from this Visa Acceptance Solutions-commissioned PYMNTS Intelligence report is a genuine warning sign hiding inside what otherwise reads as a growth story. Only 23 percent of merchants can clearly identify both AI-driven traffic and the purchases that traffic actually generates. Another 21 percent can see that agent traffic exists but cannot connect it to completed sales at all. That means a majority of merchants are already operating a commerce channel they cannot properly measure, at exactly the moment that channel is beginning to scale.
The size gap compounds the problem. Large merchants identify agent-driven purchases at 27 percent versus just 19 percent for small and mid-sized businesses, meaning the retailers with the least visibility into this emerging channel are also the ones with the fewest resources to build the measurement infrastructure needed to close that gap. Without intervention, agentic commerce risks becoming another advantage that compounds in favor of already-large players, purely because they can afford the analytics investment smaller competitors cannot.
Why building the flashiest agent is the wrong first investment
The report's core insight deserves to be the headline rather than a supporting point: the real competitive advantage in agentic commerce comes from building the measurement systems needed to actually track agent activity across the commerce funnel, well beyond having the most sophisticated shopping agent on your own storefront. That is a genuinely counterintuitive priority for retail technology teams accustomed to competing on customer-facing feature sophistication rather than backend attribution infrastructure that never shows up in a product demo.
It makes sense once you consider what an agent actually does differently from a human shopper. An AI agent may compare your product against competitors, evaluate your pricing, and make a purchase decision entirely outside a human-visible browsing session, through channels and interaction patterns your existing web analytics were never built to capture. A retailer investing heavily in an impressive on-site shopping assistant while ignoring the attribution gap is optimizing the wrong end of a funnel it cannot see the middle of.
What merchants expect agents to actually do at checkout
The behavioral expectations merchants report are specific enough to plan around directly. 61 percent expect AI results to influence purchase decisions more than traditional search, a meaningful shift in where retailers need to focus discovery and search engine optimization investment going forward. 58 percent anticipate agents will select payment methods based on fees and rewards structures rather than a human's stored preference or habit, which has direct implications for how payment options are surfaced and prioritized in checkout flows built for agents rather than humans.
Most significantly, 56 percent expect agents to complete transactions fully autonomously, without a human confirming the final purchase step at all. That expectation alone should be driving a rethink of checkout security, fraud detection, and confirmation workflows at any retailer processing meaningful ecommerce volume, since a checkout flow designed around a human pausing to review an order summary does not map cleanly onto an agent executing a purchase without that same pause.
The two-year revenue number that should reset budget priorities
The most consequential figure in the report is the 38 percent of merchants who forecast agent-driven purchases will exceed 15 percent of their total sales within two years. That is not a speculative, far-future prediction, it is a near-term revenue forecast from merchants themselves, and 15 percent of total sales is a large enough share that most retail finance and technology leadership teams would treat any other channel commanding that share as a standing budget line with dedicated headcount and measurement infrastructure.
Any retailer that currently treats agentic commerce as an emerging trend worth monitoring rather than an active revenue channel worth measuring today is likely to find itself two years behind exactly the peers who took the 38 percent forecast seriously enough to invest in attribution capability now, while the channel is still young enough that building measurement infrastructure is a manageable project rather than an urgent catch-up effort.
What to build first
The report's practical recommendation is specific enough to act on directly: develop attribution capabilities that connect agent activity to completed purchases, payment method selection, loyalty program usage, and total transaction value, building early visibility into this channel before it scales further. That means auditing your current analytics stack specifically for its ability to distinguish agent-originated traffic from human traffic, and identifying wherever it currently cannot make that distinction cleanly.
For retail CIOs and CTOs setting technology priorities for the next budget cycle, this report is a strong argument for moving attribution and agent-traffic measurement ahead of customer-facing agentic commerce features on the priority list, at least for now. You cannot optimize a channel you cannot measure, and right now the data says most of the industry, including a majority of large merchants with real resources to spend, still cannot measure this one.
