Agentic commerce has moved from buzzword to balance sheet, and the numbers prove it. 

• AI-driven commerce is projected to grow from under $50 billion this year to nearly $300 billion by 2030, rising from roughly 3% to 15% of total US e-commerce sales.

At EMARKETER’s Future of Digital Summit this week, EMARKETER analyst Sky Canaves led a panel with Bluefish CEO Alex Sherman, WPP Enterprise Solutions' Molly Schonthal, and the IAB's Caroline Giegerich who laid out the scale of the shift.

Retailers, not chatbots, may win first

Where that growth lands is the surprise as retailer-owned assistants take the larger share.

"More than half of these sales are not coming from the big AI platforms, the ChatGPTs and Geminis, but actually from retailer native assistants," Sherman said, “Think Amazon's Alexa for shopping or Walmart's Sparky.” 

• EMARKETER survey data backs that up: 27% of US digital shoppers have already used a retailer assistant, compared with 16% who've shopped through a general AI platform.

Retailers hold the edge on three fronts: proximity to the transaction, integrated fulfillment, and, Schonthal argued, the biggest advantage of all, unique access to brand data. Retailers can go directly to brand partners and demand richer product detail than a standard page provides, and she expects them to "really exploit that advantage" over the next six to 12 months.

Trust is the bottleneck, not technology

Panelists pointed to Visa research showing that consumer willingness to let an AI complete a purchase autonomously jumped from 23% to 61% once Visa itself was named as the party executing the transaction. 

Shoppers already trust their payment providers and retailers. They have not extended that trust to AI platforms, and that gap sets the pace at which agentic commerce scales.

The hidden accuracy crisis

Beneath the growth numbers, the panel flagged a quieter problem: AI accuracy. 

Schonthal cited inaccuracy rates of "anywhere from 10 to 20 percent across overall AI responses," with the worst spikes showing up in precisely the categories where precision matters most including financial services, healthcare, and pharma. That's pushing models toward a "flight to quality," including a documented pullback from lower-quality sources like Reddit in favor of brand-verified content.

Marketers should treat the brand as a "data surface," Schonthal said. Every channel an AI agent might pull from, including product pages, social, and earned media, needs to tell the same story. Contradictions become liabilities once an agent starts synthesizing across sources: "low caffeine" in one place and "zero milligrams" in another.

A framework for measuring AI visibility

To help brands track their footing, IAB’s Giegerich offered four measures worth watching: presence (are you showing up at all), prominence, portrayal (is the information accurate), and persuasion (does it actually drive action). Together, they map an awareness-and-consideration funnel that now runs partly through AI surfaces rather than a brand's own website.

Breaking down the silos

The distance between ecommerce and PR teams is a recurring failure point, said Sherman, drawing on their time at Warner Music Group, Estée Lauder, and Time Warner.

"Organizations don't do really well at communication across departments," he said. 

Schonthal framed the fix as building an "agentic marketing practice" around three pillars: insight into what drives AI performance, coordinated activation across teams, and measurement, which she said remains the most neglected of the three.

Giegerich added a practical closing frame for brand teams: get clear internally on the "where, what, when, how, and why" of the brand, since an AI agent is always answering a specific, contextual question rather than a generic one. 

The real alignment question for marketers, she said, is what questions the brand should be the inevitable answer to, and whether every team agrees on that answer.

Watch the full session.