February 23, 2026

Most agentic commerce is not agentic. JPMorgan made that case this month, describing current implementations as not actually autonomous but "another form of embedded shopping, similar to social commerce." That assessment, from one of the world's largest payment processors, should land hard. The industry has spent the past year branding every AI-adjacent shopping feature with the word. JPMorgan's analysis maps the real evolution in three stages, from discovery-only agents that recommend products to web-crawling models that simulate guest checkouts to protocol-based direct integrations where agents transact natively through standardized APIs. Most implementations sit somewhere between the first two. Autonomous purchasing, where agents complete transactions without human approval, "will take even longer" to achieve at scale. "It took more than two decades to build the digital commerce ecosystem we see today." Faster evolution is expected, but JPMorgan warns it "won't be at scale tomorrow."

Publicis Sapient's Simon James disagrees on pace. He predicts agentic commerce will reach 10% of retail in three to five years, far faster than the two-plus decades e-commerce took to reach 16% of global retail sales. James points to the distribution infrastructure already in place. ChatGPT serves 800 million weekly active users. Google AI reaches 1.5 billion monthly. He highlights Shopify's September 2025 OpenAI partnership as an early-mover case, arguing it is "disproportionately important to the valuation of their business overall" even if it barely moves quarterly revenue. His warning to slower players is blunt. "If you sit out for two or three years, you're at generation zero." Early movers generate data exhaust that improves their systems while late entrants face the same learning curve on a compressed timeline with competitors already controlling market position.

Small businesses are not waiting for the debate to settle. The US Chamber of Commerce reports nearly 60% now use AI, up from roughly 30% in 2023, with 84% of high-tech small business adopters reporting sales and profit gains. Eighty percent are accelerating technology adoption specifically because of competitive pressure. The Chamber analysis shows AI drove 20% of retail sales during the 2025 holiday season, generating $262 billion. Chatbot traffic doubled year-over-year. Search conversations with AI run two to three times longer than traditional queries, which means users are spending more time inside AI environments before they ever reach a merchant site. Coresight Research's John Harmon, quoted in the report, puts the implication plainly. "If you're not sharing your product information with the chatbots, you're at a big disadvantage."

B2B procurement is outpacing retail in the march toward genuine autonomy. CPOstrategy describes transactional procurement going from "near touchless" to "fully autonomous," with buyer and supplier-side agents negotiating, placing orders, managing compliance, and resolving exceptions through digital dialogue without manual touchpoints. The same analysis highlights Zip, which launched 50 purpose-built AI agents spanning procurement, finance, legal, and security. Not one department's copilot. Cross-functional orchestration where approvals become "trust-based, outcome-led, and exception-driven," and agents initiate actions unless told otherwise. Tonkean takes a different cut at the same problem, positioning itself as agentic orchestration for the Fortune 1000 and claiming its agents can autonomously pursue long-term business goals without undermining human agency. CPOstrategy's Andrew Woods captures where this leads. "Tomorrow's insights won't be extracted. They'll be embedded and automatic."

Payment infrastructure is catching up. PYMNTS frames the shift as "commerce triggered by prompts, not clicks," with dynamic tokenization enabling agents to transact through single-use, real-time tokens rather than stored card credentials. Visa partnered with PYMNTS on a "Prompt Economy" tracker that positions this as the next credential evolution. Agents don't need card numbers. They need cryptographically scoped permissions that authorize specific transactions within defined constraints. PYMNTS reports that credential-on-file systems are already improving approval rates and reducing fraud for merchants who adopt them, and describes agentic transaction infrastructure as "rapidly maturing."

B2B marketplaces face a more fundamental shift. As agents take over sourcing decisions, competitive advantage moves from rich imagery and narrative copy to precise, unambiguous metadata. Specifications, compatibility data, warranties, return policies, and real-time availability now determine whether an agent even considers a listing. Payment terms, credit availability, and settlement speed become "central inputs to the purchasing algorithm itself." A PYMNTS Intelligence study found a stark readiness divide. Seventy-five percent of technology firms report extreme familiarity with agentic AI, compared to 33% of goods firms and 38% of services firms. Merchants whose data is already structured for machine consumption will be the ones purchasing algorithms evaluate first.

JPMorgan is right about the current state. Most agentic commerce is embedded shopping wearing a new label. But underneath, payment networks are tokenizing for agents. Procurement platforms are automating multi-round negotiations across buyer and supplier organizations. B2B marketplaces are re-architecting around metadata over merchandising. The gap between the label and the reality will close faster than the two decades it took to build traditional digital commerce. Merchants without agent-grade data, identity, and payment infrastructure will not just lose market share. They will lose visibility to the algorithms that increasingly decide where money flows.