Something shifted at Shoptalk Spring 2026. A year ago, the conference theme was "Retail's New Golden Age" — confident, forward-leaning, a little triumphant. This year's unofficial slogan, repeated by speakers, panelists, and exhibitors alike: "Keep humans in the loop."

That's not a retreat from AI. It's a more honest reckoning with what AI in retail actually looks like when it's operating in production rather than in demos.

From Theory to Production

PwC's Shoptalk recap captures the shift cleanly: "Retail AI shifted from experimental to operational, with AI systems live in production, generating measurable revenue, reducing costs, and automating decisions at scale." Retailers presenting results at Shoptalk 2026 made their investments in 2024. The implication for anyone still in evaluation mode is uncomfortable.

Specific deployments got airtime in ways that generic AI strategy sessions didn't last year. Macy's Chief Digital Officer's stated formula for retail success — "relevance, experience, and value" — was notable for what it didn't say: scale or efficiency, the keywords of the 2024 AI retail conversation. This year's success metrics were customer-facing, not cost-center-facing.

Agentic Commerce: Real, But When?

The most debated concept at the conference was agentic commerce — AI agents that can browse, compare, and purchase on behalf of consumers, bypassing traditional search and checkout entirely. The disagreement was not about whether it arrives, but when and how fast.

eMarketer projected 8.8% of e-commerce will be agentic by 2029. Merkle predicted up to 50%. That 600% gap in forecasts tells you something important: the underlying technology is advancing faster than anyone's ability to model consumer adoption curves.

The panel consensus, as Coresight Research documented, is that agentic commerce is better understood right now as a powerful discovery and research tool rather than an autonomous purchase-executor. David's Bridal's use case — AI agents helping brides navigate 300 wedding-related decisions over 18 months — represents a credible near-term application: high complexity, high stakes, long cycle.

The Infrastructure Problem Nobody Wants to Talk About

For agentic commerce to reach either forecast, retail data infrastructure has to change. AI agents can only purchase from retailers whose product data is structured, accurate, and agent-readable. Right now, most retail product catalogs are not built for machine consumption. January Digital's Shoptalk summary puts it plainly: "Retailers who make their product data discoverable and agent-readable will capture purchases that bypass traditional search and browsing entirely."

The retailers building that infrastructure today are making a bet that agentic traffic will matter by 2027 or 2028. The retailers waiting for the technology to mature before investing may find that the agents have already developed habits — and loyalty — toward early movers.

The Human Connection Wins

Chain Store Age's Shoptalk coverage highlighted what may be the most counterintuitive finding of the conference: in an AI- saturated retail environment, human connection has become a competitive differentiator, not a cost to be eliminated. Dutch Bros CEO summed it up simply: "We sell emotion, not coffee."

Victoria's Secret's presentation described a shift from prescriptive brand positioning to genuine customer conversation — using AI to listen better, not to replace listening with automation.

The implied warning for retailers building AI-first experiences: the tools are getting better at efficiency. They are not getting better at warmth. The retailers that figure out how to use AI to create more room for human connection — rather than substitute for it — may be the ones still standing when the agentic era fully arrives.

The decision window, as multiple conference speakers noted, is narrowing. The investment cycle for the results you'll be presenting at Shoptalk 2027 started yesterday.