NRF 2026: Agentic AI Comes to Physical Stores — And Most Are Not Ready

At NRF 2026 in New York, the phrase “agentic AI in retail” was everywhere. From the keynote stages to the Javits Center floor, leaders from Google, Microsoft, Shopify, and Stripe laid out a future where AI doesn’t just assist shoppers — it shops for them. Autonomous agents that research, compare, assemble baskets, and complete transactions without human input.

The demos were impressive. The ambition was real. But as someone who has spent the past three years deploying physical retail infrastructure across four countries, I kept asking the same question the entire week: is the store floor actually ready for any of this?

The honest answer is no. Not yet.

The Gap Nobody Talked About on Stage

Here is what the keynotes got right: agentic AI represents a structural shift. Google announced a Universal Commerce Protocol with Walmart, Shopify, and Target. Microsoft launched Copilot Checkout. OpenAI partnered with Target to let shoppers build carts and check out directly inside ChatGPT. These are not experiments. They are infrastructure moves with real capital behind them.

But here is what the keynotes left out: agentic AI needs to talk to your store in real time. It needs accurate inventory data, live pricing, and a POS that can process transactions initiated by a machine, not a human. For that to work, your kiosks, your electronic shelf labels, and your POS terminals need to be connected, updated, and feeding clean data into a system that AI can actually trust.

Nikki Baird, VP of Strategy at Aptos, said it plainly from the NRF floor: “You’re not going to get very far if your data is not good.” Nedap’s General Manager made the same point: retailers who expect agentic AI to perform well on fragmented, siloed data will find themselves disappointed.

I have seen this firsthand. In several deployments across Latin America and the U.S., we have found retailers whose electronic shelf labels run prices that differ from their POS by 5 to 15 percent. Inventory counts update once a day, not in real time. Kiosks run software that nobody has patched in over a year. None of that is compatible with an AI agent making purchase decisions on behalf of a customer in seconds.

What “Agentic Ready” Actually Means for Your Store

There is a checklist that most NRF conversations skipped entirely. Before any retailer invests in agentic AI pilots, four operational questions need honest answers.

Four Questions to Ask Before Any AI Investment

First: does your inventory data update in real time? AI agents need a reliable map of what is in the store right now, not what was there yesterday morning.

Second: do your POS and shelf label systems stay synchronized? Price discrepancies between shelf and checkout already hurt customer experience. For an AI agent completing a transaction autonomously, that gap becomes a failure point that breaks the entire flow.

Third: can your store-level hardware handle the processing load? Cisco’s analysis ahead of NRF noted that agentic workflows generate significantly more network traffic than traditional checkout systems. Your edge infrastructure needs to support that load.

Fourth: does your operation have an open API layer that external systems can connect to? Without it, no AI agent can plug into your store regardless of how sophisticated the software is.

These are not software questions. They are hardware and infrastructure questions. Retailers should direct them at their technology partners right now, not after signing an agentic AI contract.

The Store Floor Is the Foundation, Not the Last Mile

The most consistent message from NRF 2026 was that this is “no longer future talk.” AI is moving from pilots to production. That is true. But production-ready AI and production-ready stores are two different things, and the gap between them shows up in kiosks, shelf labels, and POS systems.

The retailers who benefit from agentic commerce will treat their physical infrastructure as a prerequisite, not an afterthought. Clean data. Connected systems. Hardware that communicates with the stack above it.

The Path to Agentic Readiness: Where to Start

Getting to agentic readiness doesn’t require replacing everything at once. It requires sequencing the work correctly. Here is how I advise retailers to approach it.

Start with a Data Audit, Not a Technology Purchase

Before signing any AI contract, map your current data flows. Where does inventory data live? How often does it update? Does your POS communicate with your ESL in real time, or do those systems run independently? The answers tell you more about your agentic readiness than any vendor demo will.

In my experience, most retailers find two or three critical disconnects in this audit that nobody knew existed. One retailer we worked with discovered that its price update cycle ran 48 hours behind across 30% of its shelves. That gap was invisible in day-to-day operations but would have completely broken any AI-driven pricing or inventory agent.

Prioritize Connectivity Over Capability

The instinct when modernizing is to buy the most capable hardware available. But for agentic AI, connectivity matters more than processing power right now. A kiosk that pushes and pulls real-time data via an open API delivers more value today than a faster processor running in isolation.

Ask every hardware vendor the same three questions: Does this device have an open API? What data does it expose? How frequently can it sync with external systems? If they can’t answer clearly, that is your answer.

Evaluate Upgrade Versus Replace Honestly

Not all legacy hardware needs replacement. Some kiosks and POS systems from the past three to five years support the connectivity agentic AI requires after a software or firmware update. The question is whether the underlying architecture allows it. My general rule with clients: if the hardware supports real-time API calls without a complete system rebuild, upgrade it. If the architecture is fundamentally closed, plan a replacement cycle over 18 to 24 months and start with your highest-traffic locations.

Have the Budget Conversation Now

The most common mistake I see is retailers approving an agentic AI pilot budget without approving the infrastructure budget that makes the pilot work. Those two line items belong together in front of leadership. An AI pilot running on unprepared infrastructure will underperform, and the lesson the team takes away will be “AI doesn’t work in our stores,” rather than the correct one: the foundation wasn’t ready.

The retailers who will win in the agentic era are building that foundation today, quietly, while their competitors are still watching keynotes. The agentic era is coming to the store floor. Whether your store is ready to receive it depends on decisions you make this year, not next.

If you are evaluating your retail technology infrastructure in light of this shift, connect with me here or reach me on LinkedIn.

---

Adriana Rivas is the COO and Chief Biwitech Development Officer at BIGWISE Corp, founder of the retail hardware brand Biwitech, and award-winning author of How to Implement Self-Service Without Failing (Amazon #1 Hot New Release, Silver Nonfiction Book Award 2025, Gold Stevie Award winner for Thought Leader of the Year 2026). She designs and deploys self-service, POS, and digital signage technology across the U.S. and Latin America.