NRF 2026: AI development gets practical

Technology providers are looking to push AI into retail back offices and stores, but only if retailers fix the plumbing first, writes Miya Knights, Retail Technology Publisher.

Across vendor briefings at NRF 2026: Retail's Big Show this week, a set of recurring themes cut through the hype: AI is being pushed hard into store and operations workflows. 

However, vendors were increasingly insistent in their interviews with Retail Technology magazine that outcomes depend on data quality, interoperability and governance foundations. 

"Return on intelligence"

Ben Gale, Diebold Nixdorf Retail Senior Vice President, EMEA and APAC, framed the shift as moving "away from touch point technology to try and drive intelligence… a return on intelligence as well as a return on investment".

He said the company is applying AI "pretty much in every single element" — from fraud prevention and product recognition to age verification — to reduce friction and interventions at checkout and self-checkout.

Gale argued that retailers are no longer patient with long-dated paybacks: "What we're trying to do is… add bottom-line value today so that retailers don't have to wait… years to get a return on their investment." He also positioned Diebold Nixdorf's approach as a platform play rather than a set of point solutions: "Some people try to solve individual problems… our desire is to build out a platform… [that] can combine data so you could see someone in an aisle or someone at a checkout."

That same "practical AI" lens came through strongly in the vendor's focus on loss prevention and store safety, including in-aisle computer vision analytics designed to detect escalating incidents and trigger appropriate responses, while also protecting privacy through facial imagery redaction and on-site data processing.

Data accuracy and trust 

Several briefings converged on a simple message: retailers cannot "AI their way" out of poor source data. Andres Avila, Global Retail Strategic Marketing Lead for Honeywell Productivity Solutions & Services, positioned the mobile computer, printer, data-capture hardware and software provider's role as "oriented to provide accuracy in the data that we are capturing in day-to-day operations". He argued that "all the algorithms… would have the best decisions based on the… high-quality data".

Avila also challenged the industry's narrative about modernisation: "The question for the retail industry is not being modern, it's being profitable." In his view, retailers need "dynamic commerce". That is, flexible store and supply chain operations that adapt to volatility and are underpinned by accurate, real-time data to deliver consistent experiences across channels.

From the systems integrator side, Ruth Harrison, Global Retail & Consumer Goods Executive for Microsoft-focused IT services consultancy, Avanade, was equally direct that AI programmes fail without foundations: "Your AI is not going to be effective… if you don't have solid foundations. You need to know that the data can be trusted and protected, with good governance structures."

Harrison added that retailers should think pragmatically about build-versus-buy: "If it's core, just buy it… where you want to differentiate, you build."

Store associate as interface

In-store enablement emerged as a significant "why now" driver for AI investment. Jim Musco, Zebra Technologies Industry Principal, Retail North America, positioned the vendor's agentic workflows as a way to reduce task-switching and accelerate frontline execution, supporting associates from back-of-store receiving through to customer-facing service.

Critically, Zebra's conversations kept circling back to deployment realities. Musco said retailers' existing "systems of record" and the state of their data determine how straightforward integration will be. That point was reinforced by Rob Desantos, Vice President, Infrastructure and Workplace Solutions, Total Wine & More. 

Desantos described early-stage progress in his pilot with the Zebra Companion (part of the Zebra Frontline AI Suite) but cautioned against unleashing agents onto messy source systems: "I don't think [the] plan is… to turn this agent loose on your data where it sits… unless it was super clean. But if you have curated data that is clean, you can integrate."

On the customer-facing side of store journeys, Clifford Perlman, Jumpmind Growth and Partnerships Vice President, argued that retail still undermines its own selling capabilities by forcing a cold, transactional moment at checkout: "You've now established a barrier — a cash wrap — and now that whole interaction has a barrier between you and them." 

Perlman said Jumpmind's CX Connect was designed to keep shoppers engaged through checkout via a customer-facing digital screen that supports loyalty identification, personalisation prompts and, now, integrated payments, reducing device sprawl while keeping the associate focused on service.

Standards avoid "agent soup”

If NRF 2026's buzzword was "agentic," its subtext was "integration". The industry's emerging agentic commerce standards are attempting to reduce the bespoke work required to connect new AI shopping surfaces to retailer systems. 

Google's newly announced universal commerce protocol (UCP) — introduced with partners, including Shopify and several large retailers — is explicitly positioned as an open standard to help AI agents complete commerce journeys across discovery, checkout and service. 

Jose Gomez, Google Cloud Retail & Consumer Managing Director, told Retail Technology that retailers' imperative is to meet customers wherever they now choose to engage: "The focus on the customer hasn't changed. The customers are just in different places than where they used to be… they didn't use to be on mobile, they didn't use to be on search, and they didn't use to be with agents, and so retailers need to show up there."

Avoiding risk and operational burden

Jason Cottrell, MACH Alliance President and CEO of composable commerce consultancy and systems integrator, Orium, warned that without coordination, "it's going to be this messy agent soup… if we let things go as they are". 

He shared a concern that retailers and their IT providers will build agents "in silos" around individual back-office or customer-facing use cases, multiplying complexity for retailers.

Cottrell argued that composable architectures should function as "the best way to lay those foundations for today and for the agentic era," particularly as most enterprises will run "existing enterprise platforms alongside… composable-oriented innovation".

The Alliance leader also highlighted risk and operational burden as agentic commerce expands: the "pace of innovation across multiple standards" can create new failure points and "additional attack vectors," making it harder for retailers to maintain reliability without a more modular approach to switching integrations on and off.

Collaborative standardisation efforts

At the same time, the Order Network eXchange launched under the Commerce Operations Foundation, a standard governance nonprofit, to provide a shared operational standard for commerce fulfilment networks. The Foundation has brought a broad cross-section of the commerce ecosystem together alongside its founder, Kelly Goetsch, President of Pipe17 and President of the Foundation. 

Foundation members include order management system (OMS) providers Manhattan Associates and IBM Sterling, 3PL and logistics providers Radial and Ryder, composable commerce platforms such as commercetools, system integrators such as Orium, and brands and retailers Allbirds and The Vitamin Shoppe.

The bottom line: NRF vendor conversations repeatedly returned to the same conclusion. AI may be the headline, but data foundations, operating model change, and integration discipline are the decisive factors separating "demo theatre" from beneficial, measurable, store-level outcomes.