NRF 2026: Retail's Big Show - It’s the data, stupid
Artificial intelligence is only as "smart" as the data supporting it.
The annual National Retail Federation conference and exposition took place January 11-13 at the Jacob J. Javits Convention Center in New York City. In my preview column for NRF 2026: Retail’s Big Show, I said NRF conferences fall into two categories – “gee whiz” shows that speculate on what technology might do and practical shows that focus on what technology does.
I predicted that the 2026 edition would fall into the second category and I was right. I wasn’t completely correct – the lack of humanoid robots other than a proof-of-concept android greeting visitors to the Verizon booth laid to rest any predictions of a "Star Wars" scenario where humans and robots routinely interact on the horizon – but this was one of the most practical NRF conferences I’ve ever attended, dating back to 1999.
[READ MORE: How humanoid robots may bring retail closer to ‘Star Wars’ after all]
Of course, Agentic AI was on everyone’s mind. But to sum up the theme of the show as simply being "agentic AI" would be oversimplifying matters.
Instead, exhibitors and attendees alike focused on how retailers can ensure they achieve ROI from agentic AI investments and make the most of this still-emerging model, which builds upon the prescriptive capabilities of generative AI to streamline enterprise workflows by analyzing massive amounts of data in near-real-time and automatically taking action based on the results.
To put it briefly, the answer is, "It’s the data, stupid." Agentic AI solutions are only useful if the right data is fed into them and then used to perform the correct tasks. Let’s take a deeper look at what this means in practice:
Garbage in, garbage out
"Garbage in, garbage out" (or GIGO) is a phrase that goes all the way back to military computer programmers in the 1950s. Quite simply, it means that a computer is only as good as the information flowing into it, and poor data leads to poor calculations, recommendations and decisions.
Next-gen AI solutions are a lot more advanced than the vacuum tube-based mechanical behemoths that occupied entire rooms in the 1950s. But they still are completely reliant on the data fueling their algorithms, calculations, decisions and actions.
No matter how sophisticated or "smart" a retail AI solution may be, if it is provided incorrect, outdated, irrelevant or otherwise faulty data, its output will be unhelpful at best and harmful to the business at worst.
Before deploying an AI solution, retailers need to first determine what types of data they will need to process, where that data is being sourced, and how it will be reviewed, filtered, organized and synthesized for further AI analysis and operationalization.
Respect your workflows
Once you have established a secure, steady flow of high-quality data, you need to ensure that any AI solution you implement fits into your established workflows.
All too frequently, businesses will see a "shiny, new" technology such as generative or agentic AI and decide they need to implement it within their enterprise as soon as possible to get a jump on their competitors. But this approach to AI is self-defeating.
Instead, you should evaluate any AI solution from the perspective of how your organization already operates and see how it would make existing workflows faster, more accurate, more efficient, or more productive. AI is there to amplify what you already do. If implementing AI would require changing core processes or not really fit into your existing architecture, find a different solution (or don’t implement one at all).
Make the results matter
Finally, even when an AI solution fits into your enterprise and your workflows, it will not produce ROI unless you can take its output and achieve something. The results need to matter.
Results that matter can take many forms. They might be more accurate supply chain forecasts, or higher resolution rates for customer service inquiries, or stores that are optimally staffed at all times of day.
In addition, "phygital" retailers seeking to seamlessly blend brick-and-mortar and virtual channels can effectively unite all the different data streams required to provide a unified shopper experience using leading-edge AI.
In my previous column, I said that the Big Show would reveal which of several popular phygital retailing models is taking precedence. It turns out the specific model is less important than having the correct data-driven foundation underneath.
Perhaps the biggest lesson of NRF 2026: Retail's Big Show is that AI isn’t so much about doing something new; it’s about doing the same thing in a new way.
