Investment in AI is accelerating at a frantic pace, and efficiency gains are promised everywhere. But meaningful ROI for retailers hasn’t yet universally materialized.

Now, the conversation is shifting from what AI can do to why it feels more like hype than reality. Increasingly, the answer comes back to the same issue: innovation is moving faster than the systems and data structures required to support it.

AI is Making an Impact Today, in Certain Areas

This isn’t to say AI is having no impact today. It’s just happening in environments where the problem is narrow, the data is well defined, and outcomes can be measured in real time.

Loss prevention is a clear example. Computer-vision-based systems have become effective at identifying abnormal patterns at checkout, such as missed scans or unusual transaction behavior, and triggering timely intervention. In these contained scenarios, AI is already helping retailers meaningfully reduce shrink because the decision context is clear and the feedback loop is immediate.

AI will only perform as well as the commerce foundation beneath it; fragmented systems create fragmented intelligence and uneven customer outcomes.

The challenge changes as retailers move beyond these isolated use cases. Applications like hyper-personalized experiences, shelf compliance, planogram adherence, and staffing optimization span multiple systems, roles, and time horizons. Their effectiveness depends on structural and operational alignment. To work, they need a shared understanding of products, pricing, inventory, and who is responsible for acting. When that alignment is missing, even strong models struggle to be effective or scale properly.

Fairly or not, that gap between promise and execution is why the conversation around AI has grown more cynical.

The Problem with Fragmented Data

This gap exists because most retail tech stacks are highly complex. For decades, retailers adopted best-of-breed systems to solve immediate needs. Each additional solution delivered new capabilities and created important new value in the moment.

But over time those individual additions created fragmented data, duplicated logic, and inconsistent execution across channels. This is a big problem when table-stakes capabilities like self-checkout and personalized promotions depend on shared transaction logic, accurate inventory, and consistent pricing rules across the enterprise.

AI Magnifies the Problem

For years, retailers lived with this fragmentation because they could. Between a complex web of APIs, manual workarounds, reconciliation processes, and exception handling, running the business in a way that met customer expectations was fully possible, even if it was an internal headache.

But because of the way AI works, that’s no longer true.

AI depends on consistent, accurate, real-time data to deliver value. When inventory, pricing, and transaction logic differ across systems, AI can’t navigate the nuances and compensate for those gaps. Instead of making things easier and more efficient, it amplifies the issues and makes them worse.

In an AI-driven operating model, those same issues actively undermine performance by automating the wrong decisions, personalizing based on flawed signals, and accelerating inconsistency across channels. And you won’t necessarily have a person there to catch them.

That turns what should be a massive opportunity into a direct barrier to progress.

Without a unified foundation, AI introduces more risk than reward.

Going Back to Basics

This tension surfaces constantly in conversations with retail leaders. Many are eager to pursue AI-driven forecasting, dynamic pricing, and more personalized engagement, but they aren’t yet able to answer basic operational questions:

• Which system is authoritative?

• Why does available inventory differ by channel?

• Why does pricing logic change depending on where a transaction starts?

AI assumes and requires trust in the underlying system. Instead of troubleshooting discrepancies, it just rolls with whatever it decides is most likely correct. As you can imagine, that creates significant margin for error.

As a result, priorities are changing. More leaders are moving away from asking, “What should we add next?” to asking, “What needs to be simplified?”

Unified Commerce as an Enabler, Not an Endpoint

In this environment, competitive advantage is determined less by which innovations retailers adopt and more by how reliably they can deploy, govern, and scale them across stores and channels.

This isn’t an argument for choosing stability over innovation. Retailers need both. However, a unified commerce foundation creates the only conditions in which innovation can effectively scale.

When new technologies can plug into consistent data and logic instead of competing with it, retailers gain the confidence that the intelligent tools and capabilities they launch will behave as expected in real world.

That leads to faster rollouts, cleaner integrations, more reliable customer experiences, and fewer surprises when new initiatives move from pilot to production.

What Will Separate Leaders in 2026

Some of the most exciting conversations I’m having today are about what the pace of AI innovation will enable in the next five or ten years.

But the most productive conversations are the ones focused on what it takes to create the unified foundation that enables those grand visions.

Unified commerce may not be the most visible force shaping retail in 2026. But it will likely separate the innovations that deliver results from those that never live up to expectations.