AI dominates nearly every board meeting, executive retreat, conference, and strategic planning session these days. Retail leaders are evaluating platforms, launching pilots, and asking their teams to find use cases that promise measurable value.

The urgency makes sense. No one wants to fall behind.

But after the excitement fades, a lot of companies land on the same question:

“We’re doing a lot with AI, but are we seeing the business impact we expected?”

The instinct is to blame the models, the data, or the technology. It’s rarely any of those. The problem started long before AI entered the conversation.

Right question, but a single answer isn’t the answer.

Start with the outcome. Fix your data. Redesign the workflow. Each of those is right, and each gets pitched as the fix. None of them is the fix by itself. Retail has never been simple enough for one answer to carry a whole strategy and treating any single one as the answer is how you end up with a stack of pilots and a business that hasn’t moved.

Plenty is changing. The scoreboard stays the same.

Consumer behavior is shifting as AI reshapes how people discover, compare, and buy. Product discovery is being rebuilt around agents that never load your site, and the traffic coming through them is growing faster than any other source. Expectations will keep moving. Channels will keep moving. Anyone telling retailers to sit still is wrong.

But the customer at the pump doesn’t care about the AI that’s saving you five basis points on a Visa transaction. They care whether the pump worked, whether the price was fair, and whether the line moved fast enough to get them to work on time.

Margin. Catalog exposure. Ease of doing business. Service. Cost to serve. Retention. Dwell time. That list has been the same for decades, and it will be the same in a decade. AI changes almost everything about how you hit those numbers, while changing nothing about which numbers count.

The harder question

Most retailers already know what they want to improve. Ask a CFO what matters, and you’ll get an answer in about four seconds.

The harder question is “Who in the organization knows how to move it?”

• Somewhere in your business, someone knows when to take the markdown and when to hold.
• Someone knows why the auth rate dipped in a region and what to do about it.
• Someone knows which interchange qualification is leaving money on the table.

That judgment is real, but it’s also stuck. It lives in a handful of people who can’t be in every store, on every transaction, every day.

Where AI earns its keep is in distributing judgment at scale. Taking what your best people know and making it repeatable, consistent, and present at the point where decision gets made.

That comes with a catch. You can’t scale what was never articulated. The merchant who knows when to mark down often can’t tell you why. Getting that out of someone’s head and into something a system can act on is the real work, and it’s harder than any tool selection.

Judgment doesn’t work alone, either. It needs data it can act on, a path that carries the decision to the store or the terminal, and someone accountable for the call once it gets there. Miss any one of those and the judgment stays where it started.

Where this leaves you

The job has never been to predict the future perfectly. It’s to be positioned when it arrives. Whatever the next three years look like, the thing that determines whether you get there is the same: whether the judgment that runs your business can reach the places where decisions actually happen.

In our next Point of View, we’ll get into what has to be true inside a business before that judgment can scale, and why most readiness checklists ask the wrong questions.

In the meantime, here are two questions worth sitting with:

Who are the three people in your business whose judgment you’d want present in every store, every day? And could you explain to anyone else how they decide?

Let’s Get Started