The Next Retail Divide is How Retailers Operate

Every retailer I talk to has an artificial intelligence story. There's a new tool on the store floor, an AI marketing pilot, or a vendor pitching a solution for merchandising, pricing, or customer service.

I think the industry is asking the wrong question. The question is not whether a retailer uses AI. Nearly every retailer will. The more important question is whether AI changes how the company operates or simply gets added to the way it already works.

In my view, that will define the next divide in retail. Some companies will rebuild their operating model around machine assistance. Others will bolt AI onto old processes, incentives, and org charts. From a distance, those approaches may look similar. Inside the business, however, they will produce very different results.

More Tools, Same Silos

Most retail AI programs still live inside individual functions. A pricing team buys a tool. Marketing tests AI-optimized media. Store operations pilots computer vision. The CIO rolls out Copilot licenses.

There's nothing wrong with those projects on their own. The problem is that the operating model usually stays untouched. Teams still own separate pieces of the business, and no one owns the tradeoffs between them.

Merchandising makes assortment decisions on seasonal cycles. Media teams optimize channel-level return on ad spend. Inventory plans from historical sell-through. Store teams receive plans they had little role in shaping. Everyone has more dashboards and recommendations, but the important decisions remain scattered across departments.

What should launch? Where should inventory go? How quickly should a product be marked down? Which channel deserves the next dollar? Those questions still move from team to team, each with its own target and version of the truth.

I've seen this movie before: a performance review full of charts and explanations, but no clear owner for the economics of the whole business. AI produces more suggestions, while the organization remains responsible for less.

That's not transformation. It's a more sophisticated version of the same fragmentation.

Start With Decisions

Retailers that want to build an AI-based operating model should start with decisions, not tools.

I would begin by asking which decisions matter most, who owns each one, what information is required, and what happens when the decision is wrong. Many companies have never answered those questions clearly because their organizations grew around functions and systems rather than outcomes.

AI shouldn't become another department. It should become part of the way merchandising, pricing, inventory, e-commerce, and marketplace teams work together.

Machines can handle forecasting, pattern recognition, and scenario analysis. People still make the judgment calls. They decide which tradeoffs are acceptable, when to override the model, and who is accountable for the result.

That accountability matters. A model can recommend a markdown, price change, or shift in inventory. It cannot explain the decision to a brand partner or take responsibility when the economics go wrong.

This approach also requires less exciting but necessary changes: clear decision rights, data organized around customers and profitability, and incentives tied to profitable growth rather than performance inside one channel.

Without those changes, AI simply makes the old organization faster at passing work from one silo to another.

From Reports to Action

In a traditional setup, the report is still the end of the process. Teams wait for an end-of-week recap. Campaigns are reviewed once a month. Markdown strategies are revisited a few times each season. AI may produce the report more quickly, but the business still moves at the same speed.

I believe that's one of retail's biggest missed opportunities. AI shouldn't just help teams understand what happened. It should help them respond while there's still time to change the outcome.

A product begins to outperform its plan. A region starts accumulating excess inventory. An ad unit's returns fall below its margin target. Those signals should lead to a decision, not another slide in a meeting.

That could mean changing marketplace bids when margins tighten, moving inventory toward a market where demand is building, or adjusting a promotion as buying behavior changes. The point is not to automate every decision. It is to shorten the distance between a change in the market and a response inside the company.

The operator's job doesn't disappear. It becomes more demanding. Leaders have to know when the model is useful, when it's missing context, and when a short-term gain creates a larger problem somewhere else. They still own the outcome.

Marketplaces Show the Difference First

Marketplaces such as Amazon.com tend to expose this gap early because the feedback loops are fast and the variables are connected. Demand, advertising, pricing, ranking, content, and inventory all affect one another.

When a brand treats those areas as separate vendor assignments, AI can make the fragmentation worse. One partner optimizes advertising. Another works on content. A third manages operations. Each may report improvement in its own area, but no one can answer the more important question: Is the business gaining profitable share?

A marketplace operator needs one view of the business and one team responsible for the P&L. That team can use AI to process marketplace data, spot demand changes, test pricing scenarios, and decide where the next dollar is likely to do the most good.

People remain responsible for the more difficult decisions: which products belong in the mix, how much inventory risk to take, and how the marketplace fits with the rest of the business.

This is where I see the real value of an AI-based operating model. It's machine assistance paired with clear financial ownership. The technology matters, but the accountability matters more.

Redesign the Work

Retailers cannot change a legacy operating model by sprinkling tools over it. A Copilot license will not remove silos. A flagship AI project will not fix unclear ownership. A vendor demo is not a strategy.

The work starts with the decisions that drive the business. Clarify who owns them. Put the right data within reach. Shorten the time between a market change and a response. Measure the outcome that matters, not just the activity of each function.

Retailers that do this well may look, from the outside, as if they simply got more from AI. The bigger change will be harder to see: they will have rebuilt how the business operates. Everyone else will still be able to point to pilots, case studies, and press releases. They just won't have an operating model that turns those investments into an advantage that lasts.

The tools are becoming easier to buy. The operating model is the part retailers still have to build. That's why I believe the next retail divide will be between companies that change how they operate and companies that do not.

Anthony Connelly is founder and CEO of Neato, a second-party commerce operator that buys inventory directly from brands and manages execution across marketplaces and related channels.