Antonio Mires Gambetta
Director – EREA Management Consulting
*Translated with DeepL AI
Is artificial intelligence streamlining operational decision-making, or is it replacing the store manager’s judgment without us even noticing?
Artificial intelligence has begun to change the way store managers and supervisors manage their daily operations. Today, it is possible to obtain performance indicators, sales comparisons, or restocking projections in seconds that previously required hours of manual analysis. That speed is real. The problem isn’t the speed of the responses; it’s what happens to the person receiving them.
When a number arrives quickly, well-presented, and appearing precise, the brain tends to assume it is correct. Verification time is reduced, operational questioning diminishes, and the responsibility for interpreting the data is delegated without anyone having decided to do so. In store management, a misinterpreted metric at the start of the day can lead to incorrect restocking, misaligned staffing, or business actions that exacerbate the problem being addressed, with a direct impact on lost sales and margin erosion.
The paradox is clear: AI increases response speed, but it can reduce the quality of operational judgment if used as an oracle rather than an assistant. That distinction changes everything.
The reality on the shop floor: analytics arrive pre-processed
Before discussing how to ask the right questions, we must identify the real problem. In most current retail operations, the store manager is not facing a conversational AI analyzing their data in real time. They are solving three simultaneous problems with an incomplete team, receiving a corporate report that has already been processed by someone else, and reading a dashboard of indicators that displays green and red lights without explaining why.
That is the most common and least discussed risk: not that the manager asks the AI the wrong questions, but that they consume, without questioning, information that has already been interpreted by previous layers. Their own judgment doesn’t disappear overnight; it erodes shift by shift, week by week, until the manager operates on assumptions they no longer verify because “it’s always worked that way.”
AI amplifies this risk because it increases the volume and speed of that processed information. More data, faster, with better visual presentation, generates greater confidence in the recipient, regardless of whether the underlying interpretation is correct. The problem isn’t the tool; it’s the lack of judgment and time to question it.
When the number is correct but the interpretation is wrong
There is a specific danger in using analytics and AI for operational management: obtaining technically accurate information, but with an incorrect interpretation of the ratios; the trend or the real causes of the problem. It is the most dangerous combination: true data, wrong conclusion.
A positive sales trend may be masking a silent erosion of margins due to a shift in the product mix. If customers are buying more volume in lower-margin categories, sales revenue rises and profitability falls. The system reports growth; the store is losing efficiency. An operating cost within the historical range can be a trap if that range was established under staffing, rent, or energy conditions that no longer exist. Being “within range” is not good if the range is already miscalibrated. And a drop in average ticket size may be attributed to the product mix when it is actually a problem with the sales skills of the team on duty—or vice versa. The cause matters just as much as the indicator, and confusing the two leads to actions that solve nothing.
In all these cases, the number is correct. The problem lies in interpreting it without context, in automatically reading an indicator without asking what lies behind it. When that interpretation is generated by a tool that appears to be analytically rigorous, the manager who doesn’t question it makes decisions based on a reality that doesn’t exist.
The difference between asking “What should I do about these low sales?” and asking “This is what I think is happening in my store today—what is the data telling me that I might be overlooking?” is the difference between delegating judgment and using it as a starting point for better thinking.
How to Formulate the Right Question
A good question to the system—“store with a drop in average ticket size over three consecutive days, without a reduction in foot traffic and with a stable category mix: what could be causing this, and what is the first action to review?”—is valid. But there is a preliminary step that most people overlook: how did the manager know that those three data points were the relevant ones and not others?
The answer isn’t in the system. It lies in the morning store walkthrough, in the conversation with the previous shift’s team, in observing which sections are out of stock and which are overstocked, and in the memory of what happened during the same period last month. That information doesn’t exist on any dashboard; it exists in the mind of the person who knows their store.
The actual process has three steps that must be followed in order. First, observe before looking at the report: walk the floor, talk to the team, identify what looks different today compared to what was expected. This observation builds the initial hypothesis. Second, approach the data with a pre-formulated question; don’t open the system to see “what’s there,” but rather to confirm, refute, or expand upon what you already suspect. Third, verify that the action suggested by the analysis is executable with the actual resources available during that shift, not with those the system assumes exist.
From oracle to assistant: the usage model that makes the difference
The opportunity isn’t in using less artificial intelligence; it’s in using it better. The correct approach to a store’s operational management is to treat it as an analysis accelerator, not as a substitute for judgment. The effective model has three stages that must be followed in sequence.
The first is the manager’s own judgment: before consulting the tool, the manager must have a hypothesis about what is happening, based on their walkthrough of the store, conversations with their team, and their reading of the day’s context. This stage cannot be skipped; it is what ensures that the subsequent interpretation has value. The second is expansion with AI: using the tool to confirm, expand, or challenge that hypothesis with data, identify omitted variables, and compare with similar periods. Here, the tool multiplies the analytical capacity of someone who already has a formed position. The third is operational validation: comparing the results with the visible reality on the shop floor and confirming that the suggested action is executable with the resources available at that moment.
High-value questions are not the most technologically sophisticated ones. They are the ones that provide the tool with the necessary context for its response to be useful: the actual situation in the store that day, the specific decision to be made, and the real constraints regarding staffing, inventory, and the behavior of regular customers. A well-constructed question does not ask the AI to guess; it asks it to work with the operational reality that only the manager knows.
The competitive advantage of the store manager who keeps thinking
In the coming years, access to artificial intelligence tools for operational management will be as universal as access to a point-of-sale system is today. The tools will converge. The competitive edge won’t lie in having them; it will lie in the operational clarity of those who use them: their ability to ask the right question, their discipline in validating what the system reports, and their judgment in distinguishing when the numbers reflect the reality of the store and when they simplify or distort it.
The stores that lose ground will not be those that fail to adopt AI. They will be the ones that stopped questioning the information they received because the system presented it with sufficient confidence. And the most valuable managers will not be those who process the most data the fastest, but those who know how to identify the real problem behind the metric that the report is not capturing.
AI does not eliminate operational judgment; it makes it more necessary. Because the easier it is to obtain a number, the more important it is to know whether that number accurately describes what is happening in the store today, during this shift, with this team, and in this context. The risk of this era is not artificial intelligence; it is the operational complacency it can bring.
Those who develop the habit of observing first, always questioning, and validating before acting will not only manage faster—they will manage better. Because in the age of AI, true operational leadership does not consist of having the most comprehensive report; it consists of knowing whether that report describes what is actually happening on the shop floor and having the judgment to act on the discrepancy.
