AI in Retail: Adoption Is Growing, but Value Is Still Missing from Critical Areas

The reality of artificial intelligence is already part of a large portion of Brazilian retail.

According to research by Zucchetti Brasil in partnership with Central do Varejo, 59% of companies in the sector are already using AI solutions. The survey, conducted with 373 professionals, shows that the technology has ceased to be a distant possibility and has begun to integrate different business activities.

But the numbers reveal an even more important question than the speed of adoption: where is this intelligence being applied?

An article published by Tiago Girelli, General Director of Wake Commerce, in E-Commerce Brasil, draws attention precisely to this difference.

While AI is rapidly advancing in communication tasks, content production, and customer service, its presence is still relatively small in processes directly related to merchandise management, replenishment, and operational efficiency.

For retail, this difference can represent a strategic opportunity.

AI has arrived in retail, but it has not yet reached the entire operation

The data shows a significant concentration of applications in activities more accessible to automation.

Among retailers using AI, the main applications cited are:

The difference is striking.

AI is already involved in creating campaigns, producing product descriptions, engaging with consumers, and automating administrative tasks.

These applications can generate significant gains in productivity and commercial efficiency.

However, activities that involve complex operational decisions still show considerably lower levels of adoption.

Inventory management, replenishment, demand forecasting, purchasing planning, and distribution of goods are processes that directly influence working capital, product availability, and margin.

This does not mean that AI applied to marketing is less relevant.

It means that a significant part of the economic potential of the technology may still be outside the areas where it is most used.

The greatest value may be where the consumer does not see

Imagine a retail chain with hundreds of stores and thousands of products distributed across different distribution centers.

Daily, this operation needs to decide how much to buy, when to replenish, where to position goods, and how to respond to changes in demand.

An inappropriate decision can lead to excess inventory in one region and stockouts in another.

The result involves tied-up capital, additional movement costs, the need for discounts, and lost sales opportunities.

In this scenario, artificial intelligence can support decisions such as:

• Anticipating demand variations by product, store, and region.
• Identifying risks of stockouts and excess inventory.
• Recommending replenishment and distribution adjustments.
• Supporting pricing and promotion analysis.
• Detecting operational exceptions that require intervention.

These applications may not be as visible as a virtual assistant or an AI-generated campaign.

But they can act on fundamental variables of a retailer's economic performance.

The value, in this case, is not in the number of automated tasks, but in the improvement of the decisions that drive the operation.

The challenge is not just technological. It is operational.

Why is the adoption of AI still relatively small in processes such as inventory, demand forecasting, and logistics?

One explanation lies in the complexity of these activities.

Producing a text with AI may require little information and direct interaction with a model.

On the other hand, correctly recommending the replenishment of a product may depend on sales history, available inventory, orders in transit, seasonality, commercial rules, supplier constraints, and replenishment parameters.

This information is often distributed across different systems.

In a large operation, artificial intelligence needs to work on consistent data and well-defined processes for its recommendations to be reliable.

Otherwise, there is a risk of accelerating decisions based on incomplete or incorrect information.

Automating a poorly structured process does not necessarily mean making it more efficient.

That is why the evolution of AI in retail also requires attention to data quality, integration architecture, and business rules.

The next frontier: intelligence connected to transactional systems

One of the most relevant opportunities lies in the proximity between artificial intelligence and the systems responsible for daily operations.

Merchandising, planning, inventory management, pricing, stores, and logistics solutions concentrate essential information about how retail works.

When this information is consistently available, more sophisticated applications of analysis and automation become possible.

Consider the process of replenishing goods.

An intelligent solution can identify a trend of increasing demand and assess its relationship with inventory availability.

Based on this analysis, it can recommend a replenishment or transfer action, respecting operational and commercial parameters.

Depending on the architecture, integrations, and controls available, certain steps can also be automated.

It is at this point that artificial intelligence begins to evolve from an individual support tool to a component integrated into business journeys.

The transformation does not simply consist of adding a chatbot to a corporate system.

It consists of connecting data, decisions, and operational execution.

Agentic AI: great possibilities, but also great risks

The evolution of artificial intelligence agents further expands these possibilities.

Unlike applications that only answer questions or generate content, agents can be designed to execute sequences of activities, use tools, and interact with different systems.

However, autonomy needs to be accompanied by controls.

According to a projection released by Gartner, more than 40% of agentic AI projects may be canceled by the end of 2027, mainly due to rising costs, difficulty in demonstrating business value, and insufficient risk control mechanisms.

The indicator specifically refers to agentic AI projects, not the entirety of artificial intelligence initiatives.

Still, the projection represents an important warning.

The availability of advanced technologies does not eliminate the need for a consistent economic justification.

Projects need to have clear objectives, controlled costs, performance indicators, and adequate governance mechanisms.

For retail, this means that the implementation of intelligent agents must consider not only what they can execute but also the operational and financial consequences of their decisions.

From experiment to measurable result

One of the main changes in approach consists of starting AI projects from concrete business problems.

Instead of asking which tools can be used, companies need to identify which processes offer relevant improvement opportunities.

This shift helps establish an objective relationship between investment and result.

In a demand forecasting project, for example, indicators may include forecast accuracy, reduction of stockouts, and decrease in excess inventory.

In a merchandising process automation initiative, execution time, volume of manual interventions, and team productivity may be considered.

In pricing applications, results need to consider not only sales growth but also the impact on margins and profitability.

The journey can follow a simple logic:

DiagramaClique para ampliar

This approach allows starting with controlled initiatives, testing hypotheses, and expanding investments only when benefits are demonstrated.

It also helps differentiate projects that truly transform operations from those that merely present an interesting technology.

The competitive advantage will not just be in having AI

As models, platforms, and artificial intelligence tools become more accessible, the technology alone tends to offer less differentiation.

Competing companies may use similar tools.

What is not equally simple to reproduce are the processes, data, business knowledge, and execution capacity of each organization.

An AI solution can be quickly contracted.

But building an operation with reliable information, consistent rules, integrated systems, and prepared teams requires a deeper transformation.

In retail, this difference is especially relevant because a large part of the economic result depends on the quality of thousands of operational decisions made daily.

The true differential will be in combining artificial intelligence with specialized knowledge about the business.

The future of AI in retail will be measured by impact

The growing adoption of artificial intelligence represents an important evolution in the sector.

But the next stage will require a shift in focus.

It will not be enough to measure how many tools have been implemented, how many employees use AI, or how many activities have become automated.

It will be necessary to demonstrate how these initiatives contribute to reducing costs, improving availability, increasing productivity, protecting margins, and generating sustainable results.

For large retailers, this means taking artificial intelligence beyond communication and relationship areas, progressively connecting it to the operational core.

The technology already shows the capacity to transform the way people work.

The next opportunity lies in transforming the way companies operate.

In retail, the competitive advantage will not be defined by who uses more artificial intelligence, but by who can apply it to the decisions that truly drive results.