AI agents move from experiment to operations in UAE retail

Breaking down data silos

While retailers have invested heavily in digital technologies over the past decade, many continue to struggle with fragmented technology environments. Merchandising applications, enterprise resource planning (ERP) platforms, warehouse management systems, e-commerce platforms and spreadsheets frequently operate independently, creating information silos that slow operational decision-making.

For organisations operating in highly competitive retail markets, delays of even a few hours can have commercial consequences. Consumer demand changes rapidly, supply chain disruptions continue to affect product availability, and competitors can alter pricing multiple times a day.

Yusuf argues that the effectiveness of agentic AI depends less on the sophistication of the model itself than the quality and accessibility of enterprise data. “One of the biggest misconceptions is that organisations need more data before they can use AI,” he says.

“In reality, most retailers already have enormous amounts of information. The challenge is connecting those systems so AI agents can understand what’s happening across the business rather than within isolated applications.”

Rather than requiring employees to manually retrieve information from different departments, AI agents can access multiple business systems simultaneously, analyse sales performance, inventory levels, customer demand, supplier information and market conditions before recommending an appropriate course of action. In some scenarios, those recommendations can then be executed automatically, provided the organisation has established appropriate governance controls.

The ability to connect previously disconnected systems is becoming increasingly important as retailers continue expanding across multiple sales channels. Customers now expect consistent pricing, inventory visibility and fulfilment regardless of whether they shop online, through marketplaces or in physical stores.

Pricing becomes increasingly dynamic

Pricing represents one of the clearest examples of where agentic AI could transform retail operations. Large retailers often manage tens of thousands of stock-keeping units (SKUs), each with different demand patterns, seasonal behaviour and competitive pressures. Monitoring those variables manually is becoming increasingly difficult, particularly as retailers attempt to balance profitability with customer expectations.

“No pricing team can realistically analyse every SKU every day,” says Yusuf. “An AI agent can continuously evaluate price elasticity, competitor activity, historical demand and current inventory before recommending pricing adjustments that protect margins while remaining competitive.”

Unlike traditional pricing software, which generally follows predefined rules, agentic AI introduces a greater degree of reasoning into the process. Rather than simply reacting to one trigger, AI agents can evaluate several business objectives simultaneously, balancing profitability, inventory turnover, promotional effectiveness and customer demand before determining the most appropriate response.

Depending on the level of autonomy granted by the organisation, pricing recommendations can either be reviewed by merchandising teams or automatically implemented within predefined thresholds, with exceptions escalated for human approval.

The same principles extend to promotional campaigns and markdown strategies, where timing is often as important as the discount itself.

“Retailers frequently lose margin because promotions aren’t aligned with inventory or changing demand,” says Yusuf. “An AI agent can continuously monitor sales performance, stock availability and market conditions, allowing promotions to evolve dynamically instead of waiting for the next campaign cycle.”

As retailers increasingly operate across digital and physical channels simultaneously, this ability to make continuous commercial adjustments could become an important competitive differentiator.

Inventory management represents another area where organisations are beginning to explore agentic AI. Traditional forecasting models have already helped retailers improve replenishment planning, but agentic systems introduce an additional layer of automation by combining predictive analytics with autonomous decision-making.

Rather than simply identifying potential shortages, AI agents can recommend inventory redistribution, initiate replenishment workflows or alert procurement teams before stock issues affect customers. “The objective isn’t simply predicting demand more accurately,” says Yusuf. “It’s reducing the time between identifying a potential problem and actually resolving it.”

For retailers operating across multiple locations, this could mean automatically moving inventory from stores with slower sales to those experiencing stronger demand, reducing stock-outs while avoiding unnecessary purchasing.

The same operational principles are increasingly being applied in manufacturing, where supply chain disruptions and inventory imbalances continue to create financial pressure.

Manufacturers face many of the same challenges as retailers, balancing production schedules, component availability, procurement costs and fluctuating customer demand. AI agents are capable of analysing these variables together rather than independently, allowing procurement decisions to become both faster and more responsive to changing market conditions.

Commercial pricing in manufacturing may also benefit from greater automation. “Sales teams often prepare quotations using information that’s already out of date by the time the proposal reaches the customer,” says Yusuf. “AI agents can continuously evaluate production costs, raw material prices, supplier availability and previous commercial agreements to generate recommendations based on current business conditions.”

Reducing quotation times from days to minutes could allow manufacturers to respond more quickly to customer requests while improving pricing consistency and profitability.

From generative AI to autonomous business operations

Much of the recent conversation around AI has focused on generative applications capable of producing text, code or images. Yusuf believes agentic AI represents a significant shift because its value comes from completing business processes rather than simply generating information.

“Generative AI helps people create content,” he says. “Agentic AI helps organisations get work done. That’s a fundamentally different capability because you’re moving from assistance towards autonomous execution under human oversight.”

This transition has important implications for enterprise technology strategies, rather than deploying isolated AI applications, organisations increasingly need platforms capable of integrating AI agents securely across multiple business systems while maintaining visibility into every decision those agents make.

The growing challenge of shadow AI

However, as organisations accelerate AI adoption, governance is emerging as one of the most significant challenges.

The rapid adoption of cloud computing introduced concerns around shadow IT, where employees adopted unauthorised applications outside central IT oversight. Yusuf believes a similar trend is now developing around AI, although with potentially greater consequences.

“Shadow AI is arguably more complex than shadow IT because AI isn’t just accessing information, it may also be making decisions or initiating actions,” he says. “If organisations don’t understand which AI agents are operating across the business, what systems they can access and what authority they’ve been given, they introduce significant operational and security risks.”

Unlike conventional software, AI agents may be granted access to customer information, ERP systems, financial applications and operational workflows. Without appropriate governance, organisations may struggle to understand how decisions are being made or whether company policies are consistently being followed.

As governments around the world continue introducing AI regulations alongside existing cyber security and data protection requirements, visibility into AI operations is becoming increasingly important.

Yusuf says governance cannot be treated as an afterthought once AI has already been deployed. “Every AI agent should operate within clearly defined guardrails,” he says. “That includes least-privilege access, policy enforcement before actions are executed, approval thresholds where appropriate and complete auditability of every significant decision.”

Decision provenance is becoming particularly important as organisations seek to demonstrate regulatory compliance and internal accountability. Being able to reconstruct how an AI system reached a recommendation may become just as valuable as the decision itself.

Building the foundations for enterprise AI

For retailers across the UAE, the growing interest in agentic AI reflects broader changes taking place across enterprise technology. As businesses continue expanding digitally while facing mounting operational complexity, automation is evolving from isolated workflow improvements towards autonomous decision-making embedded in core business processes.

Yusuf believes success will ultimately depend less on the AI models organisations choose than on the operational foundations supporting them.

“The organisations that will realise the greatest value from agentic AI won’t necessarily be those deploying the most AI,” he says. “They’ll be the ones that invest in trusted data, integrated systems and governance from the very beginning. Once those foundations are in place, AI agents can scale safely across the organisation and become a genuine competitive advantage.”