DFI taps AI to sharpen merchandising

Group is directing the bulk of its AI efforts towards business transformation.

Retailers that spread their artificial intelligence (AI) investments across too many use cases at once risk missing measurable returns, as the competitive retail landscape dictates a focused strategy.

“Retailers need to focus on initiatives that deliver tangible value rather than adopt AI simply because it is gaining traction,” Stijn Casneuf, data & AI automation director at DFI Retail Group, said during the Retail Asia Summit 2026 in Hong Kong.

Casneuf said that the company focuses its AI efforts across three key areas: AI  for merchandising, AI for customers, and AI for its store team members.

He added that DFI aims to allocate more than half [SL2.1]of its efforts to business transformation, reinvention and innovation, and the remainder to productivity improvement for internal functions.

The AI strategy covers four areas, including merchandising, customers, and storefront store support centres. The group operates over 7,600 stores across 12 markets in Asia.

Casneuf said merchandising receives the largest share of attention because improvements in this area can significantly affect both revenue and profitability.

“As an example, the company is using AI in supply chain forecasting, combining internal business data with holiday or event calendars and weather information to improve forecast accuracy,” he added. 
DFI began deploying AI forecasting around one-and-a-half years ago, with Casneuf noting that the technology has already generated significant value through improved forecasting accuracy.
Beyond selecting the right applications, Casneuf said retailers need to address fragmented data systems before they can scale AI effectively.

DFI began restructuring data processes a year ago and is now starting to see the first wave of benefits. As the company continues to advance, it expects to unlock even greater value and impact.

To coordinate its efforts, DFI established an AI Control Tower comprising representatives from functions including people and culture, technology and cyber security, finance, corporate communications and legal, and AI.

The team assesses proposed initiatives against business priorities, shareholder value, customer trust, and team member impact before evaluating technical fit, security, and legal requirements.

The Control Tower reports directly to the Management Committee, with updates also provided to the board. Casneuf said this structure helps the company concentrate resources on a prioritised set of high-value projects.

“We are also taking a hybrid approach to large language models (LLMs), assessing Chinese and Western models according to the requirements of each application,” he added.

On the customer side, DFI is exploring how AI can improve product discovery and shopping experiences.  

The group is also piloting the use of LLMs to improve product descriptions, drawing on enterprise resource planning or ERP data and predefined best practices from the market. This allows for better LLM discovery and searchability.

The group manages millions of stock-keeping units, whilst commercial and master data teams validate AI-generated descriptions, rather than creating them by hand.

“Agentic shopping remains at an early stage, but DFI has a clear strategy for how customers could eventually use AI agents to shop through its own platforms,” Casneuf said.