Fujitsu is opening trials of four retail AI agents with seven retail companies, expanding its work on tools that help store teams analyse sales, plan stock and respond to changing customer behaviour.
The Japanese technology group announced the trial environment on 6 October. A full commercial launch of the agents and their shared platform is planned for June 2027.
The four tools cover sales analysis, customer loyalty, merchandise planning and store-manager support. Demonstration projects with the seven retailers will begin in phases.
The announcement builds on Fujitsu’s earlier work with AEON Food Style. That project focused on store strategy and shelf planning. The new programme broadens the scope to include customer behaviour, inventory allocation and commercial decisions across stores.
Fujitsu is already part of Japan’s established supermarket technology sector, where suppliers are developing systems that connect store operations with central merchandising and planning.
Its sales analysis agent examines the components of sales performance, identifies potential opportunities and recommends actions. Product teams and area managers can use it to compare stores and review assortments.
The customer loyalty agent examines purchasing behaviour beyond measures such as visit frequency and loyalty-point use. It is designed to identify possible reasons for customer loss and suggest responses.
For merchandising teams, the sales planning agent supports product-level and store-level plans. It also recommends stock allocations and monitors performance after sales begin, including where transfers between stores could help prevent shortages.
The store-manager agent combines sales figures with information about local markets, store characteristics, weather and social media. It then recommends assortment and shelf-display changes.
These functions extend the direction established by the AEON Food Style store-management trial. However, a broader trial is not the same as a proven return on investment.
For grocery retailers, the practical test will be whether recommendations improve availability and sales without increasing waste, markdowns or staff workload.
A proposed stock transfer, for example, may appear attractive in a sales forecast but become less useful once transport costs, remaining shelf life and handling requirements are included. Fresh and chilled categories make those trade-offs particularly important.
Assortment recommendations also need to reflect supplier lead times, agreed promotional volumes and the space available in each store. A system can identify demand without necessarily creating the supply capacity needed to meet it.
Fujitsu’s shared execution platform connects data sources and provides storage, processing and access controls. That infrastructure is intended to support the agents as the company adds further functions.
The commercial challenge will be maintaining reliable product, customer and store data across those connections. Incomplete records or inconsistent stock figures can weaken recommendations before a store team has acted on them.
Fujitsu has not reported measured sales gains, waste reductions or labour savings from the new seven-company programme in the announcement. Those results will be important when retailers assess whether to move beyond trials.
The company plans to extend its AI-agent range over time, with ambitions reaching beyond retail into manufacturing, wholesale and logistics.
Ahead of the planned June 2027 launch, the strongest evidence will come from repeatable improvements across different stores and categories—not simply the number of tasks the agents can perform.
Editor’s note: Based on Fujitsu’s announcement dated 6 October 2026. Grocery Trade News added analysis of stock planning, store operations and the commercial tests facing retail AI adoption.
