Fujitsu puts four AI agents on the shop floor: retail trial starts today, seven chains in, full launch June 2027
Fujitsu announced on Tuesday that it will begin offering a trial environment for four AI agents built for retail — sales-structure analysis, customer loyalty, sales planning, and store-manager support — along with the execution platform that runs them, under its Uvance for Retail initiative.
Fujitsu said in a press release dated October 6 that the trial environment opens today, October 6, 2026. The four agents are designed to carry retail decision-making from data analysis all the way to concrete measures — higher sales, better profitability, faster calls by management — rather than stopping at dashboards. Fujitsu will also run demonstration experiments with seven retail companies in phases, and will exhibit the agents at the 22nd Asia-Pacific Retailers Convention & Exhibition (APRCE) at Tokyo International Forum, running October 7–10. The official launch is scheduled for June 2027.
Four agents, four retail jobs
The sales-structure analysis agent breaks revenue down by its constituent parts, visualizes the mix, flags imbalances and growth opportunities, then recommends concrete measures to maximize profit and forecasts their impact. Product and area managers can compare stores and areas side by side and rework product assortments against what the data actually shows. The customer-loyalty analysis agent mines underutilized customer data for behavioral signals — purchase motivations and attrition factors that don't show up in simple frequency or loyalty-point metrics — then tailors measures and service improvements aimed at cultivating high-value customers. The sales-planning (MD planning) agent takes on the merchandiser's most time-consuming job: building per-product, per-store sales plans from market trends, customer needs, and the competitive picture, recommending inventory allocations for each store, monitoring sales after launch, and flagging inventory transfers to prevent stockouts. The store-manager support agent folds external signals — local market areas, store characteristics, weather, social media — into store sales data to recommend optimal assortments and shelf-display changes, moving managers off manual analysis and experience-only decisions.
The platform underneath
The trial isn't just agents — it includes the AI execution platform, a common layer that connects diverse retail data sources and business functions. It provides shared data storage, processing, and access control, and is where agents perform demand forecasting, generate tailored recommendations, and automate tasks. Fujitsu says the platform is being built on its Kozuchi AI platform and a framework for developing and operating multi-AI agents, and that the longer-term aim extends beyond storefronts to the whole distribution chain: manufacturers, wholesalers, and logistics providers. That platform framing matters — it turns the agents from a consulting engagement into a packaged product the company can keep expanding.
Why retail, why now
The announcement names the gap Fujitsu is aiming at: Japanese retailers sit on sales and customer data they can't fully use for strategy. Advanced analysis and KPI development demand specialists; routine work — compiling data, preparing reports and materials — eats the day; and the cycle from identifying an issue to deciding a measure to evaluating the result keeps getting longer. The pitch is that agents collapse that cycle rather than just visualize it.
Trial now, launch later
Keep the framing honest: this is a trial environment starting today, with phased demonstrations at seven unnamed retailers, and the official launch is set for June 2027 with more agents added progressively. The release gives no performance numbers — no forecast accuracy, no measured sales lift, no pilot results — so the credible thing to watch is what comes out of the seven demonstration experiments and the APRCE showing. If the trial converts, it would be one of the first attempts by a major systems vendor to sell AI agents to retail as a platform-supported product rather than custom consulting work.