Automated replenishment: case study Danone x Flowlity

AI-powered automated replenishment is typically delivered through advanced Supply Chain planning platforms rather than basic ERP systems. These platforms combine demand forecasting, inventory optimization, and supply planning capabilities to automatically calculate replenishment quantities and timing. They often integrate with ERP solutions such as SAP to leverage existing data while enhancing replenishment decisions with AI-driven models. The architectural split is deliberate: ERPs are optimized for transactional execution, while planning platforms are optimized for decision quality under uncertainty. Running replenishment in a dedicated planning layer means buffers and order proposals reflect real demand variability and lead time risk, rather than static parameters that age quickly in volatile markets.

E-commerce brands automate replenishment reminders by using inventory management or replenishment software that monitors stock levels and demand patterns in real time. When inventory reaches predefined thresholds, the system automatically triggers reorder recommendations, purchase orders, or alerts. More advanced solutions rely on AI to anticipate demand fluctuations and adjust replenishment timing before stockouts occur. The shift from threshold-based to anticipation-based reminders is what separates basic automation from real decision support. With probabilistic forecasting in the loop, the system not only tells planners when to reorder but also how much to commit given the current demand uncertainty, lead time and service level target on the SKU.

Retailers automate replenishment tasks by connecting sales data, inventory levels, and supplier lead times within automated replenishment systems. These systems continuously calculate reorder points and quantities, reducing manual intervention. AI-powered solutions allow retailers to dynamically adjust replenishment decisions across stores, warehouses, and distribution centers based on demand variability and service level objectives. The benefit grows with network complexity: managing thousands of SKU-location pairs manually is simply not realistic, while automated logic can keep every node at its target service level with the minimum inventory required. Planners then shift their time from running calculations to handling exceptions, which is where their judgment actually creates value.

Traditional reorder point systems rely on static thresholds that must be manually updated and often fail in volatile environments. Automated replenishment continuously recalculates replenishment decisions using real-time data, demand forecasts, and safety stock logic, making it far more resilient to demand spikes, supplier delays, and seasonality. The difference shows up most clearly in service level and inventory KPIs read together. Static thresholds tend to either over-cover, generating excess stock, or under-cover, generating shortages, because they cannot adapt fast enough to changing demand profiles. Dynamic replenishment sizes buffers per SKU period based on real demand uncertainty, which is what protects service while keeping working capital under control.

Yes. Automated replenishment is a critical execution layer that connects inventory optimization and supply planning. It translates optimized inventory policies and demand forecasts into concrete replenishment decisions, ensuring alignment between strategic planning objectives and day-to-day operations. Without this layer, even the best planning logic stays theoretical, because the gap between policy and execution is where most service level and working capital losses occur. With it, the same model that sets the target service level and dynamic safety stock also generates the specific order proposals that achieve them, which is the only way to keep inventory optimization and supply planning consistent over time across SKUs, suppliers and locations.

Danone, founded in 1919 with €25.29 billion in revenue (2019) and 104,843 employees, is a leader in the agrifood industry. The project with Flowlity began in January 2020, originating from a call for project proposals issued by Danone in collaboration with Microsoft's “AI Factory For Agrifood” program. The program's aim was to respond to challenges in agriculture, logistics, and supply chain management, including waste reduction. Within that program, Flowlity was selected to work on raw material and packaging stock optimization, two key elements in the food sector supply chain. The pilot scope covered 27 products at a Danone Nutricia facility in Haps, the Netherlands. The solution optimizes stock levels (min and max) and provides replenishment and consumption forecasts to support the planning team.

Once integrated with Danone's SAP environment, Flowlity retrieved all past orders and inventory history over a two-year period. Using this data, the teams compared past stock forecasts with what Flowlity's algorithms would have proposed. On a three-month horizon, Flowlity's forecasts reached approximately 79% reliability versus around 30% for Danone's pre-existing forecasts. On a six-month horizon, the comparison was 67% versus 12%. The gap between the two approaches widens as the horizon lengthens, which matters specifically for raw materials and packaging: these categories require visibility several months ahead to coordinate with multi-tier supplier networks. The forecast improvement is the underlying driver of the inventory reduction projected over the next twelve months and of the better service level reported after the pilot.

The Danone project followed Flowlity's standard four-phase implementation. On Day 1, the team held a kick-off and organised workshops to define project scope, use case, and data integration. After two months, continuous data integration from SAP was complete, and the algorithms had been trained using the integrated data sets. After three months, the planning team had access to the application and AI-generated recommendations, and users were sharing their planning and feedback. After six months, the application was implemented and used daily by the procurement team. This phased timeline (data, then test, then production) is what moved the project from kickoff to operational use within six months at the Nutricia plant in Haps.

Six months after integration with Danone's existing IT environment, the solution was fully operational and deployed for the planning team. Flowlity supported the planners and let them dynamically adjust the company's safety stock and replenishments. It enabled replenishments to be digitalised and automated, and improved service level by reducing the risk of shortages. The teams documented a 17.28% reduction in stocks based on a six-month simulation, and projected a 28% to 40% reduction in inventory over one year as the platform continued to mature. Additional results expected included synchronization with suppliers, which became the focus of the next deployment phase.

To make Danone's digitalisation and inventory reduction goals a reality, Flowlity moved into a synchronization phase with the group's suppliers, including DS Smith, Dutch State Mines, and Ardagh Group. The objective was to integrate supplier data without ever compromising the confidentiality of information belonging to each party. This visibility gave Danone and its suppliers improved recommendations: suppliers can track past and incoming orders in real time, obtain sales forecasts for all Danone products covered by the project, and anticipate potential shortages. As a result of this synchronization, Danone measured a further inventory reduction of around 12.40%, and Danone's suppliers measured a 30% to 60% reduction in their finished goods inventory. The supplier synchronization phase therefore created mutual value: Danone reduced its own stocks further, while suppliers cut their finished-goods inventories drastically.

The Danone project began in January 2020 as a result of a call for project proposals that Danone issued in collaboration with Microsoft's “AI Factory For Agrifood” program. The program's purpose was to respond to challenges in agriculture, logistics, and supply chain management, with waste reduction as one of its priority topics. Flowlity was selected through this call to work on raw material and packaging stock optimization, two key elements in the food sector supply chain. This program-driven origin is what shaped the initial pilot scope (27 products at the Nutricia plant in Haps) and the focus on agrifood-specific challenges, before the project expanded into the supplier synchronization phase with DS Smith, Dutch State Mines, and Ardagh Group.

The agrifood industry combines high-volume operations with strict service requirements and raw materials whose availability depends on agricultural cycles. Danone's project illustrates why automated replenishment is a fitting answer: raw materials and packaging are the two key elements in the food sector supply chain, and they are the categories where forecast errors translate most directly into either obsolescence or supplier emergencies. Flowlity's deployment at Danone allowed replenishments to be digitalised and automated, with the planning team dynamically adjusting safety stock and replenishments rather than running fixed coverage rules. This shift improved service levels by reducing the risk of shortages and projected inventory reductions of 28% to 40% over one year. For food industry players considering similar projects, the Danone case shows that AI-driven replenishment can produce these results within a six-month pilot, before expanding into supplier synchronization for additional gains.