Not every inventory problem is a people problem. In a case reported by Grupo Intelsis, a specialty retailer with a large SKU count and an inventory-driven business had experienced buyers and a well-established purchasing routine. Even so, it kept running out of its best sellers, ending up with leftover seasonal products and placing frequent rush orders.
Over 18 months, the company cut total inventory by 35% with no loss of service level, recovered 3 points of sales margin and reduced off-cycle orders by more than 60%. It didn't replace its ERP or migrate data. Here is how it got there: the diagnosis, the decision, the rollout and the lessons for anyone facing similar symptoms.
Strong buyers, late information
The purchasing routine drew on four inputs: sales history, quarterly projections from the sales team, supplier agreements and the team's own experience. The results were reasonable, but the process had a built-in lag. By the time a forecast was ready, the data was days or weeks old, and every correction only took effect in the next buying cycle.
That detail changes the diagnosis. The team wasn't making poor decisions; it was making sound decisions with the information it had, and that information arrived late. In situations like this, hiring more buyers or demanding more accuracy rarely fixes anything, because the bottleneck is the flow of data, not analytical skill.
Three symptoms of the same lag
The lag showed up in three ways, always at the most sensitive moments for the business:
- Stockouts of top sellers during demand peaks, exactly when each lost sale matters most.
- Chronic overstock of seasonal products that didn't sell during the season and ended up tying up capital.
- Urgent off-cycle orders that drove up logistics costs and strained supplier relationships.
At first glance, the three problems look unrelated. In practice, they share a root. Product runs out when demand grows faster than the forecast can track, and it piles up when demand drops before the plan is revised. The rush order is an attempt to fix the first mistake at a higher cost. If you want to measure this effect in your own company, start with these four diagnostic questions on overstock and stockouts.
What the company wanted, and what it avoided
The company was clear about what it didn't want: a new ERP or a consulting project lasting many months. It was looking for an artificial intelligence tool that would connect to its existing systems, learn from historical data and recommend what to order, how much and when.
The choice rested on a clear principle: people would keep the final say. The tool would take over the repetitive work of building and recalculating forecasts, SKU by SKU. Buyers would review the recommendations, add what they know about suppliers, the market and negotiations, and make the call.
That design reduces two common risks in AI projects. The first is rejection: when a system decides on its own, teams tend to distrust it and work around it. The second is lost knowledge, because part of what a good buyer knows never makes it into the sales history.
Recommendations in less than a week
The rollout moved faster than expected. Once connected to the ERP the company already used, the solution processed its sales and inventory history and began producing recommendations in less than a week, with no data migration, no lengthy technical training and no disruption to operations.
A start at that pace depends on a few conditions worth checking before any similar project:
- Accessible history: SKU-level sales and inventory movements that can be pulled from the ERP without manual rework.
- Consistent master data: item codes, units of measure and category hierarchies that allow period-to-period comparisons.
- Defined scope: which decisions the tool will support and who approves each recommendation.
- Starting metrics: inventory, service level, margin and rush orders measured before launch, so you can compare later.
What changed in 18 months
The reported results cover the first 18 months of use:
- Total inventory down 35%, with no impact on service level. The freed-up capital went into faster-moving, higher-margin categories.
- 3 more points of sales margin, driven by fewer clearance sales and better coverage of high-demand items.
- Off-cycle orders down by more than 60%, which cut logistics costs and made supplier relationships more stable.
- Buyers focused on strategy: instead of building and correcting forecasts, the team reviews ready-made recommendations and spends its time on suppliers, negotiations and business decisions.
The gains reinforce each other. Less idle stock frees up capital; capital moved into faster-moving categories improves margin; fewer emergencies lower costs and ease pressure on suppliers. None of it came from a system replacement. It came from changing how information reaches the people who decide.
Lessons for a similar project
The main lesson is organizational. Better tools and timely information lead to better decisions, with the same team. Artificial intelligence took over the repetitive work and left the judgment calls to people. In one sentence: AI recommends, the team decides.
There is also a lesson about sequencing. Instead of a broad transformation program, the starting point was one specific decision, inventory replenishment, with clear metrics. A focused scope like that makes it easier to measure the return, earn the team's trust and choose the next steps based on data.
To go deeper, this article explains how an AI solution supports demand, inventory and pricing decisions and which kinds of companies it fits best.
Next step: assess the potential in your operation
If these signs sound familiar, the Grupo Intelsis team can help you estimate the improvement potential with your own data, understand what your current ERP already offers and decide whether it makes sense to start with a few categories. Talk to our team and bring last quarter's numbers to the conversation.