AI Can Predict Ecommerce Demand. But Can It Tell You What to Do Next?

By Johannes Panzer - Descartes.

For years, ecommerce inventory planning has revolved around a deceptively simple question: How much are you going to sell? Artificial intelligence is getting much better at helping you answer it. Feed enough sales data into the right tools, and AI can identify patterns, analyze trends, and produce increasingly sophisticated demand forecasts.

But if you run a growing ecommerce business, you already know that predicting demand is only part of the problem. The harder question is what you should do about it. Knowing you are likely to sell 500 units does not tell you how many units to order, when to place the purchase order, which supplier constraints to consider, or where those units should go.

This is where the conversation around AI in ecommerce needs to evolve. The real opportunity is not simply better forecasting. Today, you can leverage your data and technology to turn forecasts into smarter inventory decisions.

Growth Makes Inventory Planning Complicated Fast

Inventory planning can be relatively straightforward when your business is small. You review what sold, check what remains in stock, and decide what to reorder. Then you grow, and the calculations that once worked start becoming considerably more complicated.

You add SKUs and begin selling through additional marketplaces. You open another warehouse or perhaps a retail location. One product sells quickly in one region but slowly in another. Supplier lead times vary, promotions distort normal demand, and stockouts make historical sales look artificially low.

Soon, even a seemingly simple replenishment decision requires you to consider several variables simultaneously. Suppose your forecast says you will need 37 units, but your supplier only ships that product in cases of 12. You already have inventory on an open purchase order, while one location is approaching a stockout and another has enough inventory for six weeks.

You can manage these calculations in spreadsheets for a while, and AI can make those spreadsheets considerably more powerful. But as your operation becomes more complex, the limitations of this approach become increasingly obvious. You do not just need a better prediction. You need a repeatable way to translate that prediction into action.

Demand Forecasting and Inventory Planning Are Not The Same Thing

A demand forecasting solution estimates what you are likely to sell, whereas inventory planning tools determine what you should purchase and where that inventory should go. Making that second decision requires much more operational context.

You need to consider how much inventory you have now, how much is already on order, and when it will arrive. You also need to account for replenishment lead times, safety stock, minimum order quantities, and case or pack requirements. Each variable changes what a seemingly straightforward sales forecast means for your next purchase order.

Then there is location. If you sell from multiple warehouses or stores, knowing you need another 1,000 units across the business is not enough. One location might sell a particular SKU twice as quickly as another, while your warehouse might need to retain a minimum quantity for ecommerce fulfilment.

The questions quickly change from “How much will we sell?” to “Which locations are likely to run out first?” and “Where should limited inventory go?” Those are much more valuable problems for technology to help you solve.

AI Needs to Understand How Your Business Actually Operates

A mathematically sound forecast can still produce a bad inventory decision when the technology does not understand your operational constraints. Imagine one retail location suddenly sells ten units of a product. A system sees the increased sales velocity and recommends sending another 30 units, but the store only has space for 15.

Or consider an international retailer. A product entering peak seasonal demand in Australia might simultaneously be moving out of season in Europe. A global forecast that applies the same seasonal assumptions everywhere could create excess inventory in one market and shortages in another.

This is why the operational context surrounding AI matters so much. Useful inventory planning needs to account for supplier lead times, minimum order quantities, inventory already purchased, safety stock, ordering multiples, location-level sales velocity, promotions, seasonality, and periods when stockouts suppressed historical sales.

These details may not sound as exciting as asking an AI assistant to analyze a spreadsheet, but they are what turn a forecast into a decision you can actually use. Without them, you may simply be making faster decisions based on incomplete assumptions.

New Products Show Why People Still Matter

Historical sales can be extremely useful when you are replenishing a SKU you have carried for two years. They are much less useful when you are launching something completely new. Fashion, apparel, and other businesses with frequent product launches encounter this problem constantly.

A new style has no sales history, but you may have years of useful information about comparable products. AI-assisted planning can help identify those patterns. Instead of estimating a size curve from memory, for example, technology can analyze how similar products sold across small, medium, large, and other sizes. Your buyer can then use that recommendation as a starting point.

The important phrase is “starting point” because your people still know things the data does not. Maybe an influencer is about to feature a product, a new business is opening next to one of your stores, or a supplier has become unreliable. You might also know a SKU will soon be discontinued despite strong current sales. That business knowledge still needs a place in the planning process.

The Goal Should Be Controlled Automation

There is a temptation to talk about AI as a path toward completely autonomous ecommerce operations. A more practical goal is controlled automation, where technology handles the calculations and repetitive work while experienced people remain responsible for the decisions that require business judgment.

Let the technology do what technology does well. It can analyze sales history, calculate velocity, monitor stock levels, identify exceptions, and recommend replenishment quantities. Your people can establish the rules, review unusual situations, and override recommendations when they know something the system does not.

You do not need AI to eliminate human decision-making. You need it to eliminate thousands of repetitive calculations and low-value manual tasks so your people can concentrate on the decisions that deserve their attention. That combination can give you greater efficiency without sacrificing the operational knowledge your business has built over time.

Better Intelligence Requires Better Integration

As AI becomes more sophisticated, another question becomes increasingly important: What can the system actually see? Your forecasting technology needs reliable information about sales, inventory, incoming purchase orders, and potentially inventory movements between locations. If those inputs are incomplete or outdated, even an advanced model can produce a poor recommendation.

Integration matters on the other side of the decision, too. If your system recommends a replenishment quantity but somebody then has to copy that number into a spreadsheet, recreate a purchase order in another application, and manually update another system, you have not eliminated much operational friction.

A stronger model creates a continuous loop. Data flows in, demand is forecast, and your team reviews the recommendation. A purchase order or stock transfer is then created, inventory arrives, and the actual results flow back into the system to inform the next planning cycle. That is where AI starts becoming more than an interesting analytics tool and becomes part of your ecommerce operation.

The Real Objective is Better Margins

As you evaluate AI-enhanced demand forecasting and inventory planning technology, do not get distracted by the sophistication of the algorithm alone. Ask a more practical question: Does this make your operation easier to run? The right technology should reduce the mental workload of inventory planning while giving your team recommendations they can understand, trust, and adjust.

Ultimately, you want enough inventory to meet customer demand without tying up unnecessary cash. You want products positioned where customers are most likely to buy them, purchasing decisions that reflect real supplier constraints, and a team that spends less time manipulating spreadsheets and second-guessing routine replenishment decisions.

All of those improvements point toward the same outcome: protecting your margins while building an operation capable of supporting continued growth. The ecommerce businesses that benefit most from AI will not necessarily be the ones that automate every decision. They will be the ones that use integrated technology to manage complexity while giving their people better information, more time and greater confidence to make the decisions that matter.

About Descartes and Shiptheory

Multichannel sellers, direct-to-consumer brands, wholesale distributors, and manufacturers rely on Descartes’ ecommerce operations solutions to manage inventory, warehouse operations, domestic and international shipments, demand forecasting, inventory planning, and more. Learn more about Descartes technology at www.descartes.com and connect with us on LinkedIn and X.

Descartes Peoplevox and Shiptheory work together to connect warehouse operations directly with the shipping process. Once an order is picked and packed in Peoplevox, accurate order and package data flows into Shiptheory, where carrier selection, label creation, customs documentation and tracking can be automated. This integration reduces manual work and errors while giving growing ecommerce businesses a faster, more scalable way to manage fulfilment from the warehouse through to dispatch. Learn more about the Descartes Peoplevox and Shiptheory integration.

Author Bio

Johannes Panzer is Vice President of Ecommerce Marketing at Descartes, where he focuses on technology that helps ecommerce businesses simplify operations, scale efficiently and protect profitability. Drawing on more than 20 years of experience in ecommerce fulfillment and shipping, he provides practical insights into inventory management, warehouse operations, shipping software and the technology that supports sustainable ecommerce growth.