Reducing Demand Forecast Review and Formatting Work: How AI Automation Can Support Order Decisions

Introduction

My name is Mia Sato, and I work on AI research at GDX Inc.

At GDX, we have developed an in-house AI-powered order decision support system to help reduce the workload involved in order decisions, inventory checks, and pre-meeting alignment in EC and retail operations.

For global footwear brands, or retail and EC companies selling products across multiple regions, sales trends can vary significantly depending on country, region, SKU, size, color, season, and campaign timing.

As a result, before placing orders, teams need to review a wide range of information, such as sales data, inventory levels, past order history, inbound lead times, and promotion plans.

In this article, I will introduce how our AI-powered order decision support system helps organize the information needed before order decisions are made, and how it makes it easier for order managers to identify the key points they should review.

Why Order and Inventory Management Often Requires So Much Review Time

In order and inventory management, there are many situations where numbers alone are not enough to make a decision.

For example, during a Monday morning order meeting, a team member may open several Excel files and reports while checking questions such as:

“Which products should we reorder?”
“Which products have higher inventory risk?”
“Which SKUs need human review?”
“Why is this region showing a different sales pattern?”

When the number of SKUs is small, these checks may still be manageable based on human experience.

However, when the number of products is large and sales regions are spread across multiple markets, the number of items to review increases quickly.

In addition, order decisions cannot be made by looking at sales volume alone.

Teams also need to consider current inventory levels, past order timing, inbound lead times, regional sales pace, campaign plans, and seasonal factors.

For example, in one country, demand for sneakers may increase during a specific season every year.
In another country, even the same product may sell better in different sizes.
In another region, demand may increase temporarily only during a promotion period.

If these differences are organized manually every time based only on human experience, both pre-meeting preparation and meeting discussions can take a lot of time.

This is exactly the kind of issue that GDX aims to solve with its AI-powered order decision support system.

The goal is not to let AI make the final decision.

Rather, the goal is to organize the information that order managers need before making decisions, reduce the number of items they need to check, and make it easier to prioritize what should be reviewed.

What Is the AI-Powered Order Decision Support System?

The inventory optimization system is an order support system developed in-house by GDX, using machine learning and AI.

Based on sales data, inventory data, and past inbound and order data, the system analyzes demand and provides reference values for order quantities.

What is especially important is that the system does not simply look at overall sales.

The AI-powered order decision support system combines information such as:

• Sales performance by region
• Current inventory levels
• Past order timing
• Inbound lead time
• SKU-level movement
• Sales differences by size, color, and specification
• Demand fluctuations caused by seasons and campaigns

Using an LLM module, the system analyzes sales data, inventory data, and yearly inbound and order data, then proposes order quantities based on the ordering frequency of each region.

In other words, the AI-powered order decision support system is not designed to automatically finalize orders.

It is a system that prepares the necessary information before the order manager makes the final decision.

Which products should be reviewed first?
Which SKUs have a higher risk of stockout?
Which products may lead to excess inventory?
Which proposed values should be checked again by a human?

By organizing these points in advance, the person in charge no longer needs to review every product with the same level of attention.

What Can AI Help Organize?

The order proposal report generated by the AI-powered order decision support system serves as a working draft for order decisions.

For example, users can check information such as:

• Recommended order quantities by SKU
• Demand trends by region
• Products that may soon face inventory shortages
• Products that should be ordered urgently
• Products likely to lead to excess inventory
• Differences from past order results
• SKUs that should be reviewed by humans

What matters here is that the numbers generated by AI are not meant to be used as-is.

Demand forecasting does not end simply because numbers have been generated.

Why was this order quantity suggested?
Which products should be prioritized?
Which areas require human review?

Only when this level of context is organized does the information become truly useful for order managers.

For example, instead of checking a full list of all products, the team can prioritize products with a high risk of stockout or products that show a large difference from previous order results.

Similarly, the system can highlight only SKUs with significant demand changes, while lowering the review priority of products that are moving as usual.

In this way, the role of AI is not to replace human judgment.

Its role is to support the preparation that comes before judgment.

It helps reduce the number of items that need to be reviewed.
It helps organize priorities.
It helps identify outliers and products with large changes.
It helps align discussion points before meetings.

The AI-powered order decision support system is designed to reduce this kind of upstream operational workload.

Client Validation Case

The AI-powered order decision support system has also been tested with an actual client company.

In one client PoC, we compared traditional manual ordering with order results generated using our AI-powered order decision support system.

The validation covered approximately 1,000 SKUs.

Over a period of five months, approximately 5,000 order decisions were compared.

As a result, traditional manual ordering recorded:

• Stockout count: 380
• Stockout rate: 7.6%

By contrast, ordering supported by the AI-powered order decision support system reduced the results to:

• Stockout count: 130
• Stockout rate: 2.6%

Although this was still at the PoC stage, the results showed a tendency to reduce stockouts to roughly one-third.

What is especially important is that the system does not simply “generate a forecast.”

It helps organize which products should be reviewed first by taking into account factors such as:

• Sales trends
• Inventory status
• SKU-level differences
• Inbound lead times
• Campaign impact

In order operations, even experienced team members can find it difficult to oversee everything when the number of SKUs increases.

We believe one of the major values of the AI-powered order decision support system is that it helps support order decisions while reducing the workload involved in checking items and aligning information before meetings.

Of course, results may vary depending on the industry, product characteristics, and regional differences. The same results cannot be guaranteed for every company.

However, for companies managing a large number of SKUs, being able to organize demand fluctuations and inventory risks based on sales data, inventory data, and order history can provide significant value.

Expected Benefits After Implementation

Based on the client validation results, the AI-powered order decision support system can be expected to deliver the following benefits.

1. Reducing Stockout Risk

The system makes it easier to identify products that are likely to face stockouts or inventory shortages in advance, helping teams make earlier decisions about additional orders or inventory adjustments.

As a result, it can help reduce:

• Lost sales
• Missed sales opportunities
• Urgent response work

2. Optimizing Inventory

By organizing sales performance, inventory levels, inbound lead times, and inventory status across stores and warehouses, the system helps teams identify issues such as:

• Excess inventory
• Slow-moving SKUs
• Uneven inventory allocation

These issues can be detected at an earlier stage.

This is an area where companies with a large number of SKUs are especially likely to feel the impact.

3. Contributing to Sales Growth

By making it easier to place high-demand products in the right location at the right time, the system can help reduce missed sales opportunities.

It also makes it easier to respond to situations such as:

• Inventory shortages during campaigns
• Unexpected demand increases
• Regional differences in sales trends

As a result, it can contribute to sales growth.

Where Is This System Useful in Practice?

The AI-powered order decision support system is especially suitable for companies with many products, large regional differences, demand fluctuations, and time-consuming order decisions.

One example is industries such as apparel, footwear, and sporting goods, where the number of SKUs tends to be large.

In these industries, even the same product can sell differently depending on size, color, model, gender, and region.

In one country, larger sizes may sell well.
In another region, basic colors may sell steadily.
In another region, demand for a specific model may increase sharply only during a campaign period.

Checking these differences in Excel every time and deciding order quantities manually can be a significant burden.

With the AI-powered order decision support system, teams can receive suggested order quantities based on the sales pace, inventory levels, and past order records of each region and SKU.

Instead of reviewing all products using the same standard, teams can start by checking the products that require attention.

The system is also useful in industries with frequent replenishment, such as food, beverages, and daily necessities.

In these areas, demand is easily influenced by seasons, temperature, events, and consumer habits.

If a company orders too much because it expects strong sales, it may face inventory or disposal risk.
If it estimates demand too low, it may run out of stock during a high-demand period.

By using the AI-powered order decision support system, teams can more easily determine which products should be replenished and when, based on sales pace, inventory levels, inbound lead times, and past order data.

The system can also help organize order decisions in industries such as home appliances, electronics, and parts, where products often have higher unit prices and longer lead times.

If a popular model runs out of stock, the company may lose sales opportunities.

On the other hand, if too much inventory of an older model remains, inventory burden may increase after the launch of a new model.

In other words, both stockouts and excess inventory are problems.

To balance these risks, it is important to organize past sales data, inventory data, inbound data, and order lead times so that replenishment timing can be understood more clearly by product.

What GDX Values

What GDX values is not AI implementation for its own sake.

We focus on questions such as:

Where does the operational burden occur?
Which decisions take time?
What information needs to be organized so that the person in charge can act more easily?

We believe it is important to understand this operational design first, and then apply AI and machine learning.

In order and inventory management, trying to automate the final decision from the beginning can make it difficult for the system to be accepted by the field.

This is because order decisions often include elements that cannot be judged by data alone, such as promotion plans, local sales conditions, relationships with business partners, and past experience.

That is why we believe the first step should be to organize the upstream process so that people can make decisions more easily.

Organize sales data.
Make inventory risk visible.
Narrow down the SKUs that need to be reviewed.
Summarize discussion points before meetings.
Create a draft of order quantities.

This sequence is more likely to fit naturally into actual operations.

The AI-powered order decision support system is also not intended to fully automate ordering.

Ultimately, sales companies and order managers review the order proposal report, make adjustments as needed, and finalize the order details.

AI is used to create a state where people can make better decisions.

This is the balance that GDX values.

Contact Us for Implementation Consultation

The AI-powered order decision support system may be useful for companies facing challenges such as:

• Too many SKUs to review every product in detail
• Different sales trends by region
• Too much time spent on order meetings and pre-meeting alignment
• A need to organize stockout and excess inventory risks
• Heavy workloads from Excel-based aggregation and checking
• A need to share the basis of order decisions more easily between headquarters and sales companies

For example, this may apply when team members spend weekly order meetings reviewing multiple documents to decide which products should be reordered and which products carry inventory risk.

It may also apply when sales trends differ by region, and even the same SKU requires different order decisions depending on the country.

In these situations, using sales data and inventory data to organize the upstream process before order decisions can help reduce checking work and pre-meeting alignment time.

The value of the AI-powered order decision support system is not limited to demand forecasting accuracy.

In practice, its value often appears in reducing the time required to prepare before each order, making it easier to identify the SKUs that should be reviewed, and helping meetings move directly into concrete discussion points.

If your company would like to use sales data and inventory data to organize the upstream process before order decisions, please feel free to contact GDX.

This article was partially created with the support of ChatGPT and edited by the author. It is intended to introduce the overview of the AI-powered order decision support system developed in-house by GDX. The information is based on details available at the time of publication. Service content and specifications may change. For details, please contact GDX Inc.