Sometimes AI and IoT are presented as if they involve the same thing but they are actually two different areas. IoT connects and monitors physical devices while generating information about their operation. AI can consume this and related information, spot patterns, and possibly take actions.
Together, they create an interesting stack for inventory management.
Inventory AI + IoT Pipeline (Simplified View)
We can present an abstracted architecture resembling the interaction of AI and IoT for inventory operations
Physical Inventory
↓
Sensors, RFID, Barcodes
↓
IoT Gateway, Edge Devices
↓
Cloud, Data Platform
↓
AI and ML
↓
Analytics, Actions
↓
Inventory, Warehouse
Each of these inventory layers presents different concerns.
IoT is used for collecting data.
A data platform stores and manages the information.
AI/ML consumes and analyses it.
Inventory applications make sense of the outcomes.
Data That Can Be Collected with IoT for Inventory Applications
Depending on the use case scenario, connected inventory environments can potentially process and analyse information from various sources
RFID Tags and Readers
Barcode Scanners
Temperature Sensors
Position Sensors
GPS Devices
Cameras
Connected Warehouse Equipment
Inventory Transactions
ERP and WMS Systems
The Inventory Master Describes Inventory and Supply-Chain Systems Involving Technologies Such as IoT Sensors, RFID, Barcode Tracking, Cloud Platforms And Inventory-Management Software
Meaning that an inventory application stack does not necessarily need to rely solely on manually inputted entries. It can potentially consume and analyse information provided by various connected systems.
The Role of AI / ML in Inventory Forecasting and Planning
With the availability of data, machine-learning or analytical models can be used. There are different methods and approaches, but for inventory planning, there are specific functions that have been developed.
A possible application for forecasting demand could look like this
Demand History
+
Current Inventory
+
Sales / Orders
↓
Forecasting Model
↓
Expected Demand
↓
Planning / Replenishment
The Inventory Master Describes AI/ML-Based Inventory Forecasting Including Trend Analysis, Seasonality Analysis, and Multi-Location Planning
The question is not what AI/ML could do for inventory, but rather what business problems these technologies could address.
As long as relevant information is available to the model and the task corresponds to its purpose, it can be applied. This is where the accuracy of historical data and the relevance of input features start to become an important consideration.
Anomaly Detection in AI-Driven Inventory Activities
Another example can be the detection of anomalous patterns of inventory activity
Let us imagine a scenario where a series of inventory movements are processed by a system:
09:10 → SKU A → Warehouse 1 → Movement
09:15 → SKU A → Warehouse 1 → Movement
09:18 → SKU A → Warehouse 1 → Movement
09:19 → SKU A → Warehouse 1 → Unusual movement
A simple analytical model can look for patterns that have previously occurred and determine that an inventory item has, for example, moved at an unexpected time or place. Instead of having to manually review each single movement, an employee can look into an isolated set of inventory data that represents an issue.
The Value of IoT in Warehousing and Inventory Operations
One strength of IoT is that inventories and warehouses reside in a physical environment.
Inventory items are tangible and their movements count as a physical operation but software systems track them on a digital environment. The introduction of connected devices is advantageous in that it can provide the software with insights about what is physically happening.
RFID Reader
↓
Inventory Event
↓
IoT Infrastructure
↓
Inventory Information
↓
Analytical Process
↓
Monitoring Dashboard
This helps build a more continuous and updated view of an inventory state. The Inventory Master Describes IoT-Enabled Supply-Chain Systems Utilising Connected Devices, Sensors, Gateways, Edge Computing, And Cloud Infrastructure
The Use of Computer Vision in Inventory and Warehousing
Computer Vision provides yet another modality for connecting physical inventory items with digital data.
Instead of relying purely on RFID or Barcode readers, we can equip cameras and utilise models to detect objects, read labels, and recognise patterns.
Applications could potentially include automated inventory counts, product and object recognition, barcode and QR detection, pallet and container monitoring, visual anomaly detection
A potential architecture could be structured as follows
Camera
↓
Image, Video
↓
Computer Vision Model
↓
Object Detection
↓
Inventory Event
↓
Inventory / Warehouse Management
This is particularly an interesting development in scenarios where physically counting inventory can be a lengthy and resource-draining process.
Edge vs. Cloud Architecture (IoT Inventory Use Cases)
Another interesting consideration is whether the processing should happen in an edge or cloud environment.
A cloud-first approach might provide more analytical possibilities at a central location while an edge setup might allow for faster processing of information that is closer to where it originates.
For an inventory context, the selection can be based on aspects such as latency, bandwidth, network reliability, data size, available infrastructure, operational environment, and application-specific considerations. There is no universally applicable reference architecture for every inventory operation, warehouse, or logistics environment
The Challenge of Implementation Around an AI Core
The difficulty of actually implementing these ideas often resides not in the core AI but the surrounding context
A logistics or inventory environment already features a complex set of data and processes.
ERP
├── Orders
├── Purchasing
└── Finance
WMS
├── Locations
├── Pick, Pack, Ship
└── Other Activities
IOT
├── Sensors
├── RFID
└── Other Equipment
Inventory Platform
├── Operations Activities
└── Other Functions
Any AI model that is supposed to understand the inventory environment needs to consume relevant information and will be as useful as the data that is fed into it. This is why Application Programming Interfaces (APIs), data integration, data pipelines, standardised object identifiers (SKU), and data quality are important elements for AI/ML inventory projects
AI Alone Cannot Make up for Data Problems
A principle that should always be remembered is that
AI is only as good as the quality of input data.
Before starting to think about introducing AI to inventory operations, we need to assess whether the data that is used is relevant, high-quality, and consistent across sources. If there are issues with inventory data (duplicated items, missing transactions, incorrect value, delayed or wrong information, etc.), an AI model will simply produce questionable results
Therefore, instead of asking
where can we introduce AI to improve our inventory operations?
we should consider
do we have the data infrastructure to allow AI to provide value?
These two perspectives shift the implementation design heavily.
Inventory Software + AI + IoT – An Example Architecture
The inventory environment can look different depending on the size and scale of operations but a conceptual modern environment might feature
THEORY DATA SOURCES
RFID | Barcode | Sensors | Cameras | ERP | WMS
↓
DATA INGESTION PIPELINES
↓
Io T / Edge Layer
↓
Cloud / Database
↓
ANALYTICS / AI / ML
↓
DASHBOARDS / ALERTS / APIs
↓
INVENTORY OPERATIONS
As can be seen, the technology will vary greatly depending on the use case.
A small environment will involve only inventory software and barcode scanners. A larger warehouse can potentially benefit from RFID tags and readers, IoT sensors, vision systems, forecasting models, and various automation components in combination with inventory applications and analytical tools. Everything else should follow from the required application and not the opposite.
Final Word
The strength of AI and IoT for inventory and warehouse management lies in their combination. IoT can provide information about the physical environment. AI can process that and potentially identify patterns. Inventory applications can consume this information and apply it to specific scenarios. This combination results in new opportunities to make inventory and warehouse operations smarter, but only if the technology is applied with a precise purpose in mind.
For an overview of how inventory software, IoT, RFID, AI forecasting, and similar technologies apply to modern inventory operations, The Inventory Master provides more information at the following link:
Top comments (0)