AI Logistics (Part 1)
What are the practical applications and misconceptions surrounding artificial intelligence in logistics?
AI has become a buzzword in various industries, and logistics is no exception. When we talk about AI in logistics, we often think of advanced technologies and futuristic applications. But it's important to understand the specific use cases of AI and how they differ from other related technologies.
When someone asks me about the data and technology that powers AI in logistics and other industries, I like to explain it in terms of the human sensory anatomy: Natural Language Processing (NLP) is the tongue of AI, Computer Vision (CV) is its eyes, Machine Learning (ML) is its brain, and large datasets are its ears. Each component plays an important role in supporting AI-powered applications.
Computer vision, for instance, provides Amazon with the datasets to identify damaged goods in its warehouses. By training AI models with this data (photos of undamaged and damaged items), the company achieved a 3x effectiveness in identifying damaged items during the picking/packing process compared to human workers. This significantly speeds up warehouse operations, improves customer loyalty, and makes employees who fulfill orders more successful at their job.
The AI checks items during the picking and packing process. Goods are picked for individual orders and placed into bins that move through an imaging station, where they are checked to confirm the right products have been selected. That imaging station will now also evaluate whether any items are damaged.
This real-world application showcases the potential of AI in optimizing logistics operations at the warehouse level, but not everything labeled as AI in logistics is truly AI. Many use cases like logistics planning, supply planning, and demand anticipation actually run on traditional data science. In other words, a human is crunching numbers and making decisions rather than an AI model and workflow. Maybe some ML is being used to reduce manual, error-prone number crunching, but without automated decision making, AI is absent.
To truly understand AI in logistics, we need to be specific and differentiate it from other technologies and trends in the field. For example, many people use the term “automated warehousing” when they’re actually referring to robotics. Robotic systems can power AI but they are not intrinsically AI.
Another example is Augmented Reality (AR) technology. Imagine an inventory manager using their smartphone camera to point at a warehouse shelf. When the camera is aligned correctly, the AR display on the screen turns green, indicating that the desired item is present on that shelf. This AR application is powered by computer vision and AI models, much like the Amazon use case.
Here, we are calling technology what it is. We’re not calling everything AI, a dirty habit that leads some people to caution against the hype surrounding AI in logistics.
This logistics reporter recently drew parallels to the blockchain frenzy in the past, arguing that companies are using AI as “shareholder-magic-dust,” pledging to implement AI without clear near-term practical applications. As I wrote about in a previous newsletter, this happened with the Blockchain in Transportation Alliance (BiTA) that 450 companies joined in 2017. Three years later, the alliance died along with the blockchain hype.
While AI undoubtedly holds immense potential for logistics applications, we must focus on tangible and realistic use cases that bring measurable benefits. This is what we’re doing at PackageX as we work toward using AI to power smart chatbots that answer customer questions as quickly as possible (great for API docs). We’re also using AI to build new features that streamline communications between shippers, transporters, and receivers without requiring truckers to be tracked.
I’ll cover these use cases more in a few weeks. In the meantime, avoid the hype of AI and focus on practical and achievable goals in the logistics industry. That’s what we’ll be doing.