AI in Logistics Management
Updated: Jun 6, 2021
You saw the title of this blog and you still came. Madame/Sir, thank you for your courage.
Artificial Intelligence (AI) is expected to add $120Bn in value in the logistics and network management fields according to Mckinsey. AI in Egypt is also expected to boost the country's GDP by 7.7% in 2030 as per the prediction of the Minister of Higher Education. Moreover, it is expected that logistics expenditure in Egypt will reach a record of $50Bn in 2024.
Logistics management is very tough on its own and by adding AI into the equation it might be a cause of panic to most. That said, AI should be thought of as nothing less than transformative for how logistics can be managed.
In this piece, I will be exploring with you some ideas of AI applications that would streamline logistics management and help companies gain competitive advantage in this arena. I will be mainly tackling the problems from an analytics point view, so none of that autonomous drones, transformers or aliens built the pyramids stuff in any way, shape or form.
So, here goes.
When it comes to managing logistics, businesses are attempting optimization to find the best solution for getting products/shipments to their customers in the fastest and least costly way possible.
Inflation, constant change in the geo-political environment, and fluctuations in the unit cost of transportation render plans which were optimum at a certain point in time not as good. Schedules always need updating and synchronization between different functions is a constant necessity.
Logistics is a market that highly depends on calculability, which AI has proven to have a significant impact upon. To that matter, AI in the Egyptian logistics market can prove to be a potential investment that will streamline the market and create a lot of efficiencies in it.
Let us first define the scope of what I mean by logistics in this piece. Here, I am considering two main horizons to solve for. The first is to transport a product from various origins or factories to different destinations, markets or outlets at minimum cost. We must find the least costly way of transportation and optimize delivery schedules to be as cost-efficient as possible.
Secondly, there is the time element, meaning the time required for transportation is important. Think of it as what is the least time required to transport the goods.
"Usually the ones who get 5% fewer of their planes shot down, or use 5% less fuel, or get 5% more nutrition into their infantry at 95% of the cost” -Jordan Ellenberg
A main branch under AI is Machine Learning. Machine learning basically studies and analyzes massive data sets that consist of loads of variables and comes up with its own understanding of said data sets, creating on it a logic by which it can optimize and predict different variables within the problem at hand.
When it comes to logistic cost management, supply chain managers usually try to optimize freight costs, on-time shipping, perfect order ratios and much more. Trying to perfect all of these variables can be a gruesome task sometimes. To that matter, a logistics dashboard might look as such:
Many of these KPIs are influenced by loads of different variables such as traffic, weather, inflation, accidents and many more. Supply chain managers usually optimize their performance indicators using forecasts and other calculations. This sometimes lacks the dynamism that is currently required to achieve an efficient logistic operation, where both speed and control are critical.
Egypt deals with major fluctuations in traffic, constant changes in road planning and many other variables that would make the use of AI in Egypt, with its ever changing business environment, particularly useful.
To that matter, AI can add a lot of calculability, predictability, control and dynamism to any supply chain process. AI algorithms will analyze and draw predictions from great amounts of data such as browsing behavior, purchase history and demographic norms as well as unrelated data sources such as weather data, social media chatter and news reports to predict customers' purchase pattern while proactively planning how products will reach their intended targets at the lowest cost and minimal time.
Now, let's take a simple use case to demonstrate better how algorithms can be effective.
AI in Demand Predictions and Transportation Cost Optimization
AI algorithms are great in correlating and optimizing big numbers of variables together to find the minimal costs and/or shortest path to a given problem.
Let's take the most common variables found in a Logistics problem in its simplest form:
X - Number of sources [factories or warehouses]
Y - Number of destinations [outlets or warehouses when shipping from factories to warehouses]
A - Number of available items [finished goods ready to be shipped]
D - Demand on the items [Outlet needing a certain volume of finished items]
C - Cost of transporting an item from the Xth source to the Yth destination
We want to minimize the total C (Cost) of transporting A (Available Items) from X (Source) to Y (Destination) based on D (the Predicted Demand).
E.g. Imagine a business owning three warehouses in the Cairo Suburbs and needs to deliver their finished products to its three shops in the city by tomorrow. This Egyptian company is using AI to dynamically and continuously predict with high accuracy the demand on these three shops.
Assuming the A.I, looking at historical sales, marketing expenditure, seasonality, geographical data, analyzing text from social source, predicted a demand on the 3 shops as follows:
Zamalek Shop: 4 Units
Maadi Shop: 2 Units
New Cairo Shop: 3 Units
Now the AI module that is connected to the ERP system, is collecting data on the current stock level of the product in three warehouses A,B & C. Now the stock available at the 3 warehouses is:
Warehouse A has 8 Units
Warehouse B has 6 Units
Warehouse C has 3 Units
Based on the historical data, the delivery costs per unit from each warehouse to each store will differ depending on the distance. Differences in cost are as follows (EGP/Unit):
Data Scientists and Machine Learning engineers use this data to build optimization algorithms integrated within the AI module that finds the least expensive way to deliver products from warehouses to shops to meet the demand. This way, customers will find the product they are looking for (whether it is now or in the future). Based on the supply constraints and the different distance-based costs, the module is able to find the optimal route. This results in the following:
Warehouse A to deliver:
No products should be delivered from warehouse A.
Warehouse B to deliver:
1 Unit to Zamalek shop:
Zamalek wanted 4 units, 25% of the demand is fulfilled from Warehouse B
Price from Warehouse B to Zamalek is 40EGP
Total cost: 40*1 = 40EGP
2 Units to Maadi shop:
Maadi wanted 2 units, 100% of the demand is fulfilled from Warehouse B
Price from Warehouse B to Maadi is 20EGP
Total cost: 20*2 = 40EGP
3 Units to New Cairo shop:
New Cairo wanted 3 units, 100% of the demand is fulfilled from Warehouse B
Price from Warehouse B to New Cairo is 20EGP
Total cost: 20*3 = 60EGP
Warehouse C to deliver:
3 Units to Zamalek shop:
Zamalek wanted 4-units, so 75% of the demand is fulfilled from Warehouse C
Price from Warehouse C to Zamalek is 20EGP
Total cost: 20*3 = 60EGP
In simple terms, this solution mimics how this problem would be tackled by the module, giving us a total cost of:
40+40+60+60 = 200EGP
For demonstration purposes, the problem here simplifies all common supply and demand constraints. In deployed algorithms, the granularity of this problem expands to higher dimensions as chained and store to store deliveries for example. Not to mention that time is not considered in this example as well, which could sometimes be of a critical nature and can directly affect the profitability of moving the products.
In these logistic problems, we run many simulations to choose the optimal way of delivering a product to the outlet considering many parameters and solving for least cost and minimal time in the process. Simulation models determine results by assessing the dynamic behavior of a system for a set parameters (warehouse operation costs, freight costs, routes etc.). With these variables always changing and the growing complexity of the inter-relationship between the required optimization and parameters of a big logistic operation, human-based optimization is difficult, error prone and requires regular revision.
A visual depiction of an AI simulation in action, showing a best case scenario for warehouse to outlet deliveries.
Source: Synapse Analytics.
The great thing about these AI modules is their ability to continuously calculate and adapt to daily situations and learn from them. This adds a dynamic layer in logistics management that can have businesses better allocate their resources and liberate massive amounts of cash from these optimizations. AI-powered Logistics can up-heave the customer experience to another level, by delivering the goods to customers before having even ordered them.
Egyptian companies who leverage AI can effectively predict demand and shorten delivery times by moving inventory closer to customer locations and allocating resources and capacity to allow for previously unforeseen demand. The cost savings enabled from these continuous optimizations will add competitive advantages to any business availing its products and up-heaving its bottom line.
AI is now more accessible than ever, and it is proving to surpass human cognitive abilities in so many domains, particularly in the problems that depend on very complex calculability and predictability as logistics management. As the car engine took us from one horse to a thousand, AI is taking us, in the knowledge worker era, from one brain to a thousand!
Founder & CEO of Synapse Analytics Ahmed Abaza on AI in Logistics Management. Synapse is a data science and AI company based in Egypt. Synapse Analytics
Want to make your operations A.I. powered?