Arthur C. Clarke made a point when he stated that beyond a certain point, technology becomes indistinguishable from magic.

AI/ML technology use and adoption in supply chain planning are still in their infancy. However, the promises of such innovative technology are already beyond the wildest dreams of many planners who still work in outdated applications or excel spreadsheets.

There is no debate that artificial intelligence and machine learning provide powerful solutions to address planning challenges. Several cases demonstrate the positive impact.

The challenge is to scale up, develop a large enough expert's pool to accelerate the learning curve for all of us and increase our collective expertise in this field.

Many used cases show great promise. Below are the three main ones we see gaining momentum in the supply chain space:

 

1. AI/ML Forecasting

Forecasts used to be based on statistical time series. The techniques and applied math focused on cleansing past sales history to determine events and correlations that might drive a deviation of the future sales from history.

Former GE exec Ian Wilson has a point when he stated, "No amount of sophistication is going to allay the fact that all of your knowledge is about the past and all your decisions are about the future."

However, AI/ML forecasting is proving him wrong, beating the forecast accuracy obtained from traditional supply chain techniques and even taking out bias from consensus forecasts based on human market intelligence from sales execs and market experts.  

There is clearly a benefit from testing a multitude of methods and approaches, applying probabilities, and learning from past experimentation. 

 

2. AI/ML Demand Sensing and Shaping

Human knowledge and traditional statistical sales forecasting techniques might indeed be about the past and the likelihood some assumptions will occur.

Artificial intelligence, on the other hand, can go beyond traditional forecasting to leverage specific data indicators of future trends and market patterns invisible to a planner.

By sensing where our demand deviates from old and outdated patterns, the algorithm can auto correct these demand signals for the current month and beyond, and continuously add new insights to the market forecasts. 

Tracking predictions in real-time gives AI/ML the ability to sense the changes and apply intelligence from contextual data previously unavailable to forecasting teams to learn and improve their initial predictions.

Some of the examples are shopper data at the point of consumption, shopping basket analysis, competitor price changes, logistics transactions, weather data, number plate registrations, store visits, eCommerce traffic or any other information (features) that could help us to continuously adjust our knowledge about the near future to explain the order book changes.

By keeping a finger on the consumer's pulse, and identifying predictors of pattern changes, AI/ML can adjust and learn in real-time how the order book materializes and continuously improve its predictions, much faster than a planner would do and based on far more parameters. 

Artificial intelligence cannot only interpret far more data about the dynamics in the market, but it can also learn, influence, and interact with the consumers in real-time with targeted customer-specific promotional and pricing tactics.

By influencing the consumer based on their buying behaviour and known preferences, machines learn to shape the demand towards the business objectives or support maximizing sales based on business strategy and product availability.

The combination of demand sensing and shaping over time will help us achieve far better and less biased forecasts, making the order book a dynamic and influenceable snapshot - subject to scenarios and probabilities - rather than a static picture based on a single range of numbers. 

 

3. Self-Healing Planning Data Management

A digital twin representing your company's supply chain is the foundation for optimized planning decisions and operational plans that can be executed. 

Data quality, accuracy and timeliness is the number one challenge in trusting a supply chain plan. Delayed information updates across a multitude of databases, duplicated data, and information prone to errors are common in supply chain management.

With a self-healing artificial intelligence and machine-learning capability, one can continuously scan the network layout data, including complex planning parameters like lead times, inventory targets or yields, to identify and auto-correct inconsistencies. 

That's the promise of AI/ML. The combination of demand sensing, demand shaping, and self-healing supply chain might sound like magic, but the capabilities are available today.

And they are about to go mainstream !

 

This article was contributed by our expert Alex Rotenberg

 

Frequently Asked Questions Answered by Alex Rotenberg

Q1. Will AI take over the supply chain?

Artificial Intelligence will never take over supply chain management. This is a false assumption fed by technology vendors making the analogy with autonomous cars. But they ignore the fundamental difference between cars and supply chains. Cars drive towards a predefined destination, which means the algorithm can optimize how to reach the specified destination. 

Supply chains provide different options to business leaders about how to best satisfy the customers, and the outcome is not a given. Therefore, the ultimate decisions in the supply chain will always have to be taken by planners and their management.

AI will play a critical role in speeding up and automating part of the decision-making for business-as-usual and running parts of the operations on autopilot.

 

Q2. Which companies are using AI in the supply chain? 

Many supply chain leaders in manufacturing and retail are already using some types of AI to improve specific functions or areas of their supply chain, mainly in planning support or process automation.

There are various used cases where technology provides incredible benefits compared to more traditional technologies and techniques. The challenge today is that these used cases are often implemented on a limited scale and are difficult to deploy across the company, industry or other functions.

In most areas, we are still mainly in the pilot phase, with a few leaders gaining immense expertise versus the rest of the industry. 

 

Q3. How is AI influencing customers and transportation? 

Customer service is the primary goal of supply chain improvement initiatives. AI can be a great support to help ensure the availability of products and proactively track the delivery status or support commitments for specific delivery dates. 

To date, supply chain AI technology focuses on prediction and automation of back office planning and execution, rarely on the customer interface, which is more in the improvement scope of customer service teams. 

 

Q4. How can IoT help in the supply chain?

IoT is a critical investment to receive real-time status updates about production, sourcing, transport or product quality and, therefore, a critical data feed to improve the quality of AI/ML support. IoT can feed data to improve predictability and automation in various areas. Think about IoT in a store, warehouse or container to track what is occurring. 

 

Q5. What benefits would happen if we implemented AI in logistics?

AI in logistics provides better traceability and more proactive solutions to address logistics disruptions, delays, and problems.

 

Q6. What is the latest supply chain technology?

There is no latest technology. Gartner publishes a hype cycle of all technologies available for the supply chain with various maturity levels. The convergence and combination of these technologies to address specific used cases are how the industry innovates in supply chain management.