While Generative AI has sparked great excitement, a form of artificial intelligence called Causal AI might offer much greater potential. Causal AI offers the tantalizing promise of being able to unravel the intricate web of cause-and-effect relationships that govern business operations. Georgia-Pacific (GP) has demonstrated an application of Causal AI to dramatically improve touchless commerce.
Georgia-Pacific and its subsidiaries manufacture and distribute a wide range of consumer products, including bath tissue, paper towels, napkins, tableware, paper-based packaging, cellulose, specialty fibers, and building products. With a sprawling network comprising over 150 facilities and a workforce of approximately 30,000 employees, the company ranks as the world’s second-largest forest products company.
Executing a Perfect Order is Difficult!
According to Mike Carroll, a vice president at Georgia-Pacific, “In creating a more seamless order management process, we needed a capability that enabled us to navigate the complexities of the myriad of individual orders that we receive every minute, hour, and day, with unparalleled precision. We needed to identify nuanced patterns, anomalies, and automation opportunities in near real-time and enhance our operational efficiency and customer satisfaction. More importantly, we needed to capture the knowledge of our subject matter experts on how to make all of that happen.”
Achieving a perfect order is a struggle for almost every business. Orders come in, and customers expect their product to be delivered when they want it. The goal is to deliver 100% of the orders on time and in full. But for that to occur, all of the dominos must fall into place: the product must be available to promise, there needs to be enough shipping capacity, and there needs to be an understanding of how inventory will be moved between mills and distribution centers. To solve this problem effectively, the system must also proactively identify potential bottlenecks, resource constraints, and delivery delays.
IT systems can be part of the problem. The complexities associated with an enterprise resource planning system’s available-to-promise functionality can lead an ERP system to be unable to process an order, requiring time-consuming human intervention before an accurate promise can be relayed to customers.
Available-to-promise functionality seeks to ensure a company has the requisite manufacturing and logistics capacity to deliver an order to a customer on the date they want it. However, complex process manufacturing presents a much more difficult ATP problem than is typical in discrete industries.
According to Ron Norris, Director of Innovation at GP, Causal AI was used to detect and correct “common and uncommon order errors or discrepancies in near real-time. It can do this because it was taught by the subject matter experts inside the order management team. It learned how to solve problems from the people who solve those problems every hour of every day. Causal AI can serve as a valuable agent to help in their decision-making processes.”
One of the key strengths of the implemented system lies in its ability to incorporate customer-centric insights into the order management process. It analyzes new and historical order data, customer preferences, and transactions. This allows the system to personalize and tailor the processing of orders to each customer’s expectations.
Driving User Buy In
One of the challenges associated with implementing any new solution is getting users to trust and use the system. Ideally, the Causal AI operates autonomously. Sometimes that is not possible.
According to Mr. Carroll, when the system “can’t perform an autonomous decision, the system will provide recommendations on how it would solve the problem and then explain to the employee why it is giving that recommendation. The employee can then validate the system’s recommendation or make changes to it. It’s a different way of working.”
What is Causal AI?
GP describes Causal AI as a mixture of Knowledge AI and Data AI. Knowledge AI incorporates the domain-specific expertise and best practices the subject matter experts provided. Data AI empowers the system to analyze vast amounts of data, identify patterns, and generate probabilistic outcomes in near real-time. The combination of these two things is what allows Causal AI to solve very difficult problems. This writer has been covering supply chain management for over 25 years. In my view, the problem GP tackled is not something any supply chain planning engine could hope to solve.
The true power of Causal AI comes from understanding how knowledge and data help determine causality. Causal AI utilizes sophisticated causal models to make decisions on multiple levels. A causal model graph represents a network of interconnected entities and relationships, enabling the system to understand how various factors influence each other to create an optimized outcome. By leveraging causal knowledge and data graphs, Causal AI can navigate complex business scenarios, anticipate outcomes, and recommend optimal courses of action.
Using words to describe Causal AI only takes you so far. Seeing the layers of knowledge modeled in a knowledge graph is more powerful. The following figure helps demonstrate the depth of causality that is modeled with these systems. For GP, softness is one of 12 critical product attributes for any of their paper products. Softness itself has 10 attributes called Influencing Attributes (IA) that can affect the Softness of the product. Further, each Influencing Attribute has many items that can affect them. BULK is one of those Influencing Attributes. But BULK, in turn, has many “Conditional Attributes” that affect it.
Georgia-Pacific used technology from Parabole.ai to build its Causal AI solution, and Vassar Labs built the interface. Before working with GP, Parabole.ai provided solutions for the financial industry.
GP Significantly Improved its Capabilities
GP’s goal was to see whether Causal AI could combine subject matter experts’ tacit knowledge with production data to make more intelligent and automated decisions. They demonstrated a 10X increase in touchless order throughput. Some order management errors that used to take days to resolve were now resolved in seconds. Of course, getting a promise right is vital. However, good customer service also demands quick answers to customers’ order inquiries.
The integration of Causal AI also enables transportation monitoring and optimization, automated replenishment, and can improve the alignment of the demand forecast with production plans.
Georgia Pacific decided that attacking the order management processes was a good place to test the capabilities of Causal AI. Mr. Carroll stressed, however, that while “we looked first at order management, Causal AI can be utilized in multiple areas of an enterprise where complex business problems are prevalent.” GP, for example, has also had success in demonstrating that Causal AI can solve difficult sourcing problems.
