Manufacturers frequently rely on traditional enterprise resource planning (ERP) to achieve visibility over order fulfillment, and preempt supply chain disruptions. Yet many have yet to deploy artificial intelligence to tap into and interpret the critical data that’s stored within those systems to determine actual order status. 

IBM, not surprisingly given its history of embracing AI, is ahead of the game. Its Supply Chain & Technology Systems (SC&TS) initiative, residing within the Chief Information Officer (CIO) organization, maintains the applications and data that drive manufacturing, while developing AI models for real-time visibility of key processes. 

One such AI innovation is EmbarcAI, which ingests manufacturing and order fulfillment table data and provides real-time predictions as to whether sales orders will be shipped late, early or on time. The IBM team has also implemented business-rule automation, alongside the model’s predictions. In the process, EmbarcAI advises where orders are in the supply chain, predicts their estimated time of arrival, and explains why orders are delayed. 

According to IBM, the innovation ensures customer satisfaction, creates healthy manufacturing metrics, and avoids millions of dollars in misallocated revenue.  

Users can access the predictions in the form of donut charts, which tell if an order will be late. For orders that are awaiting interventions in the manufacturing process, the charts show if they’re shippable, not shippable or at risk of going on hold. 

The model uses historical data from several common SAP manufacturing and order-fulfillment tables. At the beginning of the process, a Java-based data pipeline runs queries against 23 tables, looking at more than 400 fields that cover two primary hardware manufacturing plants. 

The data-preparation pipeline refines the 400 fields to eight significant features, from which two are calculated and two are preprocessed. They encompass the SAP manufacturing and shipping status, error codes, alterations of order and order value, which are essential to the time required to move from manufacturing to delivery and shipping complete. 

The models are constantly evaluated for accuracy and error metrics. They are measured by comparing the differences between predicted remaining hours for shipping and the historical data for the planned sales order ship date (PSSD). Route-mean-square error (RMSE) is calculated with the variability of the differences between estimated remaining hours for shipping and the historical value of PSSD. 

According to IBM, most companies using an application such as SAP Ariba tend to lack the resources to develop AI models that can harness the power of the data in their SAP tables. Regardless of whether orders are delayed due to missing parts or data error, an AI model such as EmbarcAI can make real-time predictions. 

The model is currently used in the two largest IBM hardware manufacturing plants in Guadalajara, Mexico and Poughkeepsie, New York. (IBM said it plans to expand the solution to its Europe plant in Hungary.) The company shares usage among its order-fulfillment analysts, as well as with shipping and logistic professionals from other supply chain partners. 

Previously, the process of checking orders was manual and time-consuming, using multiple tools to obtain order status, and causing many orders to miss their shipping deadlines. When that happened, IBM wouldn’t be able to claim revenue from the order for the month in question, affecting its financial reporting as a public company. 

Manual validation of order status used to take about 30 minutes to run and validate queries. Now, users can now see in near-real-time precisely which orders need attention before the PSSD. 

Implementing AI and business-rule automation has helped IBM to better predict and reduce the number of orders delayed in shipping, enhancing customer satisfaction and enabling the correct allocation of forecasted monthly revenue, the company said. As a result, it has been able to capture millions of dollars of revenues generated by sales orders within the appropriate financial period. 

Resource Link:

IBM CIO, https://www.ibm.com/case-studies/cio-office-turbonomic