The Bravado of Artificial Intelligence
By Jon and Matt Schreibfeder
Companies need good software to manage their operations. Organizations can choose from literally hundreds of enterprise resource planning (ERP) systems. Each of these software packages has a unique set of capabilities and features. It is a daunting task to choose the best solution for a particular business.
Software companies’ unrealistic claims complicate this decision. Today we visited a web site for a new AI based system that claimed if you used their “cutting-edge” technology, you would “Never Stock Out” and “Never Overstock”. Do they really feel that their proprietary algorithms will be able to predict and compensate for every future supply-chain disruption, or change in your customers’ buying habits? They must believe their system has “perfect knowledge” because they claim to provide 100% accurate forecasts of future demand of products.
However, in their discussion of safety stock they vary service level targets based on an individual product’s rank:
Product Rank Service Level Target
A 95%
B 90% C 85%
D 80%
Remember that the definition of service level is the percentage of customer product requests that can be completely filled from stock inventory. How could they promise that you would “never stock out” yet only plan to totally fill eight of 10 (i.e., 80%) of product requests for “D” ranked items? And, if their forecasts were 100% accurate why would they need any safety stock (which provides protection against stockouts) at all? It is obvious that this software company is making both unrealistic and contradictory claims concerning their system’s potential for improving inventory performance.
Artificial Intelligence (AI) can be utilized to predict the future based on available data from a variety of sources which it can access. But it cannot perform miracles such as preventing all stockouts. And it cannot consider information it doesn’t have access to such as whether you will experience unusually high demand for a product due to an unforeseen circumstance. Or demand for another item will dramatically decrease due to a customer changing suppliers. Best practice forecasting is achieved by combining a comprehensive analysis of available data with market knowledge.
This market knowledge is acquired from salespeople, customers, vendors and other sources. This information must be examined, evaluated, and selectively applied to produce the most accurate forecast possible. It is possible for an AI system to “learn” from the results of this collaborative information. But the sources are too numerous and vary in accuracy as well as format to completely automate this task. Experienced buyers and inventory planners are necessary to evaluate market knowledge and selectively apply it to predictions of future demand. And no matter how much technology and human effort is utilized in managing inventory, predictions will never be perfect. In order to consistently meet customers’ expectations of product availability, you will always need some safety stock. That is, “insurance inventory” to prevent stockouts due to unexpected demand or delays in receiving a replenishment shipment.
Forecasting and replenishment systems that utilize AI will help your organization improve inventory performance. But to achieve success, these systems must be supplemented with knowledge acquired from employees and the marketplace. Beware of consultants, software companies and others who state that their technological solution can provide “inventory perfection” in a matter of weeks, without the help of knowledgeable human beings.