What data does AI need for supply chain?
AsscherAi needs access to the purchase order table, inventory by location, and demand signal to function effectively in a supply chain deployment, and a logistics manager can use it to determine the best way to schedule deliveries without disrupting the supply chain, by checking the lead times and delivery history.
Some people object to using AI in supply chain management because they think it requires a complete overhaul of their existing systems, but this is a misconception. The reality is that AI can work with the data from the systems that are already in place, as long as those systems are providing the necessary information. The question is, what data does the AI need to function effectively in a supply chain deployment?
The systems that matter
The AI needs access to data from several key systems, including the purchase and delivery history, lead times, inventory by location, and demand signal. The purchase and delivery history provides information on what has been ordered and delivered in the past, while the lead times give the AI an understanding of how long it takes for orders to be fulfilled. The inventory by location data shows the AI where products are currently stored, and the demand signal indicates what products are likely to be needed in the future. A shift supervisor, for example, might check the maintenance log to see if any equipment is due for maintenance, and then use the AI to determine the best way to schedule that maintenance without disrupting the supply chain.
These systems are crucial because they provide the AI with a complete picture of the supply chain. Without them, the AI would not be able to make informed decisions about how to optimize the supply chain. For instance, if the AI does not have access to the inventory by location data, it may recommend ordering more products than are actually needed, leading to waste and inefficiency.
What happens when a system is missing
If one of the necessary systems is missing, the AI's ability to provide accurate answers will be compromised. For example, if the demand signal is not available, the AI may not be able to predict future demand accurately, leading to stockouts or overstocking. This can have serious consequences, such as lost sales or wasted resources. A category manager, for instance, might use the AI to compare two suppliers, but if the lead times are not available, the AI may not be able to give a complete picture of the suppliers' performance.
Most people get wrong the idea that the AI can simply fill in the gaps if a system is missing. This is not the case - the AI needs complete and accurate data to function effectively. If a system is missing, the AI's answers will be incomplete or inaccurate, and may even be misleading.
Defining the terms
Before the AI can be deployed, it is essential to agree on definitions for key terms such as "inventory" and "lead time". These definitions must be consistent across all systems and departments, to ensure that the AI is working with accurate and consistent data. For instance, if the inventory by location data includes both finished goods and raw materials, but the AI is only programmed to recognize finished goods, the AI's answers will be incomplete. A operations manager might need to work with the IT department to ensure that the definitions are consistent and accurate.
This process can be time-consuming and requires careful consideration, but it is essential to get it right. If the definitions are not agreed upon, the AI's answers will be inconsistent and may even be contradictory.
Quality problems that surface later
Even if the AI is deployed successfully, quality problems can still surface later. For example, if the data from one of the systems is not accurate or up-to-date, the AI's answers will be compromised. This can happen if the data is not being updated regularly, or if there are errors in the data entry process. A data analyst might need to review the data regularly to ensure that it is accurate and complete.
These quality problems can be difficult to detect, and may only become apparent after the AI has been in use for some time. It is essential to have a process in place for monitoring the AI's performance and detecting any quality problems that may arise. If you have concerns about the quality of your data, you can contact us for more information.
When AI is not the right choice
AI is not always the right choice for supply chain management. If the supply chain is relatively simple, or if the data is not available or accurate, AI may not be the best solution. In these cases, other approaches such as manual planning or traditional software solutions may be more effective. It is essential to carefully consider the specific needs and circumstances of the supply chain before deciding whether to deploy AI.
Additionally, AI may not be suitable for supply chains that are subject to frequent changes or disruptions. In these cases, the AI may struggle to keep up with the changing circumstances, and may provide inaccurate or outdated answers. It is essential to carefully evaluate the suitability of AI for the specific supply chain before deploying it.
Frequently asked questions
What data does AsscherAi need to optimize supply chain management?
AsscherAi needs data from key systems, including purchase and delivery history, lead times, inventory by location, and demand signal.
Can AsscherAi function effectively without access to all necessary systems?
No, if one of the necessary systems is missing, AsscherAi's ability to provide accurate answers will be compromised, leading to potential stockouts or overstocking.
How does AsscherAi handle inconsistent definitions across different systems and departments?
AsscherAi requires consistent definitions for key terms, such as "inventory" and "lead time", to ensure accurate and consistent data, and a operations manager may need to work with the IT department to agree on these definitions.
What happens if the data from one of the systems is not accurate or up-to-date?
If the data is not accurate or up-to-date, AsscherAi's answers will be compromised, and a data analyst may need to review the data regularly to ensure its accuracy and completeness.