What AI actually costs in supply chain
Supply chain managers can optimise AI costs by focusing on integration and data work, rather than the technology itself, and attributing spend to the process being supported, such as demand forecasting, to ensure accurate assessment of costs and benefits, and make informed decisions about where to invest.
Every week, supply chain managers sit through meetings where the cost of AI deployments is discussed, and every week, the conversation seems to miss the point, focusing on the upfront cost of the technology rather than where the spend actually accumulates in a supply chain deployment.
Most people get wrong that the cost of AI in supply chain is mainly about the technology itself, but this is wrong, the cost is mainly about integration and data work, which is often overlooked.
This misunderstanding leads to a lack of planning for the actual costs that will be incurred, and as a result, the deployment of AI in supply chain can be more expensive than anticipated, and the benefits may not be fully realised.
Integration and data work as the dominant line
The integration of AI systems with existing supply chain systems and the work required to prepare and maintain the data are the dominant costs in any AI deployment, and this is particularly true in supply chain where the complexity of the systems and the data is high.
A shift supervisor checks the maintenance log to ensure that all equipment is properly maintained, and this requires integration with the AI system to provide real-time updates, which can be a significant cost.
The category manager compares two suppliers, and this requires data on the suppliers' performance, which must be collected, cleaned, and maintained, adding to the cost of the AI deployment.
Run cost against query volume
The cost of running an AI system in supply chain is closely tied to the volume of queries that the system must handle, and this can be a significant cost, particularly if the system is not properly optimised.
For example, if the system is handling a large number of queries from different parts of the supply chain, the cost of running the system can increase rapidly, and this must be taken into account when planning the deployment.
By understanding the relationship between run cost and query volume, supply chain managers can better plan for the costs of their AI deployment and ensure that the benefits are fully realised, and this can be done by monitoring the system's performance and adjusting as needed.
The cost of a wrong answer nobody caught
A wrong answer from an AI system can have significant consequences in supply chain, particularly if it is not caught and corrected, and this can lead to additional costs and reduced efficiency.
For example, if the AI system provides an incorrect forecast of demand, this can lead to overstocking or understocking, which can be costly, and the cost of correcting the mistake can be high.
Therefore, it is essential to have processes in place to detect and correct errors, and to ensure that the AI system is providing accurate and reliable information, and this requires careful planning and monitoring.
Attributing spend to a process rather than a model
When deploying AI in supply chain, it is essential to attribute spend to the process that the AI is supporting, rather than to the AI model itself, and this requires a clear understanding of the costs and benefits of the deployment.
For example, the cost of deploying an AI system to support demand forecasting should be attributed to the demand forecasting process, rather than to the AI model, and this allows for a more accurate assessment of the costs and benefits.
By attributing spend to the process, supply chain managers can better understand the costs and benefits of their AI deployment and make more informed decisions about where to invest, and this can be done by working closely with the AI deployment team to ensure that the costs are properly attributed.
When AI is the wrong choice
AI is not always the right choice for every supply chain problem, and there are times when other solutions may be more appropriate, and this requires a careful assessment of the problem and the potential solutions.
For example, if the problem is relatively simple and can be solved with existing systems and processes, then AI may not be the best choice, and the costs and benefits of the deployment should be carefully considered.
In these cases, it is essential to consider alternative solutions and to ensure that the chosen solution is the most appropriate for the problem, and this requires a careful evaluation of the options and a clear understanding of the costs and benefits, and this can be a complex and time-consuming process, but it is essential to ensure that the right solution is chosen.
Frequently asked questions
What are the main costs associated with AI deployments in supply chain?
The main costs are integration and data work, which are often overlooked and can lead to a lack of planning for actual costs incurred.
How does query volume affect the cost of running an AI system in supply chain?
The cost of running an AI system is closely tied to query volume, and handling a large number of queries can increase costs rapidly if the system is not properly optimised.
What are the consequences of a wrong answer from an AI system in supply chain?
A wrong answer can have significant consequences, such as overstocking or understocking, and can lead to additional costs and reduced efficiency if not caught and corrected.
How should supply chain managers attribute spend when deploying AI?
Spend should be attributed to the process being supported, rather than the AI model itself, to ensure a more accurate assessment of costs and benefits, and make informed decisions about where to invest.