Should we deploy AI/ML in the Demand Sensing engine of Supply Chains?

Most leading CPG companies have started using Demand Sensing for deployment of finished goods across their network, even though they are still using conventional time series tools for Demand Forecasting.

These leading companies are still debating whether to deploy AI/ML based algorithms for the Demand Sensing engine. We need a framework to decide when to deploy these advanced technologies to extract significant benefits for the company.

In my opinion, the following 5 pre-requisites should be in place to leverage AI/ML in Demand Sensing.

  • FG deployment algorithm should be based on inventory norms set at various supply nodes
  • Inventory norms should be based on near term demand prediction done by the Demand Sensing engine, instead of monthly demand forecasts
  • Inventory norms should incorporate the granular level impact of various demand drivers
  • Differential inventory norms should be set at the most granular level
  • Inventory norms should be dynamic (refreshed frequently) for fast response to demand shifts

If the above pre-requisites are in place, we should go ahead and use AI/ML in Demand Sensing. Else work on putting these in place first by modifying the supply chain processes.