• Written by Azhar Aliyar, Director

In theory, a system that answers questions about an organisation's supply chain should be able to take action to manage that supply chain, but in practice, most systems stop at answering questions, leaving the action to humans. This limitation is not inherent, but rather a result of the way these systems are designed, often as chatbots that provide information but do not take multi-step action.

What separates an agent from a chatbot

An agent is separated from a chatbot by its ability to use tools, maintain state, and work towards a goal. A chatbot, on the other hand, is typically designed to answer a single question or provide a simple response, without the ability to take further action or maintain context over time. For example, a shift supervisor checks the maintenance log and sees that a machine is due for maintenance, an agentic workflow could automatically generate a work order and assign it to a maintenance technician, whereas a chatbot would simply provide the information and require the supervisor to take action.

This difference is critical in supply chain management, where the ability to take multi-step action can be the difference between meeting customer demand and failing to do so. By automating these actions, an organisation can reduce the risk of human error and free up staff to focus on higher-level tasks.

Where a human approval step belongs

Many organisations are hesitant to automate supply chain management tasks, citing the need for human approval and oversight. However, this approval step often belongs at the beginning of the process, where the goals and parameters of the workflow are set, rather than at the end, where the workflow is executing. Removing the human approval step altogether is usually premature, as it can lead to unintended consequences and a lack of accountability.

For instance, an agentic workflow that automatically orders inventory based on demand forecasts may need human approval to set the forecast parameters, but once those parameters are set, the workflow can execute without further human intervention. This allows the organisation to balance the need for automation with the need for human oversight and control.

The supply chain data this depends on

An agentic workflow for supply chain management depends on a range of data, including purchase and delivery history, lead times, inventory by location, and demand signal. This data is used to inform the workflow's decisions and actions, and to ensure that the organisation is meeting customer demand while minimising waste and excess inventory. For example, a category manager compares two suppliers and sees that one has a consistently shorter lead time, an agentic workflow could automatically adjust the purchasing parameters to favour that supplier.

Access to this data is critical, and organisations that do not have it may need to invest in data collection and integration efforts before they can implement an agentic workflow. This can be a significant undertaking, but it is essential for achieving the full benefits of agentic workflows in supply chain management.

What it changes about lead time variance and inventory

An agentic workflow can help to reduce lead time variance, which is often hidden behind a stable average, and inventory held against the wrong risk. By automating the supply chain management process, an organisation can reduce the variability in lead times and ensure that inventory is held against the correct risks. For example, an agentic workflow can automatically adjust inventory levels based on changes in demand or lead times, ensuring that the organisation is always holding the right amount of inventory.

This can have a significant impact on the organisation's ability to meet customer demand, as well as its bottom line. By reducing lead time variance and inventory held against the wrong risk, an organisation can reduce waste and excess inventory, and improve its overall efficiency and effectiveness. To learn more about how to apply this in your operations, visit our operations page.

Measured with lead time variance, fill rate, and inventory turns

The effectiveness of an agentic workflow in supply chain management can be measured using a range of metrics, including lead time variance, fill rate, and inventory turns. These metrics provide insight into the workflow's ability to reduce variability, meet customer demand, and minimise waste and excess inventory. By tracking these metrics, an organisation can refine its agentic workflow and ensure that it is achieving the desired outcomes.

For example, an organisation that implements an agentic workflow to manage its inventory may see a reduction in lead time variance, an increase in fill rate, and an improvement in inventory turns. These improvements can have a significant impact on the organisation's bottom line, and can help to drive business growth and success. If you have any questions about how to implement an agentic workflow, please contact us.

What this does not do

An agentic workflow for supply chain management is not a replacement for human judgement and oversight, but rather a tool to support and augment it. It is not a solution for organisations that do not have a clear understanding of their supply chain operations or that lack the data and infrastructure to support an agentic workflow. In these cases, an organisation may need to invest in data collection and integration efforts, or develop a clearer understanding of its supply chain operations, before it can implement an agentic workflow.

Furthermore, an agentic workflow is not a one-size-fits-all solution, and may not be suitable for all organisations or supply chain operations. For example, an organisation with a highly complex or variable supply chain may require a more customised solution, or may need to implement additional controls and safeguards to ensure that the workflow is operating effectively.