• Written by Sonali Subhadarsini, Product Manager

This week, a logistics manager at a large manufacturing firm has to decide which carrier to use for a shipment of critical components, and what route to take, based on the current performance of each carrier, and the likelihood of delays or losses, which they currently assess by checking the carrier's website, and calling their account manager, a process that can take several hours, and may not always yield the most up-to-date information.

Logistics complexity

Logistics operations are inherently complex, with many variables interacting in subtle ways, and the performance of carriers can vary significantly, depending on factors such as traffic, weather, and road conditions, making it difficult to predict exactly when a shipment will arrive, and what condition it will be in, which is why logistics managers often have to resort to manual checks, and phone calls, to get the information they need, a process that is time-consuming, and prone to errors.

The problem is further complicated by the need to handle exceptions, such as missed deliveries, or damaged goods, which can be time-consuming, and require a lot of manual intervention, and the fact that dwell and detention times can vary significantly, depending on the carrier, and the route taken, which can add to the overall cost, and complexity of the shipment, and make it even harder to predict exactly when it will arrive.

Agentic AI workflows

An agentic AI workflow is a system that can take multi-step actions, rather than just answering a single question, and can use various tools, and systems, to achieve its goals, such as checking the carrier's website, and calling their account manager, or even booking a shipment, and tracking its progress, which can save time, and reduce errors, and make the whole process more efficient, and reliable, and can be accessed through a platform like AsscherAi, which allows users to query their own data, and get answers in real time.

The key characteristic of an agentic AI workflow is that it has a goal, and can take actions to achieve that goal, using various tools, and systems, and can adapt to changing circumstances, such as changes in the carrier's performance, or unexpected delays, and can learn from experience, and improve its performance over time, which makes it a powerful tool for logistics managers, who need to make quick, and informed decisions, in a fast-paced, and dynamic environment.

Human approval

While agentic AI workflows can automate many tasks, and make the logistics process more efficient, and reliable, there are still some situations where human approval is necessary, such as when a shipment is particularly valuable, or sensitive, and requires special handling, or when there are exceptional circumstances, such as a natural disaster, or a major traffic accident, that require manual intervention, and judgment, and cannot be handled by the system alone, and in these cases, the logistics manager needs to be able to review the situation, and make a decision, based on their experience, and expertise.

Removing the human approval step prematurely can be problematic, as it can lead to errors, and mistakes, that can have serious consequences, such as delayed, or lost shipments, and can damage the reputation of the company, and erode customer trust, and confidence, which is why it is essential to carefully evaluate the situation, and determine when human approval is necessary, and when it can be safely automated, and to have a system in place for handling exceptions, and unexpected events, such as a support team that can provide assistance, and guidance.

Failure modes

Agentic AI workflows can fail in different ways than traditional models, as they are designed to take actions, rather than just make predictions, and can have unintended consequences, such as booking a shipment with the wrong carrier, or routing it through a congested area, which can lead to delays, and losses, and can be difficult to diagnose, and repair, as the system is complex, and dynamic, and has many interacting components.

When an agentic AI workflow fails, it is essential to log the error, and analyze the situation, to determine what went wrong, and how to prevent it from happening again, which can involve reviewing the system's performance, and identifying areas for improvement, such as updating the carrier's performance data, or adjusting the routing algorithm, and can help to improve the overall reliability, and efficiency of the system, and reduce the risk of errors, and mistakes.

Limitations

Agentic AI workflows are not a panacea for all logistics problems, and there are situations where they are not the best solution, such as when the logistics process is relatively simple, and straightforward, and does not require complex decision-making, or when the data is incomplete, or inaccurate, and cannot be relied upon to make informed decisions, and in these cases, other approaches, such as traditional modeling, or manual intervention, may be more effective, and efficient.

In general, agentic AI workflows are best suited for complex, dynamic logistics operations, where there are many variables interacting, and the performance of carriers can vary significantly, and where the ability to adapt to changing circumstances, and learn from experience is essential, and can help to improve the overall efficiency, and reliability of the logistics process, and reduce the risk of errors, and mistakes, and can be a valuable tool for logistics managers, who need to make quick, and informed decisions, in a fast-paced, and dynamic environment.