• Written by Sonali Subhadarsini, Product Manager

Can we really automate the entire production planning process in metals and mining, or are there still too many variables to account for, and if so, what does that mean for the role of the shift supervisor, who has to balance the need for consistent output with the unpredictable nature of grade variability and equipment maintenance, all while trying to keep costs under control in the face of volatile input prices, a question that gets to the heart of what makes this industry so challenging, and one that has puzzled category managers and production planners for years, as they try to optimize their processes and stay ahead of the competition.

Metals and mining complexities

The metals and mining industry is unique in that it involves the extraction and processing of natural resources, which can be highly variable in terms of quality and quantity, making it difficult to predict and plan for production, and this variability can have a significant impact on the entire production process, from the initial extraction of the raw materials to the final processing and delivery of the finished product, and it is not just the grade variability that is the problem, but also the maintenance backlog, which can lead to equipment downtime and reduced productivity, and the energy intensity of the process, which can make it difficult to reduce costs and minimize the environmental impact, all of which can make it challenging to implement agentic AI workflows, which require a high degree of predictability and control.

For example, a shift supervisor at a mining operation may have to deal with the consequences of grade variability on a daily basis, as the quality of the ore being extracted can affect the entire production process, and this can make it difficult to plan and schedule production, and to optimize the use of resources, such as equipment and personnel, and it is not just the shift supervisor who is affected, but also the category manager, who has to negotiate with suppliers and manage the supply chain, and who may have to deal with the consequences of volatile input prices, which can make it difficult to predict and plan for costs, and to optimize the use of resources, such as materials and energy.

Agentic AI workflows

An agentic AI workflow is a system that can take multi-step action, using a variety of tools and resources, to achieve a specific goal, and this is different from a chatbot, which can only respond to a single question or prompt, and an agentic AI workflow requires a high degree of autonomy and decision-making ability, as well as the ability to use a variety of tools and resources, such as databases and software applications, and to interact with other systems and stakeholders, such as personnel and equipment, and it is this ability to take multi-step action and to use a variety of tools and resources that makes agentic AI workflows so powerful, and that allows them to be used in a variety of applications, from production planning to supply chain management.

The key characteristics of an agentic AI workflow are its ability to use tools, to maintain state, and to pursue a goal, and these characteristics allow it to take multi-step action and to interact with other systems and stakeholders, and to achieve a specific outcome, and it is this ability to pursue a goal that makes an agentic AI workflow so useful, as it can be used to optimize processes and to improve productivity, and to reduce costs and minimize the environmental impact, and for more information on how to implement agentic AI workflows, you can visit our website at www.asscher.ai.

Human approval and oversight

One of the key challenges in implementing agentic AI workflows is determining where human approval and oversight are necessary, and where they can be removed, and this is a critical decision, as it can affect the safety and efficacy of the system, and the ability to remove human approval and oversight is often premature, as it requires a high degree of confidence in the system's ability to function correctly and to make decisions, and this confidence can be difficult to achieve, especially in industries such as metals and mining, where the consequences of error can be severe, and where the systems are often complex and difficult to understand.

For example, a category manager may want to remove the human approval step from a procurement process, in order to speed up the process and reduce costs, but this may not be possible, if the system is not able to make decisions that are consistent with the organization's goals and values, and if the system is not able to interact with other stakeholders, such as suppliers and personnel, in a way that is transparent and accountable, and it is for this reason that human approval and oversight are often necessary, even in systems that are highly automated, and for more information on how to determine where human approval and oversight are necessary, you can contact us at www.asscher.ai/contact-us.

Failure modes and logging

An agentic AI workflow can fail in a variety of ways, from the failure of individual components, such as sensors or software applications, to the failure of the system as a whole, due to errors in decision-making or interaction with other systems and stakeholders, and it is for this reason that logging and monitoring are so critical, as they allow the organization to understand what went wrong and to take corrective action, and to prevent similar failures from occurring in the future, and logging and monitoring can also help to identify areas for improvement, and to optimize the system's performance and productivity.

For example, an agentic AI workflow may fail due to a lack of data or a failure in the decision-making process, and this failure can have significant consequences, such as reduced productivity or increased costs, and it is for this reason that logging and monitoring are so critical, as they allow the organization to understand what went wrong and to take corrective action, and to prevent similar failures from occurring in the future, and it is also for this reason that the organization should have a clear understanding of the system's failure modes, and should have a plan in place for responding to failures, and for minimizing their impact.

Limitations and alternatives

Agentic AI workflows are not a solution for every problem, and there are many situations in which they are not the best choice, such as situations in which the system is highly complex or dynamic, or in which the consequences of error are severe, and in these situations, other approaches may be more suitable, such as rule-based systems or human-in-the-loop systems, and it is for this reason that the organization should carefully consider the limitations and alternatives of agentic AI workflows, before deciding whether to implement them, and should have a clear understanding of the system's capabilities and limitations, and should have a plan in place for responding to failures, and for minimizing their impact.

For example, an organization may decide to use a rule-based system instead of an agentic AI workflow, if the system is highly complex or dynamic, or if the consequences of error are severe, and this decision should be based on a careful consideration of the system's capabilities and limitations, and should take into account the organization's goals and values, and the potential risks and benefits of the system, and it is for this reason that the organization should have a clear understanding of the system's limitations and alternatives, and should have a plan in place for responding to failures, and for minimizing their impact, and for more information on how to determine whether agentic AI workflows are the right choice for your organization, you can visit our website at www.asscher.ai.