• Written by Muhammed Abdulla NC, CEO

This week, a production manager at a large chemicals manufacturer must decide whether to adopt a new enterprise AI system to optimise yield in their continuous processes, and the decision will be based on the system's ability to handle the complexities of yield drift, feedstock variability, and plant-to-plant recipe differences.

Chemicals' unique challenges

The chemicals industry poses unique challenges for enterprise AI adoption, primarily due to the tight yield and safety envelopes that govern continuous processes. Yield drift, which refers to the gradual decline in production efficiency over time, can be difficult to predict and correct. Feedstock variability, where the quality and composition of raw materials vary, can also significantly impact production outcomes. Furthermore, plant-to-plant recipe differences, where identical processes are implemented differently at various sites, can make it challenging to develop and deploy a single, effective AI solution.

These challenges are compounded by the fact that even small changes to a process can have significant consequences, making it difficult to implement and test new systems. As a result, many capable systems go unused, not because they are ineffective, but because they are unable to account for the intricacies of the chemicals industry.

Integrating new capability into existing routines

A key factor in successful enterprise AI adoption is the ability to attach new capability to a routine that already exists. In the chemicals industry, this means integrating AI systems with existing process control and monitoring systems. A shift supervisor, for example, checks the maintenance log and production schedules to identify opportunities for optimisation, and an AI system can be used to analyse this data and provide recommendations for improvement.

By building on existing routines and systems, organisations can reduce the complexity and risk associated with implementing new technology. This approach also allows organisations to leverage the expertise and knowledge of their existing staff, rather than requiring significant retraining or recruitment.

Enabling the right people to ask questions

Most people assume that the primary consideration for enterprise AI adoption is identifying who needs the answer, but this is wrong. In reality, the key to successful adoption is determining who needs to be able to ask the question in the first place. This requires a deep understanding of the organisation's workflows and decision-making processes, as well as the ability to empower staff at all levels to interact with the AI system.

A category manager, for example, may need to compare two suppliers and determine which one is likely to provide the most consistent quality of raw materials. By enabling the category manager to ask the AI system for this information, organisations can ensure that the right data is being used to inform decision-making, and that the AI system is being used to its full potential. For more information on how AsscherAi can support this, visit our website.

Rebuilding trust after the first wrong answer

Even with the best-designed systems, there will inevitably be instances where the AI provides an incorrect answer. When this happens, it is essential to have a clear process in place for rebuilding trust. This requires transparency about how the AI system works, as well as a clear understanding of its limitations and potential biases.

By acknowledging and addressing the first wrong answer, organisations can demonstrate their commitment to using the AI system responsibly and effectively. This, in turn, can help to rebuild trust among staff and stakeholders, and ensure that the AI system is used to its full potential. To discuss how AsscherAi can support your organisation in rebuilding trust, contact us.

When enterprise AI is not the right choice

While enterprise AI has the potential to drive significant improvements in the chemicals industry, there are certain situations where it may not be the right choice. For example, in situations where the process is relatively simple and well-understood, or where the data is incomplete or of poor quality, other solutions may be more effective.

In these cases, organisations should consider alternative approaches, such as traditional process optimisation techniques or more targeted data analysis. By taking a nuanced and informed approach to enterprise AI adoption, organisations can ensure that they are using the right tools for the job, and achieving the best possible outcomes. The decision to adopt enterprise AI should be based on a clear understanding of the organisation's needs and goals, rather than a blanket assumption that AI is always the best solution.