• Written by Muhammed Abdulla NC, CEO

A common objection to AI in chemicals is that it will replace human workers, but this concern is misplaced, as AI has actually landed in the sector's office functions, where its impact is mostly on tasks that produce text, such as reporting and analysis, rather than on physical work, according to the US Census Bureau's Business Trends and Outlook Survey, which found that 12.4% of US manufacturers, the Census sector that includes chemical producers, reported using AI in the previous two weeks. This is because chemicals is an industry with continuous processes that have tight yield and safety envelopes, making it challenging for AI to directly impact physical operations.

Where AI has actually landed in chemicals so far

AI has mostly been applied to functions whose output is text, such as sales and marketing, where 14.3% of US businesses used AI in the previous six months, according to the US Census Bureau's Business Trends and Outlook Survey. In chemicals, this means that AI is being used to analyze data and generate reports, rather than to control physical processes. For example, a category manager might use AI to analyze sales data and generate reports on market trends.

This is not to say that AI has no potential to impact physical operations in chemicals, but rather that its impact so far has been largely limited to office functions. As the technology continues to evolve, we may see more direct applications of AI in physical operations, but for now, its impact is mostly on tasks that produce text.

Yield drift

One area where AI has had a significant impact in chemicals is in addressing yield drift, which occurs when the output of a process deviates from its expected value. By allowing people to ask the data in systems such as LIMS, batch control, ERP, and environmental health and safety systems direct questions, rather than waiting for a report, AI can help identify the causes of yield drift and enable corrective action to be taken. For example, a shift supervisor might use AI to ask the LIMS system about the current yield of a particular process, and receive an answer in real-time.

This can be particularly useful in chemicals, where yield drift can have significant economic and safety implications. By enabling people to quickly identify and address the causes of yield drift, AI can help to minimize its impact and ensure that processes are operating within their expected envelopes.

Feedstock variability

Another area where AI has had an impact in chemicals is in addressing feedstock variability, which occurs when the characteristics of the raw materials used in a process vary. Currently, decisions about how to adjust processes to account for feedstock variability often wait on a person stitching together data from two systems, such as the ERP system and the batch control system. AI can help to automate this process, by providing answers to questions about feedstock variability in real-time.

For example, a production planner might use AI to ask the ERP system about the current inventory of a particular raw material, and receive an answer that takes into account the latest data from the batch control system. This can help to ensure that processes are adjusted in a timely and effective manner, minimizing the impact of feedstock variability.

Plant-to-plant recipe differences

One area where AI has not yet had a significant impact in chemicals is in addressing plant-to-plant recipe differences, which occur when the same process is operated at multiple plants, but with different recipes. This is because recipe differences can be complex and nuanced, and require a deep understanding of the underlying chemistry and physics of the process. While AI can help to analyze data and identify patterns, it is not yet capable of fully capturing the complexity of plant-to-plant recipe differences.

Furthermore, in order for AI to be effective in addressing plant-to-plant recipe differences, a number of underlying systems and processes must be in place, such as standardized data collection and management, and robust analytics and modeling capabilities. Without these underlying systems and processes, AI is unlikely to be able to provide meaningful insights or recommendations.

Why most adopters have not changed anything underneath

Despite the potential of AI to transform chemicals, many adopters have not changed anything underneath, according to the US Census Bureau's Business Trends and Outlook Survey, which found that 64.3% of US businesses that used AI made no change of any kind to adopt it. This is because AI can often be applied to existing systems and processes, without requiring significant changes to the underlying infrastructure.

However, this also means that the benefits of AI are often limited to incremental improvements, rather than fundamental transformations. In order to realize the full potential of AI, chemicals companies must be willing to make more fundamental changes to their systems and processes, such as changing how they collect or manage data, which only 6.8% of US businesses that used AI did, according to the US Census Bureau's Business Trends and Outlook Survey.

What a sensible first quarter looks like for a chemicals business

A sensible first quarter for a chemicals business looking to adopt AI might involve focusing on a single question, such as "what is the current yield of our process?", and a single system, such as the LIMS system. By starting small and focusing on a specific use case, companies can begin to build their capabilities and expertise, and lay the foundation for more widespread adoption of AI.

It's also important to involve a single team, such as the production planning team, in the adoption of AI, in order to ensure that the technology is properly integrated into existing workflows and processes. For more information on how to get started with AI, contact us or visit our production page.

What this does not do

It's also important to be clear about what AI does not do, and when it is the wrong choice. AI is not a replacement for human judgment and expertise, and it is not a solution to all problems. In some cases, other technologies, such as IT investments, may be more appropriate. Furthermore, AI is not yet widely used in production of goods, with only 1.7% of US businesses using AI in this area, according to the US Census Bureau's Business Trends and Outlook Survey, and 2.2% of US manufacturers, which is a lower percentage than the 14.3% of US businesses that used AI in sales and marketing.