Artificial Intelligence in the Chemical and Coatings Industries: Practical Applications, Benefits, and Risks

Over the last few years, the use of artificial intelligence (AI) in the chemical and coatings industries has become quite a polarizing topic. The launch of Open AI’s ChatGPT in 2022, with its easy-to-use, large language model (LLM) platform, gave everyone access to the vast capabilities of AI. While the first versions of these LLM platforms certainly had their issues and became the butt of many jokes, numerous people found practical value in them and began to use the tools daily for tasks such as online research, writing and editing articles, or data analysis.

Fast-forward to 2025, and the advancement in capabilities and breadth of use cases for AI makes it clear that this technology will significantly change the way companies do business. However, many people still dig their heels in and refuse to embrace that change. This crowd raises legitimate concerns, a few of which will be discussed further in this article. But first, let’s look at how AI is being implemented by companies to work more efficiently or gain a competitive advantage.

AI use cases exist across most (if not all) industries, and the chemical and coatings industries are no exception. Even when not used on a company-sanctioned basis, many employees across these industries are using LLM platforms as a “virtual assistant” at various levels. With that said, the focus of this article is on the use of AI strategically at companies to simplify internal processes and/or gain a competitive edge. While potential use cases are abundant, two that are often mentioned are in the areas of supply chain and logistics, and research and development.

While Amazon does not qualify as a chemical company, the complexity of its supply chain and logistics is well-known and parallels can be drawn from the use of AI within this forward-thinking company. In a recent message from Amazon CEO Andy Jassy to his employees, he describes AI as a once-in-a-lifetime technology development that has the ability to “completely change what’s possible for customers and businesses.”(1) In addition to various use cases that are specific to an online retailer, Jassy discusses the use of AI across Amazon’s internal operations, which includes improved inventory placement and demand forecasting, both of which are critical for improving cost structure and delivery speed. It isn’t a major stretch to see how these improvements could benefit companies in the chemical industry.

Along similar lines, companies such as Valdera and Arkestro offer AI-powered software platforms that enable chemical companies to improve raw material sourcing and the procurement process. They boast that their systems typically save their clients 10-20% in overall costs by improving capabilities in supplier and material sourcing, negotiation, and streamlining the materials procurement process.(2,3,4) These companies leverage AI, data science, and machine learning to identify and qualify new suppliers based on their ability to meet specifications, rather than just an SKU or CAS number. Additionally, they use historic pricing patterns to anticipate market changes and assist in the negotiations process.

The companies report an impressive list of chemical organizations in their client lists, collectively including BASF, Dow, Evonik, Clariant, and many other well-known names in the chemicals and coatings industries. Supply chain and logistics clearly stand out as areas where AI can deliver near-term benefits.

Getting a bit deeper into the predictive capabilities, many companies are beginning to see the benefits of using AI in their product development work — an area where scientists have typically held a high level of skepticism. The skepticism isn’t around the ability of AI to analyze a data set and determine the next set of experiments or a possible solution; after all, design of experiment (DOE) has been used for many years in this manner, and there is a clear path for AI to act as a natural extension to that process. Instead, the concern is that under certain conditions,  AI can potentially “make things up” or “hallucinate,” resulting in inaccuracies or baseless recommendations for follow-up experiments.

Another concern is the complexity of the chemical and coatings formulation world — without knowing the constraints, AI may recommend combinations of materials that are not physically possible. These issues will be discussed in more detail, but first let’s examine how companies like Syensqo and Applied Molecules are leveraging AI in their product development work.

In a recent article in Chemical Processing Magazine, Jonathan Katz writes that Syensqo launched an AI-focused division about two years ago focused on implementing AI initiatives across the organization and is already reaping the benefits.(5) As an example, Syensqo’s technical team successfully utilized AI to analyze data on 4 million molecules to identify a single polymer that met their desired properties for a new product. Their team of data scientists, modeling experts, and experimental chemists were able to leverage internal and external databases to arrive at a solution in 18 months; they estimate the discovery would have taken five to six years using traditional methods.

Applied Molecules, a company that develops and offers energy-curable technologies for the CASE industry, has also been leveraging AI and machine learning to accelerate their product development.(6) They use a platform called Albert Breakthrough, from AlbertInvent, that reportedly excels at identifying non-obvious correlations within complex scientific data. This “intelligent assistant” enables them to rapidly sift through large databases of chemical and material properties to identify optimal pathways and predict performance.

Paul Snowwhite, CEO of Applied Molecules, recently stated, “Thanks to Albert Breakthrough, projects that would traditionally take three months now take as little as two days.”(7) At a recent coatings conference, Snowwhite said that “AI is like an Ironman suit. It doesn’t replace scientists, but it gives them the tools to do science better — and as they feed in more data they keep getting better results.” Since Applied Molecules desires to remain at the forefront of technology in their markets, they believe it is critical to “integrate cutting-edge lab technology into the heart of their processes.”

I have spoken with many leaders in the coatings industry, and while there are many potential AI use cases for coating manufacturers, most are hesitant to give specific examples of how the technology is being implemented within their organizations. What is clear from these discussions is that all the major coating manufacturers are at the very least experimenting with AI in their processes, including supply chain and logistics, R&D, coating formulation, color matching, and optimizing coating application processes.

What is unclear is how far along these companies are in their implementation efforts, as they tend to keep details a bit closer to the vest regarding information that they consider a competitive advantage or that may affect end-user perceptions. Nevertheless, one can conclude that coating manufacturers, much like raw material suppliers, are utilizing AI and will continue to adopt the technology on a broader scale to capture its benefits in their operations.

With all the upside potential of AI, it may seem that every company would be pushing forward to implement this technology in their business operations, but that is not the case. In fact, many are highly skeptical of the technology and believe the benefits are not worth the downside risks. What exactly is holding these people back from getting on board and reaping the benefits? In addition to the investment costs, which can be significant depending on the specific initiative, a few reasons come up repeatedly that could have clear, negative impacts if not managed properly. Among them are security of information, output accuracy, and potential overreach.

Chemical and coatings companies are a highly secretive bunch, and rightly so, given that their proprietary information plays a major role in their ability to compete. Additionally, most people’s familiarity with AI is limited to the free LLM AI platforms mentioned earlier, and it is well understood that information input into these platforms is far from confidential. In fact, that information is often used to “train” the model and can be used by the AI as output to other users, both directly and indirectly.

Based on this knowledge, “trust me, your information is secure” is not going to be enough to convince people in the chemical and coatings industries that their information won’t be shared or used in a way that is not desired. Even if the vendor is deemed trustworthy, people have seen enough data breaches to make them think twice before trusting a third party with their valuable information. Security of information remains one of the main concerns regarding the use of AI technology in the chemical and coatings industries.

Another common concern involves the accuracy of the information output from AI. If not managed properly, AI models have been known to generate and confidently present information that is not grounded in verified data, evidence, or reality. Many have experienced this issue with commonly used LLM models, but the impact of these “hallucinations” is much more significant when companies are using this information to make business decisions.

While several factors influence AI hallucinations, the issue stems from the AI model generating responses to questions that do not have a clear answer based on its available information. When the necessary data does not exist, the AI’s tendency will be to “fill in the blanks” based on patterns within the data set, causing it to draw conclusions that may be incorrect. The potential for an AI model to hallucinate creates a significant barrier to adoption, because if the AI model lacks  credibility, it will not be trusted by the scientific community or with tasks that are business critical.

The issue with AI overreach isn’t the science-fiction scenarios that are the subject matter of many books and movies. Instead, the real concern is that the AI model’s role may expand faster than governance — it may start deciding rather than advising in the business environment. That is an important distinction, because the consequences of those decisions sit with the company, not with the AI model, and thus human oversight is critical.

For example, an AI pilot program that was intended to help employees make product recommendations may evolve into making customer-facing responses, completely bypassing the human approval process. The concern here isn’t that the AI model becomes too powerful, but that it becomes powerful without accountability.

When researching the use of AI within the chemical and coatings industries, there are a few common themes regarding how companies manage the potential risks:

• Internal/private deployment—AI models for company usage should be behind firewalls or contained within a closed environment to avoid information leakage and prevent interaction with the external environment. This is of critical importance to protect IP and data confidentiality, as well as to control the information that the AI uses in its decision-making processes.
• Domain constraints and AI rules—The information that the AI model has access to and the range of activities that it is allowed to undertake must be constrained to reduce the potential for hallucinations and overreach. Access to information should be limited to data that is validated and approved for use. Guardrails should be used to limit the AI to activities that are within its scope-of-use, including the type of prompts it is allowed to answer.
• Permissions and auditing—A tiered access system should be used to restrict data and AI access by employee role to avoid cross-project or inter-company information leakage. Additionally, prompts and logs should be recorded and auditable, creating audit trails that can be used in the event of data leaks or misuse. Outputs should be spot checked for accuracy, and the AI model should be modified as necessary.
• Pilot deployment and oversight—Start with a limited scope on non-critical business activities and gradually scale while evaluating risk, performance, and governance.

It is highly recommended to engage a third party that is current on the latest AI technologies early in the process, as they will play a critical role in laying out a roadmap for the specific use case. In addition to building the roadmap, the selected partner should also be able to recommend specific types of software or develop a custom AI model as needed.