Artificial Intelligence in chemicals: Balancing innovation and practical challenges

AI has been around for a long time, and companies everywhere are now asking how they might harness it to help them.

Jorge Gonzalez

by Jorge Gonzalez

Artificial intelligence (AI) has been around for a long time, but has only really hit the mainstream relatively recently. Tools like ChatGPT have brought AI into the spotlight, and companies everywhere are now asking how they might harness it to help them.  

At Enhesa, we’ve been using AI since 2018, providing our customers with innovative benefits for many years now. However, when customers want to use AI beyond this, what do they need to know?

From broad principles to specifics

In a recent webinar on the use of AI in chemicals compliance, presenters shared some helpful pointers. 

Each presenter offered different examples of how AI was being used within their work and company — but broadly speaking, the benefits and challenges are very similar. Notably, all the biggest benefits also come with a ‘shadow’ challenge that’s effectively a barrier to achieving that benefit. 

A key benefit of AI was accelerating development time, reducing both waste and time to market. We know that time is money, so this is important.  

The flip side is that AI models take a long time to train. Hiring people who can build suitable models is just the starting point — businesses also require people and data to train them.  

Balancing inputs and outputs

Data availability is one of the key reasons AI has suddenly become mainstream. With significantly more data now available, from connected devices and on the internet, it’s become increasingly feasible to train AI-based models.  

However, as our own AI experts in Enhesa explain, it’s clear that the training process can also be costly because of the required expertise. Our AI models are trained by our legal experts at Enhesa, and other companies find that they need a similar level of input. Training an AI-based model is certainly not a job for an intern.  

The issue for any company thinking about investing in AI is, therefore, to identify the right balance between inputs and outputs. Webinar speakers indicated that many companies have a lot to gain from the right investment in AI-based models. However, you need to be able to invest time and data to get maximum value — and you need to be sure that it’s worth it.  

Once you have your model, you could experience huge benefits from increased productivity and avoiding downtime, to improved accuracy or better precision. 

From subjectivity to objectivity?

The contrast between human subjectivity, and the (potential) objectivity of an AI-based algorithm was another point to emerge from the webinar. I say ‘potential’ because it’s important to understand that any AI model is only as unbiased as the data used to train it. If biased data is used for training, the results from the model will also be biased — consider some early voice recognition models that only recognized and responded to men’s voices. However, with sufficient — and sufficiently unbiased — data to train your model, your assessments should become less subjective. 

The differences between people and algorithms are further challenges. As humans, we keep learning throughout our lives—and certainly throughout our working lives. We can develop and change over many years. We can change jobs, learn new skills, and take on new responsibilities. AI-based algorithms don’t have this capacity. Supported by machine learning capabilities, they’ll continue to learn to some extent over their lifetime. However, that lifetime is likely one to three years, and then you will have to start again — except if models and infrastructure can often be incrementally updated to adapt to new data and requirements — and that will be costly. The message I take from this is that you should be ready to start getting value from your AI system immediately, or you’ll run out of time before you’ve recouped your investment.   

What’s more, AI models have no capacity to develop new skills in different areas. As yet, there are no algorithms that fall into the category of ‘general AI’. All the current AI models are ‘specific AI’. That means they can do one job extremely well, which could be writing, generating synthetic data, or scanning regulatory databases like Enhesa Fusion. And before anyone says ‘ChatGPT’, that’s specific too: it simply finds the most likely next word in any sequence. 

Complicated — but not complex

The last point from the webinar is also linked to the scope of AI and its capabilities. All the presenters emphasized that AI models are able to help us to solve complicated problems. They can, for example, scan thousands of registry databases and check for any changes to consider. They can also do this at speed, in a way that simply wouldn’t be possible for human operators.  

However, they can’t yet solve complex problems. World peace and climate change will have to wait — and quite possibly for many years yet. Indeed, our AI experts at Enhesa note drily that AI models use so much computing power that they could be a significant contributor to your carbon footprint. In other words, these models can make climate change worse unless migration efforts are made.  

The bottom line for me is that you need to be very clear that you need AI — that is, that nothing else can do that specific job for you. It’s no good deciding that you’d like AI, without knowing how you’ll use it. You also need to be confident that you have both the expertise and the data to train your model. This is a significant investment of time, money and computing power — and you need to be ready for it.  

Related AI content

Review our latest content on AI to learn more about how this evolving technology is powering sustainability and compliance across industries.

AI resources
Regulatory content and sustainability intelligence

Using in-house legal expertise to enhance AI

Why do companies need an in-house legal expert in their AI teams?

Regulatory content and sustainability intelligence

Enhancing regulatory content with AI metadata

Unlocking content discovery with metadata, auto-classification, and chemical entity recognition.

Regulatory content and sustainability intelligence

Leveraging AI technology for compliance

Expert insight from the Chief AI Officer at Enhesa, Director of Sustainable Finance at LSEG, and a leading GRC analyst on the future of AI in sustainability and compliance.

Regulatory content and sustainability intelligence

AI-driven precision for compliance requirements

See how Enhesa’s in-house AI team has tailor-made specialist tools that improve regulatory understanding and compliance — from legal text to site inspection.