The Burden of Safety Data Sheets (SDS)

Safety Data Sheets (SDS) are critical for workplace safety, ensuring employees have access to key hazard information about chemicals they work with. However, SDS documents are often lengthy, complex, and difficult for frontline workers to interpret quickly. This is where Safety Instruction Cards (SICs) come in—concise, task-specific summaries that translate SDS information into practical, actionable guidance.

Traditionally, converting SDS into SICs has been a labour-intensive process, often taking safety professionals four hours or more per sheet to manually extract and summarize essential information. With AI, this process is reduced to 15 minutes, where 90% of the time is dedicated to expert quality control rather than tedious data extraction.

From SDS to SICs: How AI Transforms the Process

AI-driven automation revolutionizes the way SDS are analyzed and converted into Safety Instruction Cards:

1. Data Extraction – AI scans and extracts key sections from SDS, such as hazard classifications, exposure risks, and protective measures.
2. Summarization & Structuring – Using natural language processing (NLP), AI organizes the extracted data into a structured, easy-to-read format tailored for SICs.
3. Customization – AI can adapt SICs based on industry-specific needs, regulatory requirements, and company-specific safety standards.
4. Quality Control – Human experts validate the AI-generated SICs to ensure accuracy, compliance, and contextual relevance.

This workflow optimally balances AI efficiency with human oversight, ensuring safety information is both fast-tracked and reliable.

The Impact: More Time for Safety Leadership

By reducing the time required for SDS-to-SIC conversion from four hours to 15 minutes, AI frees up safety professionals to focus on:

• Proactive risk management – instead of being buried in paperwork, safety teams can engage in strategic safety initiatives.
• Training and engagement – frontline workers receive clearer, more digestible safety instructions, improving compliance and hazard awareness.
• Process improvement – EHS leaders can leverage AI for continuous improvement in safety communication and documentation.

Practical Example: Manual to AI-Powered Conversion

Let’s look at a real-world example:

Traditional Approach (4+ Hours)

• Manually read a 30+ page SDS.
• Identify critical safety information.
• Summarize hazards, PPE requirements, first-aid measures, and handling instructions.
• Format and review the document for clarity and compliance.
• Approve and distribute the final Safety Instruction Card.

AI-Optimized Approach (15 Minutes)

• AI scans and extracts relevant information from the SDS in seconds.
• AI structures and summarizes the key points into a pre-formatted SIC.
• Human expert reviews and validates the content (~90% of the time spent).
• Adjustments and approvals are completed within minutes.
• The SIC is distributed to frontline workers immediately.
• Integration with other processes such as industrial hygiene, occupational medicine, emergency response, education, production standards,…

Practical Example

step 1: download the KNIME Analytics Platform 5.4.1 (or later). In version 5.4.1. the integration of ChatGPT O3-mini and DeepSeek AI is in place (download the AI integration: link)

step 2:  download the template for SDS extraction here: link

step 3: adjust the process to your local computer (see community hub for training)

The above shows the process that is used to extract data from a safety datasheet to an Excel, in this case, the risk phrases. Although this is rudimental, it is a good way to start your journey. With the integration of large language models, one can go a step further. SASM has designed an automatic SDS converter that goes beyond this process:

As an example, we’ve taken the BASF SDS Acrylic Acid Glacial (link)

The data is automatically extracted into the template that has been designed. Most of the data doesn’t require any AI, but solely automation. AI is only useful when we need interpretation for the First Aid measures and the Collective Protective Equipment.

In our approach to threshold management, we’ve integrated local legislation with our analytical processes to ensure compliance and safety. This means that country-specific exposure limits are directly compared to the thresholds outlined in the Safety Data Sheet (SDS). If a country doesn’t specify a limit or sets a higher threshold than the SDS, our system automatically applies the strictest standard. This ensures a consistent, high-level safety approach across all locations, prioritizing worker protection regardless of regional variations.

AI in EHS: A Smart Step Forward

The automation of SDS-to-SIC conversion is a clear example of how AI enhances safety management without replacing human expertise. By handling the heavy lifting of data extraction and summarization, AI allows safety professionals to dedicate their expertise where it matters most—ensuring quality and safeguarding frontline workers.

A glimpse into the future

If your organization is still manually converting SDS into SICs, it’s time to explore AI-powered solutions. By adopting automation, you can significantly improve efficiency, accuracy, and safety culture—making workplace safety truly proactive, not reactive.

The future, however, is much closer than one might think. Automatically providing risk assessments from footage like SASM’s Molossus is already working today.

SASM specializes in bringing the future to you. We build applications and provide tools to bring AI to EHS. We streamline your processes and study where it makes sense to automate, use generative AI, and simplify processes. Do you want to leverage AI in your organisation? Contact us at ai@sasm.be