Orchestrating next-generation safety management
Introduction
Safety management systems (SMS) are frameworks for managing occupational health and safety (OH&S) risks with the goal of providing safe and healthy workplaces. Standards such as ISO 45001:2018 provide requirements and guidance for establishing, implementing, maintaining, and continually improving an OH&S management system (ISO, 2018).
However, despite the widespread adoption of SMS, major incidents like the Exxon Valdez, Texas City, Deepwater Horizon, and Piper Alpha disasters have still occurred, indicating that merely having an SMS in place does not guarantee the prevention of harm. This suggests inherent flaws in how these systems are developed, implemented, and maintained.
The challenge in establishing a robust safety management system is largely due to the difficulty in developing and maintaining effective policies and procedures. Policies are frequently written without sufficient worker involvement, leading to a disconnect between "work as imagined" and "work as done". Furthermore, vague or ineffective guidance causes errors such as confusing information or difficult execution. Finally, workers are often challenged to find the right procedure given the sheer number of SOPs that many companies have.
Hidden crisis: Thousands of ineffective procedures
Here's a reality most organizations won't admit publicly: they're sitting on thousands of procedures that are essential for compliance, safety, quality, and productivity. Yet many of these critical documents are ineffective, outdated, or worse, they actually create the conditions for error.
Companies know this. Safety professionals lose sleep over it.
But who has the bandwidth to review, let alone fix, thousands of procedures? You'd need an army of technical writers working full-time. That's been the uncomfortable truth, until now.
AI as a catalyst for enhanced safety documentation and ISO 45001 compliance
By leveraging Artificial Intelligence (AI), organizations can make a strong case for building a significantly better Occupational Health and Safety management system that meets rigorous standards like ISO 45001.
A well-structured OH&S management system, such as ISO 45001:2018, provides a framework for managing OH&S risks and opportunities, aiming to prevent work-related injury and ill health and provide safe and healthy workplaces. This standard emphasizes proportionate management of identified hazards and OH&S risks. Compliance with ISO 45001 involves establishing, implementing, maintaining, and continually improving an OH&S management system (ISO, 2018).
AI can significantly streamline and enhance this process, particularly in the realm of documented information.
Streamlining policy and procedure development
AI tools can revolutionize the development and review of safety documentation. By specifying the required content and structure, including policy titles, purpose, roles and responsibilities, definitions, detailed instructions, risk controls, exception management, and training requirements, AI ensures consistency, clarity, and completeness, making documents easier to understand and implement. This directly supports the ISO 45001 requirement for documented information necessary for the OH&S management system's effectiveness (ISO, 2018).
Accelerating ISO 45001 integration
AI can assist in developing an ISO 45001 compliant management system. For instance, by feeding regulatory standards like OSHA, ANSI, and ISO into an AI tool, it can generate draft policies that cover most compliance elements, significantly reducing the time required for policy writing. This helps organizations fulfill their legal and other requirements, a key intended outcome of an OH&S management system. Here's a practical example: Upload your policy template and regulatory requirements into NotebookLM, and ask the AI to generate new policies that automatically follow your established format and structure. One manufacturing client saved 15 days of policy development time using this approach, allowing their team to focus on worker consultation and risk assessment rather than document formatting.
Enhancing information accuracy and reliability through Retrieval Augmented Generation (RAG)
A common limitation of AI is "hallucinating" or fabricating information. However, by uploading relevant safety documentation, standards, and procedures into AI tools, organizations can create a reliable knowledge base for consultation using Retrieval Augmented Generation (RAG). This process anchors AI responses in actual source documentation, reducing hallucinations and providing numbered links to verify original sources, thus ensuring greater transparency and accuracy in safety documentation. This directly supports the need for OH&S information to be consistent and reliable, as required for communication processes (ISO, 2018).
Integrating real-world worker input
Effective safety policies require a thorough understanding of actual tasks being performed, addressing the gap between "work as imagined" and "work as done". AI can bridge this gap by synthesizing real-world input. For example, the transcription of a recorded conversation detailing a manufacturing process can be fed into AI alongside existing company procedures. The AI can then produce a well-structured first draft, aligning it with company documentation and ensuring procedural clarity and usability. This directly supports the ISO 45001 emphasis on consultation and participation of workers in the development and implementation of the OH&S management system (ISO, 2018).
This becomes even more powerful when you instruct AI to write procedures at an appropriate reading level using "operator speak" rather than technical jargon. Complex engineering terminology can become a barrier for workers, and AI excels at translating technical requirements into clear, actionable steps that workers can quickly understand and execute under real working conditions.
Evaluating and refining documentation
AI can analyze policy and procedure drafts against best practices and regulatory requirements, identifying strengths and suggesting improvements. It can pinpoint common error traps such as field decisions lacking clear criteria, vague information, conflicting instructions, or multiple or hidden actions. This capability aids in achieving continual improvement of the OH&S management system, a core commitment of the OH&S policy.
During a recent SMS project, we fed draft policies back into NotebookLM alongside ISO 45001 for gap analysis. The AI consistently identified missing elements particularly worker consultation requirements that experienced safety professionals had overlooked. This systematic review, which typically requires multiple expert reviewers and revision cycles, was completed in minutes while delivering stronger, more compliant policies.
Game changer: AI-powered procedure enhancement at scale
This is where AI becomes truly transformative. By training AI models on proven frameworks like Fisher Improvement Technologies' (FIT) Top 5 Procedure Error Traps (Fisher Improvement Technologies, 2024), organizations can now systematically identify and eliminate the most common causes of procedural failures at unprecedented scale.
FIT's five critical error traps
According to Fisher Improvement Technologies' proven methodology (Fisher Improvement Technologies, 2024), the most common procedure error traps include:
- Field Decisions: The procedure user must make a decision with little or no guidance for making that decision.
- Difficulty: Mental or physical challenges that make tasks hard to accomplish.
- Vague or Misleading Information: Abstract or undefined terms.
- Conflicting Information: Instructions that conflict with normal expectations.
- Multiple or Embedded Actions: Several discrete actions or hidden steps within notes or warnings.
AI-Powered transformation process
- Upload or scan procedures.
- Identify every instance of these error traps.
- Red-line problematic sections with annotations.
- Suggest evidence-based modifications.
- Wait for human approval.
- Generate revised procedures.
- Scale across the organization.
Indispensable role of human oversight
Despite these powerful capabilities, AI's limitations make human oversight essential. AI can still fabricate information, miss workplace nuances, or over-generalize content. Therefore:
- Check for regulatory compliance.
- Refine content with human expertise.
- Validate with workers.
The goal is to leverage AI for efficiency and consistency while combining it with human expertise to ensure clarity, precision, compliance, and usability (ISO, 2018). Think of AI as hiring an exceptionally capable intern who is tireless and has perfect recall of every safety standard. Like any intern, it produces impressive work but needs experienced oversight. AI might miss contextual nuances or make logical leaps that don't align with operational reality. The magic happens when you combine AI's processing power with human expertise and real-world understanding.
From liability to asset: The bottom line
Organizations have accepted ineffective procedures as an inevitable cost. But with AI-enhanced optimization, systemic weaknesses can be addressed, turning procedures into operational assets.
By combining the specialized insights of industry-proven human performance frameworks (such as FIT's methodology) and broader best practices, including robust regulatory compliance strategies, with AI's processing power, we can remove systemic conditions that lead to incidents, inefficiencies, and compliance gaps.
AI transforms SMS development by serving as an intelligent thought partner that actively challenges assumptions and suggests enhancements. While developing confined space procedures, AI might recommend integrating emergency response protocols or cross-referencing lockout/tagout procedures, creating logical connections across the entire system. This results in fundamentally better safety systems that address risks more comprehensively than traditional approaches.
Conclusion
AI is transforming how safety policies and procedures are developed. With thoughtful use and consistent human oversight, organizations can overcome traditional SMS documentation challenges, creating safer workplaces.
References
- Fisher Improvement Technologies. (2024). Top 5 Procedure Error Trap Assessment Guide. Concord, NC.
- International Organization for Standardization. (2018). ISO 45001:2018 Occupational health and safety management systems Requirements with guidance for use. Geneva, Switzerland: ISO.
