WEBINAR

Your 2025 OSHA Recordkeeping Checklist | February 20th, 2:00pm ET Learn More

×
Request a Demo
Request A Demo
+1.216.765.7100
close

EHSvoice

Dakota Software's Blog for EHS and Sustainability Professionals

AI and EHS Compliance: Benefits and Boundaries

October 29th, 2024 by Dakota Software Staff

AI and EHS Compliance: Benefits and Boundaries

There’s no question that AI is poised to transform Environmental, Health, and Safety (EHS) management in high-risk industries.

Automated environmental compliance monitoring, predictive analytics for safety management, AI-powered permit deconstruction—these applications and more all stand to reshape workplace safety as we know it.

At the same time, the AI gold rush has led to a proliferation of false claims, creating confusion and increasing risk for those who take those claims at face value.

To help distinguish fact from fiction, this article explores what AI applications in EHS compliance can (and can’t) do, and pitfalls to avoid when looking to leverage AI.

Practical AI Applications in EHS Compliance

A recent report by the National Safety Council (NSC) outlines a number of emerging AI applications in EHS compliance that can be integrated with EHS software workflows. NSC notes that the key advantage of these tools is their ability to deliver predictive and prescriptive insights, as opposed to the traditional focus on lagging indicators.

Let’s jump in and take a look at some of the most promising applications and how they stand to impact EHS management.

AI-Powered Permit Deconstruction

Keeping up with permit requirements is a complex and time-consuming job for EHS professionals, and one area where emerging AI tools can help. These tools make permit compliance simpler by:

  • Automatically scanning permits for actionable requirements
  • Summarizing action items from many pages of requirements
  • Incorporating compliance tasks into EHS software tools such as compliance calendars and checklists

Computer Vision for Hazard Detection

Computer vision is one of the most well-developed AI applications for EHS management currently in use.

This technology allows cameras installed in workplaces to automatically analyze images to detect hazards and unsafe behaviors that might not be caught during periodic manual inspections, such as:

  • Near-misses like unsafe forklift driving

  • Poor ergonomic positions

  • Spills, fires and poor housekeeping

  • Workers not wearing personal protective equipment (PPE)

AI Incident Reporting

As a routine task that involves sifting through multiple data sources, EHS incident reporting is an activity where AI can significantly reduce manual labor hours. AI tools for EHS incident reporting help streamline the process by:

  • Analyzing historical data and identifying trends for root cause analysis

  • Flagging safety incidents and risks that management needs to be aware of

  • Generating detailed EHS incident reports based on incident data

AI Risk Analysis

Combining IoT sensor data with predictive analytics, AI can assist with risk analysis and mitigation by:

  • Predicting high-risk situations

  • Calculating real-time risk levels

  • Alerting teams when risk thresholds are exceeded

AI Knowledge Bases

AI-powered help systems now make it easy for software users to get product support without the need to navigate help systems or wait for answers from support teams. One example is Dakota Software’s AI-powered help system, which provides:

  • Step-by-step instructions for tasks like adding users or identifying trends from safety data

  • 24/7 multilingual support for users worldwide

  • Links to source materials so users can learn more about the specific topics they need help with

AI for Predictive Maintenance

In the past, manufacturers have typically performed equipment maintenance on a scheduled basis, or when a machine breaks down. Predictive analytics for safety management represents a real game-changer here, pulling data from multiple sources to:

  • Predict when machine failures are most likely to occur

  • Provide proactive recommendations for maintenance actions

  • Automatically alert safety managers when machine performance has degraded

AI’s Limitations in Complex Regulatory Environments

While the number of AI applications in EHS compliance is growing, companies should take caution, as AI still displays significant limitations in several key areas. Regulatory analysis is a prime example, and one where companies could find themselves in hot water if they rely on AI to shortcut the process.

There are several issues at stake here:

  • AI challenges in regulatory interpretation

  • Issues with datasets used in AI models

  • Deficiencies in cross-jurisdictional understanding

  • Inaccuracy in AI-generated results

Below, we discuss each of these in more detail, and why companies should use caution before utilizing AI to replace human expertise in regulatory compliance.

AI Challenges in Regulatory Interpretation

As it stands currently, AI doesn’t actually possess the intelligence to understand the nuances of legal language, nor does it know how to apply standards that are used across different industries.

Many OSHA requirements, for example, use ambiguous terms like “reasonable,” a situation where AI doesn’t always have the context to provide reliable regulatory guidance. Compliance with standards used across different industries—for instance, waste management—also requires specialized expertise in how the requirements apply to a given organization.

Data Challenges

AI learns on large datasets, so its capabilities are only as robust as the data fed into the system. This creates several problems:

  • Data biases: If data is biased or otherwise inadequate, the results are likely to be inaccurate and biased as well. As the old saying goes—“garbage in, garbage out.”

  • Overreliance on historical data: AI struggles to recognize emerging and unexpected risks for which limited data exists. Examples here would be exposure risks to nanomaterials, or the environmental impact on companies from climate change.

  • Data workload: Collecting and cleaning large amounts of data is very labor-intensive and requires significant resources to do right.

Deficiencies in Cross-Jurisdictional Understanding

AI isn’t yet sophisticated enough to interpret complex regulatory frameworks, which can present problems for companies with a global or multi-state presence.

Conflicting requirements among different states or countries, for instance, could lead to incorrect application of environmental laws from one jurisdiction to another, creating hidden compliance risks.

Inaccuracy in AI-Generated Results

Research has shown that AI models often deliver wildly inaccurate responses, even inventing “facts” out of whole cloth across a wide variety of situations.

One large-scale study, for example, showed that AI large language models (LLMs) frequently hallucinate even when performing simple legal analysis tasks. This study showed that ChatGPT and Llama 2 returned erroneous results from 68% to 88% of the time, respectively.

Combining Human Intelligence with AI for Safe and Compliant Workplaces

AI is getting smarter every day, but claims that it can replace human expertise entirely have been wildly overblown. Given the risks discussed here, responsible AI use requires strong guardrails that address the inherent limitations of AI with real-life experience and human intelligence.

These guardrails include:

  • Human oversight: Humans still have a vital role to play in validating and fact-checking AI findings to ensure compliance and safety (not to mention a company’s reputation).

  • Collaborative decision-making: Combining the power of AI data processing with human expertise in interpreting legal nuances represents a balanced, responsible approach to AI use in EHS compliance.

  • Adaptive learning and feedback loops: Feeding real-world human expertise and insights back into AI systems promotes continuous improvement and makes AI smarter over time.

Maximizing AI’s Potential While Avoiding the Pitfalls

One of the biggest advantages of AI is its ability to automate routine tasks, improving efficiency so EHS professionals can focus more on strategic priorities.

Even so, AI is still a nascent technology, and as such demands a cautious approach. Companies that ignore the risks—especially as they relate to regulatory compliance—may find themselves faced with serious legal headaches. It’s imperative to recognize that workplace safety and compliance still require human expertise, both to protect employees and businesses as a whole.

Check out our demo video library

Be Part of the Solution

Sign up for the Dakota EHS e-Newsletter for monthly updates from our regulatory and industry experts.

subscribe