By Phil Molé, MPH

Safety teams are being asked to do something very difficult: Move faster, but don’t make mistakes.

In many ways, as EHS professionals know all too well, this is nothing new. EHS professionals have long reported they need to do a lot with a little and to work efficiently while delivering results, especially when it comes to reducing rates of serious injuries and fatalities (SIFs). But today, the pressures are more intense than ever. Risks are evolving. Regulations are evolving. Data is piling up. And EHS professionals are struggling to not only collect but analyze and use that data to improve safety.

For these reasons, more EHS professionals are curious about AI and some are already using it. Further conversations are happening around agentic AI, but so far, these discussions are not going deep enough to help EHS professionals understand what AI agents can do, and what they can’t, or at least shouldn’t do.

This is why VelocityEHS is kicking off a new two-part blog series on agentic AI. In this first installment, you’ll get a basic primer on agentic AI and on the challenges driving EHS professionals to investigate it.

Background: The Challenges of EHS Management

Let’s be direct about the state of EHS today: Many teams are stretched thin. They’re under-resourced, working across disconnected systems and responsible for more than one team should reasonably carry. That’s no surprise. What continues to build, though, is the weight of it.

Layer on the demands of regulatory compliance, including constant data collection, ongoing reporting, and the real consequences of getting it wrong, and the pressure compounds quickly. Violations and fines are only part of the equation. The day-to-day effort required just to stay compliant can be draining.

Over time, this workload shapes how EHS teams operate. When so much energy goes into just keeping up with the basics, including documentation, follow-ups, and deadlines, there’s less room for proactive work. Safety becomes reactive by default, focused on what just happened, instead of what could happen next.

As stressful as this is, stress isn’t the biggest problem in this situation. It’s the unaddressed hazards and risks, even those with the potential to cause SIFs. The U.S. Bureau of Labor Statistics (BLS) data bears this out, showing that rates of SIFs have fallen significantly less over the last 20+ years compared with rates of total injuries. And when injuries happen, you need to drop everything and do an investigation, recordkeeping and reporting, and follow up, all of which locks you deeper in the reactive safety management cycle.

The Growing Role of AI in EHS

In recent years, AI has emerged as a potential way to help EHS teams break the cycle.

AI is technology that can analyze information and make decisions or predictions in ways that historically required significant time and human thinking. The key to AI’s effectiveness is its ability to not just analyze huge datasets and use insights from the data to provide useful outputs, but to continue learning as more data comes in.

In other words, the algorithm, or internal instructions used by AI to perform tasks, is not fixed, but instead keeps leveling itself up based on patterns it sees as it processes data. As a result, AI can conduct analyses of complex problems and provide feedback not only much faster, but more accurately than humans can.

These benefits explain results VelocityEHS has seen from surveys of EHS professionals over the last year. The overwhelming majority of EHS professionals express positive sentiment about AI, and most are also already using it in their workflows.

Some of the uses of AI reported by EHS professionals include:

  • Data management, including cleaning and summarizing data
  • Generating written content, such as meeting minutes or summary reports
  • Identifying trends in incident data
  • Flagging potential risks earlier

It’s worth noting there is a mixture of uses from general AI capabilities with EHS-specific use cases in this list. For example, content writing is a general AI ability, available in common commercially available tools, including ChatGPT and Microsoft software applications. These kinds of AI functions are useful and can be a major time-saver. Even so, AI capabilities that are tailor-made for EHS offer deeper value because they provide insights that help professionals address common pain points.

For example, specific EHS use cases of AI include:

Incident Management: AI can assess the strength of an incident description and offer suggestions for improvement. From there, it can conduct root cause analysis (RCA) and identify appropriate actions to address each cause.

Job Safety Analysis (JSA): Using AI, EHS professionals can improve the consistency and accuracy of their JSAs. AI can help improve job descriptions, identify hazards associated with tasks, and recommend effective controls to reduce risks.

Ergonomics: With modern AI-enhanced ergonomics software, you and your team can conduct 3-D ergonomics assessments from anywhere using a mobile device. Then, the software will not only identify musculoskeletal disorder (MSD) risks, but also pinpoint root causes of those risks and recommend controls. AI can help teams transition from a focus on just getting assessments done, to using insights from assessments to do things.

Contractor Management: Using contractor management software with AI features, you can auto process contractor documentation such as certificates of insurance (COIs) and OSHA logs, up to 7x faster.

Chemical Management: Using machine learning (ML), modern chemical ingredient indexing services can not only extract information about constituents of your chemical products, but also cross reference them against regulatory lists. From there, you can get a levels of concern (LoC) summary of major listings and concerns for the chemical, such as whether it’s a Toxic Release Inventory (TRI)/Form R reportable chemical, a Hazardous Air Pollutant (HAP), or a chemical with established occupational exposure limits (OELs) to plan your industrial hygiene (IH) chemical sampling program. AI made for EHS is valuable because it helps EHS teams see more and know more.

Understanding Agentic AI

Given the benefits of AI, some EHS professionals are also curious about the potential of agentic AI.

But before you truly understand what agentic AI is, and what EHS professionals should be considering about it, you first need to understand what an agent is. The best way to do that is to think about types of agents already familiar to you.

For example, what do publicity agents do? They take actions on behalf of a client to get them exposure, such as setting up interviews and public book signings, or releasing statements to the press. A writer entrusts a literary agent with her work, so the agent can pitch it to publishers, and line up reviews in publications once the work has a release date. An FBI agent conducts criminal investigations and special operations authorized by the bureau chief, but only after passing all the rigorous background checks and training required to join the FBI.

Strip it down and these examples share three core elements:

  1. An agent can act. It does something. It doesn’t just sit there.
  2. An agent has some goal or objective that guides its actions. An agent that fails to accomplish its goals isn’t doing its job.
  3. An agent acts with authorization and trust. For example, you sign a contract with a literary agent, and the contract stipulates what the agent is authorized to do and the terms of service. An agent who takes your manuscript and palms it off to another client to present as their own work is violating the trust underpinning that arrangement.

Not everyone can become an FBI agent, and even if they do, they don’t have the latitude to do anything they want. They need to meet stringent requirements, and if they make the cut, they can only conduct operations authorized by the Bureau chief.

With that background in mind, we can now better understand what agentic AI is, and why some EHS professionals might be interested in it.

Agentic AI is AI that plans and takes actions to achieve a goal, rather than simply responding to instructions. It still uses algorithms trained on datasets and still improves over time as it assesses new data (at least ideally), but its output emphasizes action.

That’s why agentic AI seems attractive to EHS teams. A safety pro or company leader whose organization is stuck in the reactive safety management cycle would certainly like having a kind of virtual agent at their disposal to help them get things done.

What kinds of things can AI agents do? You’ll see a fuller discussion of use cases, and some of the nuances and special considerations involved, in the second installment of this series, but here’s a quick list for now:

  • Performing auto-fills of data fields
  • Generating summaries of safety meetings or of other safety documents
  • Assessing incident descriptions and recommending improvements
  • Conducting RCA for incidents to identify underlying risks
  • Recommending corrective actions after an incident based on RCA
  • Evaluating job task descriptions in JSAs and suggesting improvements
  • Identifying hazards and recommending effective controls in JSAs
  • Flagging root causes of MSD risks from ergonomics assessments and selecting controls for each identified cause

You’ll notice that there’s considerable overlap between this list, and the earlier list of uses for AI in EHS, more generally. Put another way, a substantial subset of AI uses cases can potentially involve use of agents. Agentic AI represents what many EHS professionals want out of their software: better efficiency, fueled by AI’s ability to act.

Here, though, is where some clarity is needed. Action is important, but in EHS, action carries a lot of weight. With human health and safety on the line, it can’t be otherwise.

Some software vendors stress the autonomy of their agentic AI and there may be times, including for low-risk, low impact functions, where autonomous agents make sense. They improve efficiency without compromising safety program effectiveness, or at worst, introducing new risks. But how much autonomy is appropriate when the stakes are high?

Let’s close this initial discussion with a couple of key takeaways to be revisited and expanded upon in the next installment.

#1: Speed alone doesn’t create value, and it doesn’t ensure EHS management success. In many cases, rushing core safety decisions can increase risk rather than reduce it.

#2: Recall discussion of the commonalities across different kinds of agents. You saw notions of oversight and trust are integrated in the concept of an agent. You need to depend on an agent doing what it’s supposed to do, and you don’t want it to overreach by taking actions you haven’t approved, especially in high-risk situations.

 

In Part Two, we’ll pick up from here with a deeper look at considerations for agentic AI, the reasons why the ability to trust an agent is paramount, and situations where human expertise remains essential. In short, you’ll learn about the types of AI agents worth having.

Looking for More Information?

Stay tuned for the second installment of this blog. In the meantime, check out other AI in EHS resources on demand, including:

Also, make sure you follow our blog for the latest EHS news and insights.

Ready to See VelocityAI in Action?

VelocityAI capabilities, powered by Vēlo, are human-centered and purpose-built for EHS, designed to integrate with core workflows, surface actionable insights, and support safety leaders in driving measurable outcomes. Vēlo is your own always-on, built-in safety assistant, delivering trusted, reliable, in-the-flow guidance inside the software. It empowers safety professionals to act faster, with greater clarity and consistency, while supporting your responsibility to protect people and prevent harm.

There’s never been a better time to check out VelocityAI because Advanced Reporting and Dashboards now can compile data from the AI incident management features, including AI PSIF Insights, AI Description Analyzer, AI Root Cause Identifier, and AI Corrective Action Advisor. They even provide ROI data on addressed PSIF risks.

If you’re ready to explore how AI can elevate your EHS program, VelocityEHS can help you take the next step. Reach out to us today to set up a meeting so you can see our capabilities in action.

Cta Insert Schedule A Live Demo