by Phil Molé, MPH

Artificial intelligence (AI) is everywhere right now. You’re reading about it in headlines, hearing about it in conference rooms, and perhaps you’re even seeing it in product roadmaps.

As an EHS professional, you’re probably also hearing a lot about “AI and EHS” or “AI for EHS.” The “buzz” has all come so quickly that it can be hard to tell what’s beneath all of it. Is the talk about AI for EHS just hype, or are there real use cases that help EHS professionals improve workplace safety?

There are even more questions queued up behind that one. Here are a few of them:

  • Can we trust AI?
  • Will AI replace human expertise?
  • Is AI secure?
  • Is AI only for massive enterprises with massive budgets?

There’s a nuanced conversation to be had here, and for EHS pros to be able to take advantage of the benefits of AI, they need to be able to separate myths from reality.

In this two-part series, we’ll do exactly that. By the end of it, you’ll have the clarity you need to understand why AI represents a generational opportunity to break out of old, reactive safety management approaches and start building resilience.

Let’s start with six of the most common myths about AI in EHS.

Myth 1: AI Replaces Human Expertise

This is the loudest myth, and the one that probably causes the most resistance to using AI. One reason for that is that it embodies one of the deepest fears we have about whether we are needed and valued, especially as technology continues to advance.

Another reason is that it’s not entirely a myth. That’s why this is a good place to start our discussion.

Look, we need to be real, here. There have already been many examples in history of businesses replacing human labor with technology, as in the reduction of the manufacturing labor pool due to automation in the 1980s. Recent news pieces have already documented staff reductions as employers have replaced human labor with AI, and economic advisers are warning that labor markets have yet to absorb all of the shocks coming from AI adoption.

But given all of that, a couple of conclusions follow:

1) AI upskilling is one of the better ways to avoid job displacement. Many stakeholders are reaching that conclusion, which is why, as one example, Singapore’s Ministry of Digital Development and Information recently announced that the National AI Impact Programme seeks to train 100,000 Singapore workers to be “fluent” in applying AI in their jobs, to reduce risks of job loss. The point here is that if it’s true that use of AI will increase job insecurity, running from AI isn’t the answer. Those who learn how to apply AI to their jobs will be in a better position than those who haven’t.

2) AI can be a great supplement to human EHS expertise. On some levels, the question about whether AI will displace human expertise presents a false dichotomy, overlooking the potential for human experts to work with AI to better leverage their expertise. The reality is that in EHS, AI works best as an accelerator, not a substitute. EHS-specific use cases for AI, built by human experts using real EHS data for use by real EHS professionals, can make a real difference in improving safety outcomes by surfacing actionable insights into the most serious workplace risks.

Putting these two points together, the real reason that AI upskilling matters is that it can give you the support needed to improve your ability to meet your core responsibility of getting people home safely every day.

Observations Secondary Hero Mobile 750x838 Gettyimages 856736866

It’s worth pointing out, too, that the best AI-enhanced EHS software still depends on human EHS professionals to decide whether or when to accept its output. For instance, it will present a curated list of applicable root causes for a specific incident based on the incident description, but you as the EHS professional still can decide if you’re going to use those suggestions.

The real power of AI is in reducing the noise so experts can focus on what matters. It gives you faster access to insights, so you can apply your judgment where it counts most.

Myth 2: AI Is All the Same

You often here about “AI” in news stories as if it’s all one thing, like some kind of giant computerized super-brain from an old Isaac Asimov short story.

But the reality is that there are countless AI models out there created by different companies, and not all AI is created equal. The value of AI depends on how it’s built, trained, and applied, and on who built and trained it.

An AI model is a mathematical algorithm trained on datasets to perform a task, often involving analysis of user inputs to provide specific outputs. In other words, it’s flexible “intelligence” that processes information and provides insights faster than humans would be able to do in the same circumstances, which is what makes it theoretically valuable. But of course, the design and training of the model is completely central to its ability to deliver that value.

Commercially available AI models designed for general productivity (e.g., common generative AI tools or AI applications within office software suites) is fundamentally different from AI that is purpose-built for EHS. Context matters. Data matters. Domain expertise matters.

An AI tool trained on generic data won’t, e.g., be able to accurately select the most appropriate corrective actions for specific incidents, given their root causes. Generic AI models also can’t pinpoint the subset of incidents that contain potential for severe injuries and fatalities (PSIF) risks, especially when clues to the existence of those risks might be lurking in the details of less severe incidents, like near misses.

But purpose-built AI for EHS can do all those things, because it’s built by people who know all the logic and special considerations to do it right, and they train the models on real-world EHS datasets. Make sure you’re not undercutting the potential value of AI to your EHS management needs with the “it’s all the same, anyway” mindset. If you do start evaluating vendors and software platforms, ask for a look “under the hood.” If you see generic tools bolted on, say “thanks, but no thanks” and keep looking until you find AI-enhanced EHS software that’s right for you.

Myth 3: AI Can’t Understand Safety Context

EHS professionals know better than anyone that safety is nuanced. It’s layered. It’s often site-specific and industry specific.

So, it’s understandable that they might assume that AI can’t possibly grasp that complexity.

The reality is that AI can analyze safety context, but only when it’s trained and structured correctly.

Think of this is a specific refinement of the previous “myth,” that all AI is created equal.

When you have modern, purpose-built AI software for EHS applications, the model can understand the context because it’s been trained on the context. That’s how it can accurately identify root causes and associated corrective actions, or flag PSIF risks.

In fact, the best AI software for EHS not only understands context but also improves context. For example, consider incident investigations. A common failure is that whoever is conducting the investigation often leaves out crucial pieces of information, so the incident description lacks the level of detail needed to support good root cause analysis and selection of corrective actions. However, AI-powered description analyzers can assess the strength of your existing description, and offer targeted improvement suggestions, so you’ll be able to use your incident description to identify and control risks.

With EHS management, context is key, and the right AI-powered EHS software can enrich context and empower you to apply it to the improvement of safety performance.

Myth 4: AI Decisions Are a “Black Box,” So We Can’t Trust Them

Trust is everything in EHS. If you can’t explain how a recommendation was generated, it’s hard to defend it, whether to front-line workers, company leadership or regulatory inspectors. For these reasons, it’s understandable that some people would be concerned about the perceived “black box” aspect of AI. Inputs go in, outcomes come out, but the process seems opaque to many people, who don’t understand how the AI model does what it does and so lack trust in the end results.

Our own surveys confirm these apprehensions. In our results, 69% of respondents indicate that the biggest downside to AI (as they perceive it) is inaccurate output from AI models.

The truth is that some AI can be a black box, namely, the AI you don’t want to use, which also tends to be same AI built from general tools trained on general datasets. On the other hand, responsible AI in EHS is transparent, explainable, and governed.

Modern AI systems can provide traceability into how outputs are generated, in a way that can be explained by the human subject matter experts who designed it. In fact, the best software providers commit to continuous improvement and are always seeking input from end users they can use to further improve the capabilities.

The most important factor to remember is that AI for EHSs supports human decision-making. You remain accountable. You remain in control.

The right AI platform is designed to augment your expertise with explainable insights, not override it with opaque automation.

Trust isn’t assumed. It’s built, through governance, validation, and clear visibility into how the system works.

Myth 5: AI Inevitably Involves Cybersecurity and Data Governance Issues

It’s common to assume that introducing AI automatically introduces new vulnerabilities. This is an increasingly relevant concern considering a growing number of regulations, including EU’s General Data Protection Regulation (GDPR) and California’s Consumer Privacy Act (CCPA), which establish requirements for cybersecurity and protection of personally identifying information (PII).

The reality is that AI doesn’t inherently create security problems, but AI lacking good planning, implementation and governance can and often does.

The additional challenge with AI for EHS is that so much of the data is PII, such as injury and illness records, or ergonomics assessments containing biometric data. That means that you need to be extra cautious when selecting a vendor, to make sure that their software platform operates within strong cybersecurity frameworks, clear data governance policies, and defined access controls. Look for EHS software vendors who can attest that they’re GDPR-compliant, and who attest to SOC 2, in which case they undergo vigorous annual third-party audits of their data protection protocols.

Partnership Handshake Tech Image

When built on secure infrastructure and aligned with enterprise-grade standards, AI can meet the same security expectations as the rest of your EHS technology stack.

Security isn’t optional in EHS. Neither is governance. The best AI for EHS aligns with both.

Myth 6: AI Is Only Useful for Large Enterprises

There’s a perception that AI only makes sense for companies with massive datasets, advanced data science teams, and enterprise-level budgets. There’s also a perception that only large, enterprise-level companies have a significant enough risk profile or long enough “to do” lists to benefit from AI.

That perception leaves many mid-sized organizations assuming AI “isn’t for us.”

The reality is that organizations across the size spectrum can benefit from AI. There’s no minimum size requirement, and the great thing about AI is that it easily scales with the growth of your organization.

While large enterprises may have more historical data, organizations of all sizes generate meaningful safety information—incidents, observations, audits, corrective actions, training records. And the amount of that data can quickly reach levels where human effort alone will struggle to extract meaningful insights from the “noise” of datasets. A perfect example is PSIF detection. Bureau of Labor Statistics (BLS) data shows that rates of severe injuries and fatalities have declined much less than rates of total injuries, which speaks to the difficulty of detecting signals for PSIF risks within incident data. Most organizations of all sizes have struggled with that challenge, but AI with PSIF detection capabilities is the difference maker that can help break these decades-long cycles.

In fact, AI can help any organization:

  • Identify trends earlier
  • Strengthen their incident descriptions
  • Improve RCA and corrective actions
  • Reduce manual analysis time
  • Level up their job safety analysis (JSA)
  • Do better ergonomics assessments that yield actionable intelligence
  • More quickly and effectively process contractor documents
  • Identify chemical ingredients in products, including PFAS, and flag concerns

You don’t need to be big to benefit from AI, and you don’t need data scientists on staff, either. You just need the right platform, designed to translate data into practical, actionable insight.

Safety excellence isn’t reserved for the largest companies. Neither is AI.

Moving from Myth to Momentum

One of the biggest takeaways from the discussion so far is that trust is the gating factor to using and adopting AI in EHS. This message is clear in our own survey data, and it’s being echoed by customers and potential customers who say they want “proof” before adopting AI for EHS applications.

To build that trust, EHS professionals first need to remember another major takeaway, which is that not all AI is created equal. AI that is purpose-built for EHS, trained on real EHS data by real human experts, will demonstrate the reliability and efficiency EHS professionals are rightly demanding.

AI in EHS, at its best, isn’t about hype. It’s about outcomes.

It’s about identifying risks sooner. The right AI for EHS software empowers you and your team to act with clarity, arming you with better information, faster.

In Part Two of this series, we’ll explore additional myths shaping the AI conversation in EHS—and the realities leaders need to understand as adoption accelerates.

In the meantime, you can check out some of our other resources on AI and EHS, including:

Ready to See What AI Can Actually Do for EHS?

VelocityAI and Vēlo deliver AI capabilities designed specifically for safety and operational leaders, grounded in real workflows, built with purpose, and focused on measurable impact.

If you’re ready to move beyond myths and see how AI can strengthen your EHS program, we’re here to help. Set up a meeting so you can see our software in action for yourself.