Episode 200: AI in Safety - Smarter Tools, New Blind Spots

Blaine J. Hoffmann: Welcome to the Safety Pro Podcast. I wanted to do this episode. It's going to be a short one, but I wanted to do it because I keep seeing all of these posts, mostly on LinkedIn, about AI this, AI that.

Yeah, it reminds me of when HD started coming out. HD, everything was in HD. So I wanted to address this issue. Look, if if you've been in safety for 10 minutes, then You've seen technology change, right?

You've seen the promises technology has made, you've seen the actual letdown that technology has has not been able to deliver in some areas. So I wanted to address this issue ⁓ with you. ⁓ man, do you remember when every software vendor told you that their platform or their app or whatever would revolutionize? your safety program or safety management.

That that promise, I've got a couple of articles ⁓ pulled up here. I we've heard all of it, right? And then of course it it sucks, right? It's it's hard to use.

We run into technology, you know, challenges because the way we've deployed stuff within our organizations, limiting factors there, which we'll get into a little bit, but man. ⁓ I I just I just don't understand it. And then most of these ⁓ systems, they're still generating like PDFs and things and reports that nobody uses and we have to come up with our own anyway. So you have to like data dump raw raw information, put it in an Excel spreadsheet, pivot all of that and come up just to be able to track what we want.

And by the way, I'm gonna I'm gonna have ⁓ give you a call back to another episode I did with Alex from Taproot called ⁓ I think it was called the ⁓ ⁓ something about ⁓ metrics. ⁓ the metrics dilemma. There you go. I don't even know what my own episodes are called.

So the metrics dilemma, and then I had a coffee topic I did, a rant about ⁓ counting the wrong things. We get requests from managers and business leaders all the time to get to tell us like, I I want to count this particular thing, and it's it's not in there, it's not. The tools we use can't produce that metric, so we've got to then go fish for it and create it manually. So it doesn't matter.

There is no one size fits all type of approach. ⁓ AI has has me intrigued, though, in some areas. So we'll talk about that. But look, ⁓ but I'm gonna be honest with you up front.

I'm not here to sell you on AI. I'm not gonna be one of those folks that tells you all of that. I'm also not gonna beat up on it too much. ⁓ I'm just going to be as practical and pragmatic as I can about how some of these tools work or don't work and where we could see us getting into some traps and ⁓ falling into some of these traps.

⁓ I think the ⁓ the big thing is the ⁓ whatever we're talking about, the same technologies that can help us spot more hazards or At least faster ⁓ to get these reported ⁓ as quickly as possible. If we're not careful, we could be creating this false sense of security that will ultimately get people ⁓ injured. So let's get into what I'm talking about. ⁓ let's start with what works, right?

What we like. ⁓ I like that. ⁓ I'll give you an example: QR-based, either hazard reporting or even inspections. QR-based stuff.

Like it's simple. I think it reduces friction. And nowadays we have systems that don't require your device to be registered or ⁓ logged into the service or the system that you could just hit the QR code and submit something. Now, some people don't like that.

Some people go, ⁓ you're gonna have somebody submitting all kinds of stuff. I don't know. There's some guardrails there, but I do like that if I'm a worker, I need to report an issue or ⁓ a hazard or a concern. Maybe it's not life threatening.

Maybe I just need some support. ⁓ I'm able to log that. ⁓ I do like that when ⁓ I have a safety tech or a safety committee member go around and check some exits, or maybe it's fire extinguishers. Boom, hitting that QR code, answering some questions, and that particular fire extinguisher at that location in our facility is marked, right?

Is marked as inspected or or needs service or whatever. I do like the use of the QR codes. ⁓ that's they've been around for a little while, but I don't know. I'll let you tell me if they're overused or overplayed.

⁓ man, this is not an endorsement. of any system, particular system, but I know I know the one is it velocity? It might be velocity. ⁓ they've they launched that ⁓ feature that allows workers to report near misses and things like that ⁓ without needing ⁓ a dedicated app or having to be logged in or registered ⁓ in there.

So ⁓ I kinda like that. It reduces the barrier, you know, the friction that workers face. ⁓ to be able to report stuff. So ⁓ I'm a fan of that.

And AI driven ⁓ stuff is out there as well. Everything's AI. It's probably just software that they're using when I call it AI, but maybe that maybe they are. Maybe they are actually using AI to ⁓ you know, speak with an AI agent or they have it embedded somehow.

But ⁓ a company, there's one ⁓ I'm a It up. Is it Navarra? Navarro Navarro Navarro Navarro. I'm gonna butcher these companies' names.

Sorry. ⁓ they released a tool called, I think it's called FormAgent, Agent Forms, forms, something like that. It will take your paper safety forms, ⁓ your PDFs, ⁓ scanned documents, doesn't matter, and it will convert them into a functional digital workflow, say like safety workflow using their AI, ⁓ Their AI model. So look, ⁓ I know you're probably saying, all right, Blaine, we've been trying to digitize our safety forms since 2010, right?

⁓ I agree. So have I. The difference now is that you're not manually rebuilding every field in the system with this. ⁓ this isn't trying to scan and create a Word doc or something or a fillable PDF.

The AI does all the translation. So, you know, that looks it's not magic. It's just it's a useful application of that technology. And I think that's what's changed over the years.

I can literally just say, hey, make this a a usable form or fillable form. And it's just a scanned image of the old paper doc, you know, and boom. ⁓ you can have it in your in your system, ⁓ maybe SharePoint or turn it into a workflow. Pretty cool.

Now, ⁓ This really is useful for companies legacy ⁓ using ⁓ legacy systems, ⁓ paper systems, stuff like that, because you don't have to start from scratch and the labor needed to upload all of your stuff. ⁓ the thing that comes to mind with this with AI here is what where people go, well, give me an example. ⁓ Safety data sheets are as easy as just scanning and and and uploading them. And then being able to open them and read them.

But think about lockout tag out procedures. If you go from one system to another, like provide service provider, and you've got your ⁓ lockout tag out procedure written in a certain format. And then this other service provider or company that's going to help manage all of these, pretty soon it's AI is going to be doing all of it and replacing all these companies. But in my opinion, okay.

But ⁓ you're able to scan these and say, Hey, I need this lockout tag out procedure with the photos, right, to match this template. And then it'll just do it. All of it. It's crazy.

It's crazy. I've heard of companies that will do this manually. They'll say, We'll take all of your old, old format, we'll upload them, scan them, and convert them. ⁓ And you have how many procedures?

And it was like a hundred, hundred thousand dollars or more for the company to to do that. AI can do this now. ⁓ that that's gonna that's gonna be seamless. It's gonna be real easy to do.

⁓ predictive analytics and computer vision. This is an area that gets weird, but there's the good and the bad. ⁓ this is where ⁓ things got like I say, go both get impressive and get ⁓ you know, there's a caution. ⁓ that you have to follow here.

All right. ⁓ the National Safety Council, in their 2026 report on emerging workplace safety technologies, highlights computer vision and predictive risk modeling as one of the most promising tools in the current ⁓ digital EHS landscape, if you will. So we're talking about camera systems that can identify when a worker isn't wearing a hard hat or ⁓ ergonomics is the is the biggest one where you get flagged, you get a notification if, hey, Blaine's over there on the dock, ⁓ not lifting with his legs, you know, that kind of stuff.

You can see where this could could go wrong real quick, right? ⁓ so predictive analytics can take some in very limited in in the where where it is right now, and I'll explain why. Incidents, near misses, things like that. ⁓ and can look for patterns.

It can help you look for look for these patterns. You could it can AI can actually look across ⁓ past incidents very easily and say, hey, where have we had issues with this equipment before? And you can kind of have a chat agent talk ⁓ within your system to say to kind of surface that those trends. so the vision system stuff, the goal isn't to surveil your work.

Right. If if you're doing that, you're using it wrong. ⁓ I do see an opportunity to ⁓ to identify. Maybe I I'm not looking at the video, but maybe I'm relying on the system, the the AI, I guess you could call you could call it, to in the background say there was a pattern of for the last two and a half hours, this worker has had to work.

at floor level with loads, with with objects, with moving objects. So it's like, okay, it's not, hey, you need to start, you need to use proper lifting techniques, right? It's w why are we not ⁓ breaking up that type of work or ⁓ making the work easier to do and access the materials that they have to loot move manually, right? And so it's it should force us to look at our process the in the situations we are putting workers in versus starting with, hey, we got a notification that you're not lifting with your legs, that you're, you know, you're doing a lot of, you know, bending at the waist.

And ⁓ don't do that. And the workers are like, I got all this crap to do. Like, and I have to do it in this amount of time. The only way for me to do that is to shave off some time with my movements.

And that means a lot of twisting and a lot of like just reaching to the side and pick. So We we miss that context. I don't think AI is not going to give us that context. That's the problem with the vision systems and everything.

It might show us what was happening, but AI is never going to be able to tell us what's going on in the mind of the worker at that moment. Right. ⁓ even with noise and and say, okay, well, maybe it was loud and they didn't hear, like for incidents or near misses, worker. Worker needs to pay attention.

If the video is showing a worker picking up an object and turning and then almost walking into somebody, and you know what AI is gonna say. They're gonna say, well, worker, worker wasn't paying attention, wasn't aware of their surroundings. That can't be the root cause, right? That can't be even if AI and vision systems can listen for, you know, for sound, for you know, and record sound.

You're gonna run into a lot of laws. States are different on this, and certainly countries, on recording audio and and sound and what workers are saying and things like that. That's why most security systems, security camera systems in buildings and in in workplaces, don't record audio because you can't. You'll get security companies that'll say, No, none of our cameras even have that capability ⁓ because it's against the law.

So, you know, we're going to get to this ⁓ point, ⁓ an inflection point, if you will. Is that a guru phrase? I could I could be a guru now and say that. We're gonna get to this phase, ⁓ this stage where, you know, we're gonna have to look at regulations and laws, and we're going to have to side with as a culture, as a society, some of these foundational principles that we have of, you know, right to privacy and and things like that.

within reason. So I I I don't see that changing anytime soon, but yet this technology is rapidly evolving. And so even if this vision system could detect noise and sound, ⁓ we're not permitted to. So the video might show us what was happening, but it cannot tell us what's going on in the mind of of that worker.

⁓ maybe it can tap into area temperature ⁓ monitors or just noise monitors to see how noisy it is, not recording sound, if you will, and say, hey, it was noisy, ⁓ it was hot. ⁓ so fatigue could be a factor. So it could help us with some of that stuff, collecting and and presenting us with that information. ⁓ we still have to take that information and work with ⁓ the the workers on the floor.

⁓ supervisors and actually go talk to people and work to come up with mitigation ⁓ solutions. I think that's the piece that's missing a lot when we talk about these ⁓ technologies. Wearables are another one. You're gonna get into a lot of privacy issues when, especially when we're talking about biometrics, ⁓ things like that.

I think that ⁓ VR and AR training is still, you're you're not going to. breakthrough. I I know a a company that worked that piloted a lot of ⁓ VR training. It was super cool.

I did some of it. It's super cool, but they learned, and the this is an industry standard, that about 20% of the population, like this will make them ill. It'll make them sick. Motion sickness, right?

They can't, they can't do it. It's just not something they're gonna they're gonna do. ⁓ so you've got to account for 20% of your population not being able to use VR or even maybe AR ⁓ to some extent ⁓ training. So how are you going to train the, you know, train?

You're only gonna net 80% of your workers at best. So again, it's not a cure all. Technology is not a cure, it just has to be a thing that we can use when it makes sense. ⁓ couple of other things.

⁓ algorithms, ⁓ man, we're gonna get a little overconfident with some of this stuff. The if you implement predictive ⁓ analytics, ⁓ some sort of platform that's just collecting data and it's saying, here's where we think your trouble spots are, go, go look. this could generate a risk score for our facilities. ⁓ maybe every morning we get a little flag ⁓ or a report that flags some areas to go to go look.

And maybe, maybe it says everything's fine, right? ⁓ someone like maybe a supervisor, ⁓ a manager, ⁓ safety tech, ⁓ maybe under a certain amount of pressure to hit production targets or goals. We look at that green score that everything is good, and we make a decision that we wouldn't have made otherwise. So ⁓ there's ⁓ some research on this actually.

⁓ Can't remember. I'm gonna have to say it was like ⁓ I think it's look it up. I had it here earlier. I was looking at it.

Pul Pilati, Pilati and colleagues, Pilates and colleagues. I've got multiple monitors here, so I'm gonna try to bring everything over to my my left side over here. It's easier for me to see that screen is closer. ⁓ Pl Pilates, drag it over in front of me.

Yeah. Artificial intelligence and emerging risks in occupational safety flags algorithmic decision making creates novel cognitive and organizational risks. So when workers and managers start deferring to the algorithm instead of exercising their own professional judgment, then you transfer risk from the hazard to the system, is what they're saying. If the AI didn't flag it, it's gotta be safe, right?

Must be safe. And so that's that could be a risky assumption. Just saying, well, the day it didn't show AI didn't pick it up, so it must be okay. ⁓ a risk score is a probability estimate, so it's not a guarantee.

That's the issue. Okay. ⁓ if your team doesn't understand that distinction, then you've got a training problem that no AI tool is going to be able to solve. ⁓ the ⁓ The kind of goes with garbage in, garbage out, right?

Which is a whole different issue. Predictive analytics is only as good as the data or what it's trained on. Okay. So EHS data in most organizations is incomplete.

It's inconsistent or shaped by under reporting culture, stuff like that. If our near miss reporting rate is low. The AI is learning from incomplete data or getting an incomplete picture. It's only going to be able to analyze what is in there, what it gets, what gets reported.

So ⁓ you could have what's known as ⁓ I think it's called the Yelp effect. I don't know if anybody's coined that phrase. I've heard that phrase though, the Yelp effect, meaning ⁓ most people that have a good experience don't say anything, right? You're only likely to get a review or somebody to be motivated to leave a review if it was negative.

So we could be looking at reviews from a restaurant that gets thousands of customers a week, but there are 125 negative reviews out of 140 that were left. So we'll go, we're gonna skip that restaurant. And they're they're doing mad traffic every week. And so But it's just the vast majority of people don't comment, don't leave a review.

That's very anecdotal. No scientific evidence or proof for that. I've heard this, makes sense to me. So, you know, you're basic you're basing your decision on those that chose or were motivated by a negative experience to leave a comment and not seeing that probably over the 10, 15 years an establishment's been open, the tens of thousands, if not more, hundreds of thousands of people that had a good experience and just didn't say anything.

⁓ so that that excuse. So if you get like if you get a two or a three out of five star rating, that's only out of the ones that rated it. Does does that make sense? So that's what I'm trying to do, is I'm trying to make that correlation of okay, our our algorithmic system here.

Is only going to base a decision or a risk score on the information that it it has entered. And if you have incomplete information, you have an incomplete risk score. So that could that could go the other way. It could score stuff crazy risky.

And then we're dumping a bunch of resources into something that's not as bad as we think it is. ⁓ anyway, you got problems. You got you just gotta be involved in in this. So Is the answer no technology?

No, I don't think so. the the low tech building of the low tech work, I should say, of building a strong reporting culture is still foundational. So we have to get that culture of and we want to hear from you by any means necessary. If you have to tell your supervisor, that's fine.

If you have to tell confide in somebody who's on the safety committee, ⁓ that's fine. You if you just rely on the technology, you're going to be missing some stuff, right? So ⁓ you need to ask, do we trust this data? Right?

How can how complete is the data? ⁓ is it representative of what's actually happening out on our shop floor or out on our work sites? You know, ask those questions. If the answer is no, then you need to invest in that low-tech, no tech culture building stuff first.

That's all. ⁓ psychosocial dimension. ⁓ this is another one I wrote down. This is these are some crazy terms.

I so I made the mistake of looking up some common issues in ⁓ organizational psychology, ⁓ common issues that come up. This one doesn't get a whole lot of attention. The same journal of that study that ⁓ that researcher I mentioned before. They highlighted psychosocial and ethical challenges from AI integration in the workplace.

⁓ and this is directly related to safety culture. Workers who feel surveilled, I mentioned it earlier. No, forget the laws and everything, but the cultural aspect. If you workers that feel like they're being surveilled by AI cameras, wearables, they may experience increased stress and psychological safety concerns, right?

And this is psychological safety really is when I hear that term, it's been overused, I think, and and mis ⁓ redefined and misdefined. Misdefined. That's another word I'm gonna coin. All right.

⁓ psychological safety. That that's what this means to me. ⁓ they may modify their behavior not to work more safely, but to avoid ⁓ not in order to work more safely, I should say. But they'll they're motivated by just simply trying not to trick ⁓ trip a flag or on the algorithm or on the monitoring.

Okay. so the the the story I and this is actually this is a story that happened the that I came up that I can remember. And this was many years ago when step monitors, ⁓ pedometer. They would issue, and this is before wearables really became popularized.

⁓ the this one company I worked with, I when I was consulting, they had this ⁓ step challenge or whatever it was called. And then they had workers competing against each other ⁓ to on these little teams to who can get the most steps, and then they incentivized it. And there was a like a at the end of the year, there was like a little ⁓ three day weekend trip somewhere or whatever, like they made it valuable. You can imagine what happened.

There were certain workers in certain departments that they didn't have a lot of opportunity to walk. So they had these pedometers and they would just sit at their desk and just shake the thing, right? One actually had a spring and he would hit it every once in a while. And that thing would just bounce for a couple of times.

And they had all kinds of crazy tricks on how to get more steps on the pedometer. And so their focus wasn't they they lost the initiative, right? That well, we want. To encourage wellness, healthy living and lifestyles, ⁓ things like that.

And what they did is they incentivized it to the point where ⁓ they made it too valuable and the folks that couldn't actually participate felt like they were missing out. So they modified their behavior, not for the wellness aspect, but ⁓ they for the the tr the monitor that they had to that that was reporting on their steps. So they they modified their behavior anyway. ⁓ Might not be exactly apples to apples, but look, that's what that's what this psychosocial impact could be.

I'm not motivated to work safer. I'm motivated out of fear. Like I don't want to get caught or flagged or in trouble, you know. And so I'm r you know, actually focused on the wrong thing.

It could be distracting ⁓ and that could lead to injuries as well. But look, viewing it as a management surveillance program rather than a Commitment to improving how we work and improving safety, that could easily that line can get blurred. And you will have no control over that once they get it in their head. Yeah.

It's gonna take on a life of its own. I've s I've seen other versions of of that play out, like I said, with simpler technology. ⁓ GPS tracking is a another one with fleet vehicles. ⁓ if you recall anybody with fleet vehicles, when if you're old enough to ⁓ Remember when the when deploying these on within a fleet of vehicles became popular, ⁓ the initial response wasn't real great from drivers.

⁓ I'm gonna drive like I'm gonna drive more safely. ⁓ that that wasn't it. It was you're going to track where I'm going, you're good. It's going to be used to say your why are you still at that ⁓ rest stop or or that restaurant, things like that.

So how you introduce these tools is as is as important as the tools themselves, right? If not more important. If AI safety technology is pushed as management watching you, or or that not pushed as management watching you, is received as management's going to be watching us, then you're gonna damage or harm your s this culture, ⁓ the psychosocial culture while trying to improve safety, which is not going to improve anything at all. So that's That's an act, that's human nature, and that's a real thing with all this stuff.

⁓ deskilling. De-skilling, something not a lot of folks think about. ⁓ and this is a long-term thing, okay, that you got to look for. As AI tools get better at identifying hazards and and flagging deviations and recommending corrective actions.

What happens to hazard recognition skills? What do you think is gonna happen? Right. One, ⁓ I'm going to be told I don't have to because I'm going to and sort of like ⁓ over reliance on the algorithm, your over reliance on the technology, okay, ⁓ while I'm working.

I don't need to pay attention to the hazards because it's gonna tell me. If that makes sense. If the camera catches the missing hard hat or ⁓ you know, is the supervisor gonna continue to start, you know, to look for that stuff? No, it's gonna be like, ⁓ I don't have to go around.

And check hard hats anymore. Not that you should be checking hard hats, but you know what I'm getting at here. It's behavior is going to change. Okay.

⁓ de skilling is a real phenomenon in high automation environments. And aviation has studied this a lot. ⁓ automation has made flying safer, arguably, right? But it has also created pilots or can create pilots who struggle when automation fails because.

Skill development is missing. And the goal of AI and safety should be to augment right human judgment, not to replace it. So building that expectation into your program, making that clear that the tools support the people, not the other way around. That's a leadership responsibility that no software vendor is going to hand you or do for you.

All right. So one, don't let the hype cycle Of all of this stuff, make decisions for you. There are real practical tools available for you right now that can meaningfully improve your program. ⁓ so particularly in incident reporting, training, and real-time hazard monitoring, pick the ones that address a specific documented problem in your organization, not the ones that have the best demo or make these wild claims.

⁓ if you should go to the if you're on LinkedIn at all, go to the tap root. ⁓ this is not a plug. I have No affiliation ⁓ at this time with ⁓ Taproot, but there ⁓ man, ⁓ when was this? Mark Mark Paradise from Taproot did a video on AI root cause, like doing the root cause for you and telling you to root cause.

Look for that video on their ⁓ LinkedIn profile page. ⁓ it's very interesting. That's an example of overselling, somebody selling you something that it can't really do. Okay.

Second, invest in your data quality before you invest in any of these AI tools. Remember, garbage in, garbage out. I I worked with a company where ⁓ they wanted to improve and it was it was a really good workflow, by the way. A sort of ⁓ guide that would guide you through some human factors questions to ask based on the type of incident that you've entered.

So it learns based on what you've entered. And they said, ⁓ well, what we'll do. is ⁓ we'll just train ⁓ the question bank on the last three years of entries and like no the three year the last three years of ⁓ incident analysis ⁓ forms and reports results were garbage. The whole reason why we had to transform the way we did incident analysis is because it sucked.

Nobody knew nobody was doing it right. They would always get like 80% Of the corrective actions, ⁓ or 80% of the root causes were ⁓ worker error. And you know, 60% of the corrective actions were coaching. And so it's like you're gonna train your your AI to do more of that.

That's the problem. So before you invest in these AI tools and technologies, invest in data quality, clean that up as best you can. ⁓ and if that means you got to go back. And correct some of some of the stuff or handpick what good looks like, because what you're telling AI is like this is what we expect.

Even if it's a handful of cases or pieces of data, if it represents what good looks like, where you want to go, then that's your benchmark. AI should be benchmarking off of that, not diluted with a bunch of garbage because that's what you're gonna get. ⁓ all right. So ⁓ and again, like the same thing with your 300 log.

If you've got reporting issues and and she's got a dirty OSHA 300 log, something as simple as that. ⁓ if that's inconsistent, fix that stuff first. All right. ⁓ when you roll out these AI tools, do it with your workers, with them, alongside them.

Don't don't force this onto them after the fact. Okay. Explain what the technology does, ⁓ what it doesn't do, more importantly. Be transparent about the data that is collected and who sees it, who can access it, how it's used.

You gotta get them involved up front. They don't want to hear about some new thing coming out hitting the floor after you've already invested and made the decision. And they have all kinds of problems with it and questions and issues. ⁓ their buy-in is the difference between a tool that improves safety and one that just wrecks your workplace culture.

⁓ Maintain those human skills. Remember that de-skilling issue. ⁓ don't let AI observations replace human observations. Keep your safety walks, do your field-level hazard assessments, keep your behavioral safety conversations going.

These aren't redundant with AI, they're the immune system that will catch what AI misses. So just think of it that way. And ⁓ of course, stay skeptical. Of these ⁓ all is good, we didn't catch anything type of scores or ⁓ reports.

⁓ build in a standing assumption that your risk profile is never fully captured by any model. Only parts of it are, and at certain parts of moments in time, I should say. The hazard you haven't seen yet is not the same as a hazard that doesn't exist. So the bottom line is: look, AI is genuinely.

changing what's possible in the workplace. The tools ⁓ that we have access to is ⁓ it's nuts nowadays, the stuff that that we can do, right? ⁓ but you're gonna run into challenges. And and I'm gonna go back to the what I opened with.

Something as simple as a QR code. If you've done any of this ⁓ in the last five or six years, and maybe even longer with QR codes, then you'll probably know what I'm talking about. What I started with was very simple. Just taking a picture of a QR code, it opening a a web page where I can submit some information or read some information, right?

That's it. Maybe it's connected to an app or something like that. But that's all it is. How many of y'all have policies in your organizations that prohibit the use of cell phones in the production area?

Right? Probably all of you. And I'm talking about. You know, workers' managers, supervisors.

Okay. So you allow them to use it because it's it's going to be for business purposes. But then maybe you give them a work phone to use for that purpose. So if you're going to roll out QR codes, hey, instantly report your safety concern.

We want to hear from you. And the way that we've done it as a safety team is using QR code codes requiring them to get out their cell phones. Now they're standing on the production floor at a poll or a post or maybe a kiosk and they have to get their phone out. And they're taking fit photos in the workplace.

You could run up against those hazards, or that the hazard actually of them violating a company policy. So it's not as cut and dry as it sounds. Make sure you consider all of these things. ⁓ and look, let me know what you think.

If you have ⁓ an example of how this has gone poorly, if you have an example of how this has gone great, ⁓ we want to hear from you about that. as well. So make sure you share this episode with others if you found it useful. And be sure to like and subscribe so you don't miss another episode.

And we'll talk to you on the next Safety Pro podcast.