As the cliché goes, work smarter, not harder. While appealing, it can be difficult for our brains process and maintain large amounts of information, especially when it’s difficult to conceive what’s even possible.
That’s where artificial intelligence (AI) and other safety technology tools come into play. These programs can quickly sift through mountains of raw data to find trends and brainstorm solutions.
It’s an exciting prospect for safety professionals, and the possibilities are still being conceived. However, it can also be overwhelming. That’s why it can help to hear from someone else on their journey.
That’s what Geoff Walter, CSP and corporate director of enterprise safety at Owens Corning, will be share how he reconnected with his former Honeywell colleague, Keith Bowers, to use AI to advance workplace safety.
Walter and Bowers will be speaking at the 2024 Safety Leadership Conference that’s taking place Aug. 26-28 in the greater Denver area. More information, including registration, can be found here. Below is a conversation with Walters on what to expect from their presentation.
EHS Today: How is Owens Corning using AI to improve workplace safety?
Walter: The first way that we started to use AI was around big data types of applications. Like many companies, we have a big database that has all of our incident reports. You can go in and do dropdown menus and some categorization, but a lot of the gold in that data is really in the textual data—all the text from the incident investigation and circumstances. You don't really capture that through dropdown menus.
The idea was to use AI tools to go through all of that historical data and try to identify some of the hidden trends inside of that. That helped to inform our serious injury and fatality strategy. We were just starting to talk about setting up a separate strategy from our high frequency, low severity incidents. That was the initial foray into AI. Since then, we've used it for some other data applications.
For instance, we just finished rolling out a new safety scorecard using a lot of existing metrics, about 18 or so existing data streams, that goes back seven to nine years’ worth of information. That's been able to help indentify where we've got increasing risks at some of our facilities of having incidents. There is no new information being generated. We are just leveraging existing data. We rolled that within the last two or three months, and we're very excited about that.
A lot of data analytics went into it to identify the correlation between those metrics and actual injuries. And then that system just keeps updating and refreshing as more data comes in. It uses AI to continue to do those comparisons, deep analytics around it and tells us which ones really matter. It’s a very, very interesting process.
What was that process like when you were looking at working with Keith and this AI tool? How did you get buy-in from management and from workers? How did you determine this is an appropriate course of action to take and to make that up-front investment?
We have a safety technology committee, a cross-functional group of safety professionals, people from IT and advanced manufacturing, that looks at promising technologies and, of course, AI is one of them. We did this on a small scale, more of a pilot, to see what the potential is with AI. When it comes to big data, there wasn't a lot of convincing that was necessary because you're just going through the data. I hired Keith to come in and gave him access to all of our data.
As you start getting more on video, the sensitivities around how that information is being used increases. One of the things we've had to do, and it's consistent with how we approach a lot of things, is making sure we've got good communication with employees. We do blur out faces. It's not about catching people doing things. We just want to see what's really going on so that we can make changes to the process and do coaching if it's necessary, but it's not around the disciplinary side of things.
We have not had any significant pushback at all. People understand. I think now video has become so prevalent in societies around most parts of the world that there isn't as much resistance to it as maybe there was at one point, especially the way that we're trying to use it. And we have to hold true to how we said we were going to use that information.
Are there any information you can share as far as things you've identified or been able to correct thanks to AI?
Through our AI traffic study, we’ve made changes in our traffic flow to reduce interactions based on what we've seen. We've taken other actions with some other, non-AI sensor technologies to try and prevent powered industrial truck and people interactions. We actually identified some unintended consequences of putting up some physical barriers in a certain way that actually increased risk. That is really helpful because we weren't getting the results that we had hoped for out of some of the other actions that we've taken.
I'm guessing that you were able to identify quickly with the video that you were achieving those unintended consequences and then immediately changed it out before anything could happen.
Yes! It shows where we've had issues. The system can find those rather quickly—we have hundreds of hours of video—but the but the AI can drill into the ones that really matter, and that's really the value in it.
Are there other safety concerns that Owens Corning is hoping to tackle with AI in the next six to 12 months?
Yes, for sure! Keith and I have brainstormed and our internal teams are thinking through some of the other potential opportunities. Some of those are in safety, some of those could be operational types of things, some could be quality related. If you can use one system for multiple purposes, that really helps to justify from a cost perspective.
One of the more innovative opportunities that we've talked about is machine guarding. We've got awesome employees, but sometimes people do some things that you don't really intend for them to do or want them to do. How do we make it impossible for somebody to, you know, to get hurt by your equipment?
AI has the potential for being able to identify when you've got people in places they shouldn't be. You can do geofencing, so if you see somebody that's in a spot they shouldn't be in because the equipment hasn't been appropriately deenergized, it can identify those situations. It's more of a secondary protection for employees.
There's a whole bunch of potential uses. As I said, it's very much the early days. We'll continue to identify new opportunities for AI.
Have you received any feedback from employees so far on what you're trying to do and how you're using AI?
The places we've done traffic studies has been just great, especially for our site EHS professionals. Being able to get that feedback, get that information and to see how that has supplemented some of the manual ways that they've done some of that same work in the past has been really good.
For employees, being able to see what's going on, being able to do some real-time coaching has been really positive as well. And then being able to actually turn that into some physical changes to traffic flow and how our people go through our facilities has been positive. We really haven't had much pushback from employees or major concerns that have come up.
The scorecard approach is still in the early days, but our folks are really excited about using that to see increasing risk. Instead of just reacting to incidents, we have the ability to see risk increasing before those incidents actually occur. I think that'll flow into the facility that we can actually take some action to prevent things from happening instead of just reacting to them.
Could you tell me a little bit about how you’re targeting higher frequency but lower severity injuries with AI and what you’re seeing as a result?
It’s actually different for every facility. That's one of the powers of this; it's not a one-size-fits-all approach. The risk interplay at a facility is a little bit different by site depending on a lot of different factors. It could be the manufacturing process at that facility. It could be the tenure of the employees of that facility.
AI looks at every facility differently from a risk perspective. For example, it looks at how much turnover a facility has, how many new employees a facility has and how many senior leadership vacancies we have. We know that as these increase, so does risk because we don't have as much leadership presence in a facility.
It looks at how a system is or how a site is operating. Are they operating well, or are they having a lot of upset conditions and manufacturing issues? We know when people are doing non-routine types of tasks or unjamming equipment that drives up risk. It actually looks at how our sites are doing on getting all their corrective and preventive actions closed out and getting all their safety work orders done. Then, it'll analyze all those interplays at every facility and it kind of generates a different kind of risk profile for every site. When we see risk of a certain type of risk increase, we know, OK, turnover is not something you can you know just fix overnight, but there are some mitigating actions you can take. We might need more leadership presence on the floor or set up a mentor process for new employees because we have a lot of new people coming in.
If you don't see the risk, you can't take action. This is really helping us see those risks in real-time before it actually results in an incident.
What's one thing you hope attendees take away from your presentation at the Safety Leadership Conference?
I don't know that an Owens Corning facility is necessarily representative of attendees’ facilities, but my hope is that they can see some of the ways that we're starting to use AI, some of the results that we've seen, and it gets their creative thinking going about how could they use it in their facility. And just to get comfortable around the term AI.
We hear a lot in the news about AI that isn't necessarily positive. In this case, the more people we have thinking of creative uses and how can AI be applied to benefit our employees, the safer our workplaces can be.
