How can artificial intelligence help with safety culture?

I spent last week at the inaugural Advancing Construction Safety Leadership conference in Dallas.  Construction is a high-hazard industry and it needs these safety professionals, these zealots, these men and women on a mission, who see their profession as more than a job, they see it as a calling. They try and do all in their power to ensure the construction workers go home at night, safe, healthy and with all their fingers and limbs. They have a unique perspective and face unique pressures, that folks who work in safer industries (like me in software), just cannot immediately grasp.

What struck me the most in their presentations is that safety is not just the sum total of rules, policies, procedures and processes, OSHA standards and various other do’s and don'ts. Speaker after speaker took the stage and spoke with great passion how the real building blocks of safety are trust, communication and culture. The see themselves not as “safety cops” or “NO-sayers” or “policy-enforcers”, but rather as communicators, educators and coaches. They work hard to earn and maintain the trust of the crews at job sites – and they want to focus on reinforcing the positive behaviors as much as possible.

Safety manager on jobsite. Image courtesy of ENR.Image courtesy of ENR

But here is a challenge. The ratio of construction workers to safety professionals is easily 60:1. They cannot (and should not!) observe each construction crew every hour of every day. And when they do visit, they want to stay away from only pointing out the faults. In fact, one safety manager I sat next to at lunch told me she would never take a photo of something unsafe at a site until she personally has helped fix the issue! “Because,” she said, “I do not want to only catalog the issues, rather I want to help resolve them, and create teachable moments”.

Which got me thinking.

Photos and videos are taken at major construction sites literally daily – for progress tracking, field observations, general documentation and reporting purposes. There is a statistic out there that 50Gb+ of digital assets are generated for each construction site. And most of it ends up unused and siloed across various systems, applications and devices.

But these photos tell a story, a story of a construction project. And an expert (such as a safety professional) could extract knowledge from those photos. It would be like a “secondary market” for construction photos – where more and more information can be extracted from them by those who know how to look. Granted one photo does not create a trend, but if a certain issue appears consistently for a project week after week, chances are the issue is real.

But who has the time and energy to go through all those photos? Certainly not the safety managers who are already putting in long hours, walking the job sites and connecting with the crews.

The answer is less of a who and more of a “what”. Technology. Computers. Artificial Intelligence, and computer vision to be precise.

The advances in AI and Machine Learning are such that a specially designed and trained system is able to find persons in the image, and also detect if the person is wearing protective equipment such as a hardhat, safety glasses and high-vis vests. It can also recognize the presence and location of ladders, rebar and standing water.

Is a machine as smart as a human? Absolutely not. But it does have important advantages over a human:

  1. It is fast
  2. It can work 7 by 24
  3. It can work in visually noisy and dynamic environments

Which makes AI a perfect assistant for a safety manager. AI can do the heavy-lifting of mining the images for potential leading indicators of risk and creating metrics that can be used by a safety manager to assess “safety flow” of individual projects. The safety risk metrics can be used to compare projects to each other, as well as to themselves across time looking for trends.

Can AI make construction projects safer? I believe so.  I believe it can contribute to the Safety culture and empower safety managers by bringing to the surface the hidden indicators of risk. And those indicators can be used by progressive safety professionals to decide which issues, projects and opportunities to focus on.

 

Putting AI into action to reduce safety risk



Written by Dmitry Grenader

Dmitry Grenader heads up product management efforts at Smartvid.io. He started his career as a software developer, but later joined the “dark side” of product management. A veteran of a startup scene, Dmitry has worked across four VC-backed startups in the Boston area. Most recently he led product management at Luminoso, an AI spinoff from MIT Media Lab focused on text analytics. Dmitry holds a Master of Science in Electrical Engineering degree from Boston University. When he is not at work he… thinks about work.

View more posts by Dmitry Grenader.

Follow me:

Related Posts

Leave a Comment: