How employers are using AI to prevent workplace accidents

Potential applications

Artificial Intelligence (AI) is slowly entering the field of workplace safety. The potential applications and advantages are numerous:

• AI can incorporate and analyze data efficiently and at rapid speed. It can generate real-time safety reports, prioritize risks, and help managers make informed decisions quickly.
• By analyzing historical safety data AI can proactively identify potential risks and suggest corrective actions and preventive measures.
• AI can automate safety inspections with AI powered camera and drones, image recognition algorithms, and thermal imaging.
• Technologies such as computer vision can actively monitor environments, detect hazards, unsafe behaviors, and fatigue, identify equipment malfunctions, and provide real-time alerts.
• AI can help keep track of regulatory changes and ensure compliance. Similarly, it can ensure compliance with company protocols and training programs.
• Conversational AI can address employees' safety inquiries instantly, enhance training, and promote a culture of safety.
• AI-driven stimulations and VR training allow workers to practice operating equipment in a realistic virtual environment without risks.
• Wearable devices can alert workers about immediate risks and provide intervention guidance, analyze patterns, and help improve preventative measures.
• AI-powered robots can be used for hazardous tasks and hazardous area inspections.
• Accurate AI can be used in predictive analytics to identify patterns and predict potential accidents and equipment failures.

Examples of companies that are leveraging AI to enhance workplace safety

Texas-based 3 Men Movers uses AI technology to detect distracted drivers and optimize route planning to reduce accident rates and insurance costs. The system is trained to monitor driver behavior, alerting both drivers and supervisors to issues like smartphone usage, eating, drinking, and even yawning while driving. Additionally, they utilize advanced routing technology to optimize delivery routes, avoiding high-risk areas. Although the rollout was not seamless and some initial products received too many false positives, the initiatives eventually led to a 4.5 percent reduction in accidents within the first three months and achieved a 91 percent accuracy in detecting distracted driving behaviors and preventing 80 percent of distractions. businessinsider.com

Cupertino Electric, a national electrical engineering and construction company based in San Jose, California uses AI to assess workplace hazards on their job sites, primarily through a tool called "SmartTagIt" which allows safety professionals and field technicians to identify, share, and mitigate potential risks in real time. Since its implementation in 2019, the company has collected nearly 74,000 field observations. The A.I. tool helps prevent injuries by assessing how often hazards should occur and when there is a higher-than-normal number of incidents. ecmag.com

A Denver restaurant uses a combination of artificial intelligence with fiber optic interferometer technology to improve kitchen safety. AI algorithms learn the sounds, look, and feel of the new environment and then continuously observe for minute deviations from any patterns in the kitchen environment. It can identify the specific sounds and movements associated with employees who were not correctly wearing gloves and recognize environmental changes that could indicate spill hazards. When deviations are detected, real-time text alerts are sent to restaurant management with a detailed description. ISHN

Facing increasing retail theft, Laurel Ace Hardware, turned to AI. By integrating Veesion's AI software with the store's security cameras, the system analyzes surveillance footage to detect theft-related movements and alerts staff in real-time. This approach has resulted in a 50 percent reduction in shoplifting incidents. sfchronicle.com

Amazon workers wear robotic tech vests in over 25 of its warehouses, which allows robots to detect and avoid workers, reducing the risk of collisions and serious injury.

Whirlpool's washing-machine factory in Clyde, Ohio, has eliminated forklifts from its production area and uses robotic tuggers to deliver parts to assembly-line workers. Other company plants are following suit. Wall Street Journal

In addition, some PPE manufacturers and automation manufacturers provide tools like video analytics and AI-powered simulations to enhance worker safety training and equipment inspection.

Barriers to implementation

While AI presents a significant opportunity for businesses to improve workplace safety, why has adoption been limited?

• Data. The foundation of effective AI, large amounts of data must be current, accurate, complete, unbiased, regularly cleaned, updated, and verified. AI must be fully trained on the company's processes and workplace procedures. Incorrect or incomplete data can lead to false positives (indicating non-issues) or false negatives (not detecting real hazards), reducing AI effectiveness and triggering employee mistrust and resistance.
• Costs. AI systems, including sensors, wearables, and computer vision tools, can require significant upfront investment. It can be challenging to calculate the expected return on investment (ROI), including reductions in incidents, liability, downtime, and on-going costs.
• Lack of AI expertise. Many companies lack in-house AI expertise, making implementing and managing AI-driven safety solutions difficult. Organizations should evaluate and selectively adopt technological solutions that fit their operational capacity.
• Integration challenges. AI systems must integrate with existing workplace safety programs, equipment, and IT infrastructure. Cyber security is another concern.
• Time and testing. AI should be rolled out slowly, with a pilot program, and closely monitored, tested, and modified. It must be continuously updated to reflect new safety standards, workplace process changes, and evolving risks. It's a long-term engagement and investment.
• Employee resistance. Workers may distrust AI, seeing it as a surveillance tool, invasion of privacy, threat to their job, and ineffective rather than a safety measure. Transparency is critical for AI implementation.
• Changing conditions. Dynamic conditions and unpredictable environments in industries such as construction and manufacturing make adoption more complex.
• Legal compliance risk. Regulations around AI are still in their infancy and constantly evolving. Employers may want clearer guidance before investing heavily.
• Appropriate balance of AI and human oversight. Overreliance on AI is a serious risk that can lead to complacency, system errors, overlooked risks, and false positive and negatives. The technology is not infallible, and it lacks human intuition and contextual awareness. It must be constantly monitored, tested, and validated by humans. It's a complementary tool that should enhance, but not replace, human decision-making.

How to begin

1. Identify the workplace safety challenges you want to address with AI. Assess the availability of current and historic data needed to train AI. Identify potential hazards and risks associated with AI implementation.
2. Develop clear guidelines for data collection, usage, and decision-making. Establish who will be inputting and updating data and who will be auditing AI insights to confirm accuracy. To be successful, AI implementation depends on well-trained and knowledgeable employees to input and interpret data.
3. Research AI solutions. Consider beginning with safety inspections or monitoring equipment. If employees are skeptical, this allows them to see how it can enhance safety. Then grow the initiative to interacting with employees. Evaluate multiple vendors and solutions to determine the best fit for the company's size, industry, and budget. To validate the technology, companies should always ask tech providers for examples, such as case studies, information on how their results compare with industry benchmarks, and who will be responsible for implementation and identifying and correcting when AI tools deliver misleading or inaccurate results.
4. Start slowly on a small scale. Begin with a pilot program - one department, facility, safety issue. Monitor AI effectiveness, accuracy, and employee feedback.
5. Be transparent with employees about when, how, and why the AI solution interacts with them and their data. Provide a system so employees can share their suggestions and complaints about each solution. If employees provide input and have a stake in the AI systems, they will be more likely to embrace these systems. Provide training programs to familiarize workers with AI-powered tools and address concerns about privacy and monitoring.
6. Ensure data privacy and compliance. Understand legal and ethical implications, including compliance with federal, state, and local regulations and data privacy laws. Stay informed about new developments in the technology and in the legal and regulatory landscape.
7. Monitor AI effectiveness, accuracy, and employee feedback. Evaluate performance metrics regularly and make adjustments and upgrades. Have a plan to address new, emerging risks.