AI

The elements of

Technology

Monochrome photography, Black, Black-and-white, Pattern, Style, Line, White

Video: VectorFusionArt / Creatas Video+ / Getty Images Plus via Getty Images.

Preventing workplace injuries with artificial intelligence

By Ryan Quiring

orkplace safety injuries in the United States alone cost $171 billion per year according to the National Safety Council (NSC). And despite companies’ investments in safety measures, the annual rate of injuries during 2015-2019 increased anywhere from 1% to 5% — except in 2017 when they remained flat.
The challenge, as many safety managers and human resources (HR) professionals realize, is that mitigation is too little, too late. What’s needed is the ability to more accurately predict where incidents may occur and prevent them from happening in the first place. For this reason, more safety and HR managers are turning to artificial intelligence (AI) technology.
Let’s look at how safety and HR professionals can apply AI to a company’s safety initiatives, provide the data necessary for obtaining meaningful results, avoid common pitfalls, and get the answers needed from an AI assistant.




W

Three main components
When applying artificial intelligence to workplace safety, there are typically three components:
1. A natural language processing/understanding algorithm lets people make requests to which an AI assistant can then respond.
2. Machine learning serves as the neural network that scans data to locate patterns.
3. A visual component may include a closed-circuit TV (CCTV) on the shop floor or work site to observe workers using equipment. Alternatively, an optical character reader (OCR) is used to convert paper records into digital data for analysis.
How AI technology is applied to safety can take any number of forms. For example, an AI assistant could scan streams of data to identify early indicators of machine or tool failures that can potentially lead to injuries. Alternatively, it could review 1,500 hazard assessments in the company’s database to pinpoint the 15 that answered “no,” so managers can focus on that group of anomalies.
The good news is that companies don’t need to invest in special natural-language and machine-learning solutions, since these technologies increasingly are being built into business applications to extend their functionality.

Collecting data
Before reaping the benefits of artificial intelligence, companies need to collect the data to be analyzed. The first step is to clean and normalize the data, structuring it to be repeatable across all of the records that exist.
The natural inclination is for businesses to take advantage of their existing data. However, if this information is stored in spreadsheets, PDFs, Google docs, or file cabinets, the AI project tends to stall before it ever gets started. That is because the data will need to be entered into a database that structures the data for consistency, a process that can take months to years to complete and cost tens or hundreds of thousands of dollars.
By contrast, companies that have implemented applications, such as environmental, health and safety (EHS), human resources (HR), and enterprise resource planning (ERP) can begin to quickly take advantage of the insights provided by AI. That is because these applications automatically normalize the data and store it in some form of database, which the AI functionality can access.
Organizations without an EHS, HR or ERP solution already in place should consider implementing one of these applications. Once they have done so, managers can start collecting data digitally and gain important insights from their safety programs in as little as four weeks. Then, it’s possible to go back over time, take those historical records most likely to enrich AI-driven analysis, and enter them into an enterprise application or directly into a database.

Since more data is better, it helps to have all employees collaborate in the collection of data. The key to workers’ participation is letting them use mobile phones to access web applications, which automatically normalize the data.

Capturing the right details
Beyond structuring their data, safety and HR managers also need to ensure that they are collecting the right information. The experience of one organization provides insight into a couple common mistakes. The team there analyzed data they had been collecting for four years, and a resulting statistic was that more injuries occurred in June—something they already knew.
The first mistake the team made is that they captured age groups instead of individuals’ ages. Second, they didn’t normalize the data around active workers, an important factor since there are more workers onsite in June. As a result, they didn’t have the necessary details for AI-driven analytics to determine if certain conditions led to a higher rate of injury per worker.
The lesson is that it’s better to collect more data because the machine-learning component of the AI assistant can scan thousands of data points to find connections people are likely to miss.

Encouraging employee collaboration
Since more data is better, it helps to have all employees collaborate in the collection of data. The key to workers’ participation is letting them use mobile phones to access web applications, which automatically normalize the data. Information employees might contribute includes data about training, work location, hours on the job before taking a break, or other factors that can contribute to safety risk.
More recently, technologies, such as Quick Response (QR) codes, are providing a way to identify individual employees via their mobile phones and effectively act as digital signatures that can be stored and tracked with other information. This can help track employees at risk because, for example, they lack or have outdated training on particular safety measures.

Using an AI assistant
The AI assistants in applications are essentially cognitive engines that train themselves on how to learn, reason, communicate, and make decisions on their own over time. As a result, they learn from the safety and HR professionals using them and improve their responses over time.
Still, it will be necessary to assist in training the system, much like training a new human assistant. Even in the initial training process, it’s possible to see early gains.
For instance, a manager may want to understand how many near-miss reports were generated in the past month. Getting the answer from the AI assistant may take three simple commands, including one clarification on the time frame:
How many near-miss reports were submitted last month?
Look at the last 30 days.
Email me the summary.
At that point, the manager has just used the AI assistant to build a report on near-miss reporting in a few minutes with three lines of communication instead of the days that such a report can take to generate manually.
AI assistants are a potential game changer in reshaping how companies maximize employee safety. By collecting the right data around safety performance today, safety and HR managers will be able to take advantage of AI assistants tomorrow.

Ryan Quiring is co-founder and CEO of SafetyTek Software. He brings more than a decade of experience as a senior automation consultant and functional safety engineer working on massive capital projects globally in the scope of process automation.

AUGUST 2021

Azure, Line, Font, Text, Blue

VOL. 55 NO. 8