AI, robotics and the new
risk management conversation
In many cases, automation does not remove risk. It changes where the risk shows up.
By Randy Boss, CRM, CIC, MWCA
Most employers are under pressure to produce more with fewer people, improve turn-around time, maintain quality, and manage documentation, hiring, scheduling, safety training, and compliance. That pressure is one reason that automation is moving so quickly into the workplace.
For many employers, automation is already here: robotic arms, warehouse equipment, software that screens job applicants, and artificial intelligence built into tools that employees and supervisors already use.
The opportunity is real. Automation can remove employees from dangerous tasks, reduce repetitive strain, improve consistency, and identify problems earlier. But as with most risk management issues, the danger lies in assuming that the tool has eliminated the risk.
In many cases, automation does not remove risk. It changes where the risk shows up.
That is where insurance advisors can bring value. We do not need to be robotics engineers or AI programmers. But we do need to help clients ask better questions before technology goes live. Who owns oversight? What takes place when the system is wrong? How are employees trained? What policies apply?
Those are advisor questions.
Safety exposures: New tools, familiar principles
The equipment may look advanced, but many safety fundamentals are familiar: machine guarding, lockout/tagout, employee training, maintenance procedures, hazard assessments, and supervisor accountability.
A robot may reduce the need for an employee to perform a dangerous task, but it can also create a new danger zone. Employees may enter that zone during setup, troubleshooting, cleaning, inspection, or maintenance. Temporary employees, new hires, and contractors may not understand the equipment as well as the people who installed it.
A recent OSHA accident summary gives us a sobering example. In December 2024, an 18-year-old employee was cleaning a sensor inside a robotic enclosure. When he entered the area, the entrance sensors stopped the robot. While he was working inside the enclosure, a co-worker reset the machine from the control panel, not knowing that the employee was still inside. The robotic arm activated, struck the employee, and the employee died from multiple injuries.
That tragedy reinforces an important point: Safety sensors, interlocked gates, warning lights, and control panels are not a true lockout/tagout procedure. When employees enter the operating envelope of robotic equipment for cleaning, setup, maintenance, troubleshooting, or adjustment, the employer needs a clear procedure that controls hazardous energy, verifies isolation, and prevents another employee from restarting the system.
I have seen this same issue for years. The written procedure looks good. The real-life practice is different. Automation can make that gap more serious because equipment may move faster, carry more force, or react in ways employees do not expect.
Advisors should encourage clients to treat automation projects like risk management projects, not just capital investments. Before equipment is installed or software is launched, the employer should complete a practical hazard review. What tasks are changing? Who will interact with the system? What happens during a jam, breakdown, emergency stop, or power loss? Are guards, barriers, sensors, and shutdown procedures adequate? Has maintenance been trained? The goal is not to slow progress. The goal is to keep progress from outrunning prevention.
Employment practices: AI does not remove employer responsibility
Automation is also changing employment practices. Many employers use tools that help with recruiting, hiring, scheduling, performance tracking, productivity measurement, and disciplinary decisions.
These tools may seem objective because they rely on data. But data does not automatically make a decision fair, accurate, or legally defensible. If an AI tool screens résumés, ranks applicants, measures productivity, recommends discipline, or helps decide who gets promoted, the employer still owns the decision. A vendor may have built the system, but the employer is using it in the workplace.
That matters for employment practices liability.
An algorithm can unintentionally create a disparate impact. A productivity tool can punish employees for things outside their control. A scheduling system can create wage-and-hour problems. A poorly explained system can damage trust, especially if employees think they are being watched but not heard.
Clients need AI-use policies that are clear and practical. These policies should define which tools may be used, who may approve them, what information may be entered, how decisions will be reviewed, and when human judgment is required. Employers should also avoid allowing managers to adopt AI tools without review from leadership, HR, legal, IT, and risk management.
A simple rule can help: AI may assist a decision, but it should not own the decision.
Liability: When automation fails, questions follow
Automation also creates liability questions. If a robotic system injures an employee, is the issue employer training, equipment design, maintenance, guarding, or misuse? If an AI-generated work product is wrong, misleading, or discriminatory, who reviewed it? If a chatbot gives an employee bad HR guidance, who approved the script?
These questions can involve several lines of coverage: workers compensation, general liability, products liability, cyber, technology errors and omissions, employment practices liability, directors and officers, fiduciary liability, and crime.
Employers often do not consider these connections when purchasing a new system. The decision may be made by operations, finance, HR, or IT. The insurance advisor may not know about it until after the system is in place.
That is backward.
Advisors should work to get technology and automation questions into the renewal conversation. Ask what has changed in the operation. Ask about new equipment, new software, new vendors, new monitoring tools, new uses of AI, and new data being collected. Ask whether contracts with vendors include indemnification, insurance requirements, limitation of liability, data protection language, and service standards.
This does not mean that the agent becomes the attorney. It means that the agent helps the client recognize when a risk has moved outside the insurance policy and into operations, contracts, HR, or leadership.
Job redesign: The human side of automation
One of the most overlooked automation risks is job redesign. When technology enters the workplace, jobs change. Some tasks go away. Other tasks become more technical. Employees may move from doing the work to monitoring the work. Supervisors may be expected to manage systems they do not fully understand.
Good employers will communicate early. They will explain why automation is being introduced, how jobs will change, what training will be provided, and how employees can raise concerns. Then they follow through.
From a risk management standpoint, this is prevention. Employees who understand the change are more likely to follow procedures, report problems, and trust the process. Employees who feel blindsided are more likely to resist, work around the system, or withdraw.
The people closest to the work often see risks that leadership misses.
Prevention first
Automation can improve safety, productivity, quality, and decision-making. But it should not be treated as a shortcut around risk management. The response is to pair automation with review, training, and control.
For insurance advisors, this is a chance to move beyond quoting coverage and become part of the client’s prevention strategy. We can help clients see that AI, robotics, and automation are not just technology decisions. They are safety, employment, liability, and leadership decisions.
The future workplace will have more automation. The real question is whether employers will put enough human decision-making around it.
That is why good advisors still matter.
The author
Randall Boss, CRM, CIC, MWCA, is a Certified Risk Manager and insurance advisor at Highstreet Ottawa Kent with more than 40 years of experience helping employers improve their cost of risk through prevention and operational discipline. He is also co-founder of Emerge Apps (emergeapps.com), developers of software tools designed to help insurance advisors move beyond quoting to become true risk advisors.
