Digital transformation is accelerating across the manufacturing sector. When manufacturers seek to turn technology investment into business value, conversations tend to focus on productivity, automation, and cost reduction — all areas where technology can make a real business impact. But another consequential opportunity is emerging as artificial intelligence systems move from digital workflows into physical environments: improving workplace safety and reducing injuries.
According to data from the National Safety Council, forklifts were the source of 84 work-related deaths in 2024 and 25,110 injuries that required time off from work in 2023-2024. Yet safety programs often remain reactive, relying on audits, manual oversight and post-incident analysis. Safety compliance remains a persistent challenge on factory floors, as traditional programs cannot always address risks in real time.
Edge AI enables a different model. Rather than relying solely on cloud computing, edge AI involves deploying AI directly on devices like sensors, bringing data processing closer to where work is happening to reduce latency and enable responses within milliseconds. By supporting continuous monitoring of industrial environments, edge AI provides greater visibility into how safety protocols are followed in practice, allowing for more immediate action when risks emerge.
From Experimentation to Real-World Impact
Edge AI creates an opportunity to move from reactive safety management to proactive risk reduction.
For example, safety systems enabled by edge AI on a high-speed production line can automatically shut down equipment when a fault occurs or a worker gets too close to active machinery. Through data from cameras, sensors or wearable devices that continuously monitor conditions, edge AI models can compare current activity against predefined safety thresholds to detect anomalies or unsafe proximity. When a risk is identified, the system can immediately trigger a response — such as stopping equipment — without needing to send data to a centralized system for processing.
A similar approach can be applied to handling chemicals, which requires strict adherence to safety protocols. Edge AI systems can identify when required procedures are not being followed — such as improper use of protective equipment, unsafe proximity to exposure zones, or missed safety steps. They can then trigger immediate alerts or interventions that significantly reduce the likelihood of incidents.
The local data-processing required to act within milliseconds is worth prioritizing when even brief delays can increase the risk of injury. In contrast, use cases around predictive maintenance or quality analysis may not carry the same safety implications, or require the same processing needs. These tradeoffs also extend to broader system design decisions. A hybrid approach enables manufacturers to run safety-critical applications on the factory floor while still leveraging centralized systems for deeper analysis and continuous improvement.
The Role of Governance and Oversight
To improve safety outcomes, manufacturing leaders need to align edge AI deployments with operational requirements from the beginning. This includes evaluating connectivity constraints, determining where data should be processed, and prioritizing use cases where a faster response can improve safety. But even with the right technical foundation, outcomes depend on how systems are implemented and managed.
Policy-as-code capabilities — which translate safety requirements and regulatory standards into machine-readable rules that are continuously enforced — can help these systems operate in a controlled, consistent way without slowing down responses in dynamic environments. This approach strengthens governance by making policies visible, explainable and auditable, creating a more resilient foundation for AI-enabled safety.
Human oversight also continues to be essential. In practice, this may involve reviewing real-time safety violation alerts — such as when a forklift operator is not following required safety protocols, including seatbelt use or maintaining safe distances from workers — and determining whether to pause operations, escalate the issue or override the alert if the violation has already been addressed. This level of oversight ensures AI-driven insights are applied in the correct context, helping organizations balance automation with accountability.
For manufacturers, the move from compliance-focused approaches to more proactive safety practices represents a meaningful shift. Edge AI has the potential to improve not only how work is performed, but how risks are identified and managed. Organizations that focus on aligning technology with thoughtful deployment, strong governance and safety objectives will be better positioned to reduce incidents and support more resilient operations over time.
Stefan Blache is practice leader for the manufacturing, communications and energy markets, Kyndryl.
