An exoskeleton moves the load, the question is where
Occupational exos don't erase strain, they redistribute it. Measure the transfer, not comfort.
The pitch for occupational exoskeletons is intuitive: strap on assistance, and the strain of lifting or overhead work goes away. But strain doesn’t vanish, physics won’t allow it. A passive device that unloads the shoulders is routing that force somewhere; one that supports the low back during a lift redirects force into the body at a new point of contact. The question was never “does it feel easier.” It’s “easier where, and harder where else.”
Relief is not elimination
NIOSH has been explicit: some devices may transfer load between musculoskeletal regions in ways that still put workers at risk, a hip-supported device can pull load off the arms while increasing load carried by the spine and lower extremities. That’s the hazard hiding inside a comfortable product: relieve the region you were measuring, quietly overload one you weren’t. NIOSH’s earlier review of wearable load-reducing devices makes the same caution, benefit at one joint must be weighed against cost at another. This is why comfort surveys are the wrong evidence: workers can prefer a device that reduces perceived shoulder effort while raising compressive load on the lumbar spine, the very load ISO 11228-1 on lifting and carrying exists to bound. Preference tracks felt exertion, not tissue loading; injury follows the second.
The test
Before crediting an exoskeleton with reducing MSD risk, ask: has anyone measured the load it adds to the regions it transfers force into, and does the net biomechanical picture improve, not just the assisted joint? If your evidence is a satisfaction score or a single-joint measurement, you haven't tested the actual question.
There’s a real place for these devices, for sustained overhead or repetitive lifting, a well-matched exo may genuinely lower a specific exposure. But treat it like any engineering control: characterise the hazard, apply the device, re-measure the whole body against a manual-handling reference. A control that just relocates the risk hasn’t reduced anything.