At Productivity Matters, we have observed the rapid rise of AI automation in industry, either via the integration of robots, for example, in manufacturing, the use of exoskeletons in highly repetitive tasks or those involving high force, AI-powered lifting devices to reduce the overall force required of workers and in the assessment of manual tasks. There is still ongoing debate about whether these technologies truly improve safety or whether they are creating new risks, such as an overreliance on automation or musculoskeletal complaints from improper use.
We provide the following list of AI and automation examples that currently exist in various industries:
- Warehousing & Logistics
- AI-driven robotic pickers reduce the need for workers to repeatedly bend and lift heavy items.
- Wearable exoskeletons help employees lift with less strain, but some worry about the long-term ergonomic effects of their use. Are they also reducing the musculoskeletal fitness of workers, placing them at higher risk of tasks that are non-exo-skeleton assisted?
- Healthcare
- Patient lifting devices reducing back strain for nurses and caregivers. These are usually introduced with training.
- AI-powered mobility aids (like robotic patient hoists) are being introduced but can malfunction or require extra supervision.
- Magnetic dishwashing conveyor systems that remove cutlery from patient trays in food services areas of Hospitals.
- Waste maceration units in hospital kitchens, where food waste is converted to liquid form, pumped into a storage tank, and pumped into tanks on trucks to be two local pig farms.
- Manufacturing
- Cobots (collaborative robots) assist workers in repetitive lifting but require safe integration.
- AI-driven ergonomics assessments track movements and suggest safer handling techniques.
In our experience, not all AI and automation integrations have gone well. Successfully integrating any of the above examples requires careful planning and extensive consultation with the workforce.
Those who have integrated AI and automated some manual tasks have done so successfully by following these steps:
- Conducting a Risk Assessment Before Implementation
- Identify tasks that can be automated without creating new hazards.
- Evaluate whether AI or robotic systems introduce new ergonomic risks (e.g., reliance on awkward postures).
- Involve end users in the identification of possible risk controls. Sometimes their suggestions can deplete the need for automation or AI intervention.
- Implement Gradual Adoption & Testing
- Pilot new technologies in small-scale tests before full implementation.
- Gather worker feedback on ease of use and potential safety concerns.
- Pilot new technologies for days and weeks (not minutes and hours).
- Provide Accessible Comprehensive Training
- Workers should understand how to safely interact with AI-powered systems and lifting aids. Training should be tailored to meet the needs of a workforce with Non-English-Speaking Backgrounds.
- Train employees on manual handling best practices, even when using automation.
- Maintain a Human-Centric Approach
- Ensure workers have a say in how automation is integrated into their tasks.
- Measure the acceptance of AI/ Automation
- Integrate learnings from the introduction of AI and Automation at the design stage of every change.
The one area we are a little critical of use of AI is in the use of Wearable sensors (technology) or AI movement detection technologies to assist in risk assessments of manual tasks. It is our view that it might have a place, BUT, there are a few disadvantages of use:
- Although wearable technologies claim to give real time feedback to users that a task they are completing is using a particular muscle group or posture to excess, there is no evidence to support in the long term that they will change their approach to the task after receiving feedback.
- Wearable tech costs a lot.
- Wearable sensor kits are needed to collect data.
- Logistics of delivery and safe keeping of sensor kits is difficult.
- Dependent on a good Wi-Fi connection, if you don’t have this you can lose data.
- The sensors can look and feel obtrusive in a work setting.
- Some workers have refused to wear sensors (they are not keen on their employer ‘tracking’ their every movement).
- Whilst they may produce a plethora of data on how the user was working, they don’t necessarily provide a risk score.
- In the absence of training on how to then use the data to identify higher order (more effective) risk controls that design out the hazardous component of the manual task, they are useless.
Ultimately, while AI and automation offer exciting possibilities for reducing body stressing injuries, their true value lies in how thoughtfully and collaboratively they are implemented—keeping people, not just technology, at the centre of every solution.
