Artificial intelligence (AI) and data analytics are rapidly changing the way warehouses work. Thanks to scanners, sensors and smart systems, companies know exactly who picked which product and when, and how much effort it took. This offers opportunities for more efficiency and better tuned work processes. At the same time, the use of personal data raises questions about privacy, autonomy and well-being. How can AI be deployed in a way that not only increases productivity but also keeps people central?
AI finds its way into the warehouse
An increasing number of articles are appearing in the scientific literature on the use of personal data and AI (or other optimisation algorithms) when deploying people in the warehouse.
AI in parts-to-picker systems
AI and optimisation algorithms can be applied, for example, to parts-to-picker order picking systems. In such a system, a shuttle, crane or robot retrieves a product load carrier from storage and brings it to a picking station. There, a human must remove the product from the load carrier and place it in a customer tray. The system keeps track of everything. Which product is picked exactly when, from which location, and where is it put away when.
Can AI help improve that picking process? The answer is: yes. The system knows exactly how fast someone picks a certain product, how much energy that person consumed, how many metres that person walks, how often that person has to bend or stretch, or how much injury risk a certain movement adds to that person. That information can be used for various purposes. For example, to increase productivity (the person who can handle certain products faster is assigned the relevant order), or to minimise a picker's energy consumption so that the picker gets less tired, or to minimise injury risk. Research has been published on each of these aspects.
AI in picker-to-parts systems
AI and optimisation algorithms can also be used in picker-to-parts systems. In such a system, the collector walks or drives an order-pick cart through the warehouse and collects items in a route. In recent research, based on data from a Finnish retail warehouse, it was found that if the right order (taking into account order properties such as the number of items needed, pick height, walking distance, volume and weight) was assigned to the right picker (taking into account skills of that picker related to those order properties) the total picking time needed per day could be reduced by almost 10%.
More than productivity: also well-being as a goal
Although most research looks at improving productivity, other goals are also possible. The algorithm can also balance system performance with an individual's well-being or preference when making the allocation decision.
Available data in modern warehouses
The necessary data is automatically collected and recorded in many warehouses. After all, everything is scanned, confirmed and logged and linked to the individual. The data is collected by the WMS, or with the help of cameras, or with sensors attached to the warehouse truck, or even directly on the person (think tags being carried, or wearables such as smartwatches).
So the data is there. The algorithms, at least in the scientific literature, are reasonably well known. However, that does not mean that they are already widely used.
Constraints: regulation and ethics
In practice, the AVG directives and the new EU AI directive, as well as requirements from trade unions and works councils, are barriers to using data and algorithms in business processes against the interests of employees. When companies do this anyway, it often leads to negative publicity. Amazon was fined in France last year for excessive tracking and misuse of personal data, e.g. toilet visits (see Amazon fined for 'excessive' surveillance of workers). Outside the EU, however, employees are often less protected and then personal data can be collected and used unhindered. Usually only to improve efficiency. Much less to improve the employee's work and workplace.
Human-centred applications of AI and data
However, personal data can also be used for the latter. For example, by warning of imminent danger after detecting someone's location. Or by measuring load trying to deliberately relieve the employee. Or by including individual work preferences, which could potentially change by the day.
Conclusion: opportunities and responsibilities
There are great opportunities for practitioners. Both process improvement and employee well-being improvement are possible. Also, far from all possibilities for this have been scientifically researched. So there is work to be done. What is important, however, is that implementation takes place transparently and with the consent of the employees.
Want to know more about Smart Warehousing? Then come to ICT&Logistics on 4-6 November 2025 at Jaarbeurs, Utrecht. Register for free here.
