POLICY BRIEF

AUTOMATING PHYSICAL TASKS USING AI-BASED SYSTEMS

IN THE WORKPLACE: CASES AND RECOMMENDATIONS

Advanced robotics in the workplace

Advanced robotics are becoming increasingly present in today’s world of work. Industrial robot sales
increased by 31% in 2021, compared to 2020. 1 Worldwide sales of professional service robots grew by
37%. 2 This growth is not limited to individual sectors. Sales for medical robots, including surgery robots, robots
for rehabilitation and non-invasive therapy, and robots for diagnostics, increased by 23%. Eighty-five per cent
more hospitality robots were sold in 2021 compared to 2020. Demand for robots in agriculture (6%), inspection
and maintenance (21%), cleaning (31%) and logistics (45%) increased as well. While some of these robots
are systems operating independently from humans, an increasing number is not only capable of some form of
interaction but specifically designed for it (for example, healthcare robots). The International Federation of
Robotics (IFR) defines collaborative industrial robots as ‘a class of robots that perform tasks in collaboration
with workers in industrial settings’. 3 The shorthand ‘cobot’, however, is often used on a wider variety of
systems. Some experts discuss three types of human–robot interaction (HRI). 4 The first type of HRI is
called ‘coexistence’, where a human and a robot share a workspace for a limited time, without sharing a
common task goal. The occurrence of a nurse passing a mail delivery robot in the hallway would be described
as coexistence. The second type is a ‘cooperative’ robotic system that works towards a shared goal with the
human worker. However, their tasks can be independent from each other. Pick-and-Place robots at a
workstation that prepare parts for human workers reflect this type of interaction. A collaborative HRI is
represented by the human and robot working towards a common goal and additionally their tasks and subtasks
are shared in time and place. Furthermore, human–robot collaboration is indicated by the creation and use of
synergies. 5 An example is lifting a heavy object collaboratively. All three scenarios include advanced
robotics, which can be described as cobots, capable of perceiving and reacting to their surroundings.
Some of these systems rely on a complex but deterministic backend software to perform their tasks, while
others use AI-based systems.
Figure 1: Three types of human–robot interaction
When looking at current case studies,human– by Onnasch and Roesler
robot collaborations in the ways described by
Onnasch and Roesler 6 are rare to find at the
workplace. Cooperative scenarios are the most
common. However, as there is rapid growth in
all sectors of robotics application, this
distribution might change in the future. As the
technology continues to expand into more and
more workplaces and unstructured
environments, companies might face difficulties
and challenges during the implementation
process. To reduce these barriers, one can
consult case studies that have already
successfully implemented advanced robotics in
their workplaces.

1
International Federation of Robotics. (2022). Executive Summary World Robotics 2022 - Industrial Robots. https://ifr.org/free-downloads
2
Ibid.
3
International Federation of Robotics. (2019, January 24). IFR Secretariat Blog. https://ifr.org/post/international-federation-of-robotics-
publishes-collaborative-industrial-rob
4
Onnasch, L., & Roesler, E. (2019). Anthropomorphizing robots: The effect of framing in human-robot collaboration. Proceedings of the
Human Factors and Ergonomics Society Annual Meeting, 63(1), 1311-1315. https://doi.org/10.21934/baua:fokus20160630
5
International Federation of Robotics. (2019, January 24). IFR Secretariat Blog. https://ifr.org/post/international-federation-of-robotics-
publishes-collaborative-industrial-rob
6
Onnasch, L., & Roesler, E. (2019). Anthropomorphizing robots: The effect of framing in human-robot collaboration. In Proceedings of
the human factors and ergonomics Society annual meeting, 63(1), 1311-1315. Sage CA: Los Angeles, CA: SAGE Publications.
https://doi.org/10.21934/baua:fokus20160630

1
 Automating physical tasks using AI-based systems in the workplace:
cases and recommendations

As part of EU-OSHA’s research on advanced robotic and AI-based systems for the automation of tasks and
occupational safety and health (OSH), 11 case studies and 5 short case studies were developed that focus on
workplaces that use these technologies. The following presents three summarised case studies from across
Europe based on companies of varying sizes that use advanced robotics with varying degrees of automation.
They each automated a different task with different systems.

Case studies
Case 1
The Slovenian-founded company for Case 1 operates on a global scale, in the field of automotive and industrial
technologies. They currently have locations in over 55 countries. Within Slovenia, they employ around 1,700
workers in their production sites. They use a six-axis, one-arm, medium-sized collaborative robotic system
from a third-party supplier.
The cobot in this case study handles parts weighing between 2.5 kg and 3.5 kg at an estimated pace of
600-700 times per day. Workers, as part of the production and quality control process, previously handled
these parts. The cobot lifts a product part from the production line and moves it to a storage pallet. This
process is interjected by a worker, who performs a quality control check on the presented part. The cobot
performs an object-related repetitive task with physical manipulation of the aforementioned 2.5 kg and
3.5 kg parts. For the task of lifting the parts, human labour is substituted and the workers’ focus shifts towards
the cognitive task of quality control.
The decision that a collaborative robotic system will be introduced to the worksite in this first case study was
made by the company’s management, in cooperation with the technology department. The first testing phase
was initialised. This included a risk assessment and extensive safety testing. After the safety-related
certification and approvals were acquired, and in consideration of safety regulations and requirements,
workers were included to get to know the system before using it in everyday operations. Before starting the
production with the cobot, there was also an education period for the operators.

Case 2
The company in this case is an automotive supplier operating on a global scale and specialising in the field
of drive and chassis technology. They provide customised integrated solutions for automobile manufacturers,
mobility providers, and other companies with a focus on transportation and mobility. This case study is located
in one of their factories in Portugal. In total, the company has more than 100 locations in over 30 countries,
with around 150,000 employees worldwide.
This factory deploys a cobot to automate the sewing of bags. The cobot performs the sewing task. It is a
one-armed multi-axis robot from a third-party supplier that was customised to fit the factory’s needs. The
worker supplies the cobot with the necessary material to perform the sewing task. Once the sewing is
completed, the cobot performs an initial quality control check to assess if all seams are of sufficient quality.
The worker can then start a new sewing cycle for the cobot while they perform further quality control checks
and assembly steps on the bag. The cobot performs physical object-related tasks with a non-AI backend
software that is highly repetitive.
Before introducing the cobot, the company assessed their production line for possible automated
processes. Worker involvement came after the decision had been made regarding which tasks to automate
and which technology to use. Parties involved in these decisions were the system developers and an
engineering team. The introduction took place in two phases. Prior to using the cobot, workers received
specialised training. At first, both the cobot and a manual machine were available to perform the task. After a
week, the manual machine was removed. Then, workers started using the cobot full time. The impact the
cobot’s performance had on production cycles has allowed the company to restructure their workers’
shifts. Instead of working at one station for eight hours, they now rotate every two hours. The company
reported initial hesitancy of workers to operate the cobot, due to subjective safety concerns. As their workers
gained experience with the technology, the concerns resolved.

Case 3
The third company is active in a variety of sectors like automation and digitalisation in industry, infrastructure
for buildings and decentralised energy systems. Today it can be classified as an international company with
branches in 190 countries, however, it was founded in Germany.

2
 Automating physical tasks using AI-based systems in the workplace:
cases and recommendations

Their cobot systems automate tasks through six-axis articulated robots or four-axis scale robots. Within
their locations, there are both self-designed robotic applications as well as systems bought from third
parties. Cobots are used in slow-cycle assembly processes for a variety of product parts. The cobot
typically provides physical support by holding the workpiece. The cobot’s function is to alleviate workers’
physical strain, and assist them in their main task, by reducing physical load. At these cobot workspaces,
workers still generally perform their previous task, or take on additional skilled tasks in the assembly
process. The cobots perform an object-related repetitive task with physical manipulation of the workpiece.
The company noted that automation through advanced robotics has, in some cases, led to task consolidation
for workers. While in older factory set-ups the workers completed all relevant tasks concerning one workpiece,
they now perform more and more ‘side tasks’.
The company explicitly encourages workers to bring forward ideas and suggestions for further process
automation. Workers can express the wish for a robotic system at their workstation to an automation specialist
(‘key user’), who then makes an initial feasibility assessment. Approved suggestions are forwarded and a
project team is created, including a project planner, workers council and data protection officers on site, and
advisory security staff. When the impulse does not come from a worker, their involvement typically starts after
a concept phase. However, this does not end with the final roll-out of the robotic solution but continues into the
ongoing production. Workers are trained to optimise production, hence, there is a continuous open
feedback system to submit suggestions for changes, optimisation or innovation.
The most influential factor on trust was the early involvement of workers during the introduction process.
In the company’s experience, this increases acceptance towards the systems. Additionally, providing
information early on and communicating clearly about both the intention of the automation and its
practical functions is vital.

Recommendations
Advanced robotics are becoming more diverse, and more companies in Europe are using these technologies.
Hence, they encounter different structural and technological hurdles along the way. The implementation
process can be just as diverse, with different strategies being the most efficient for different companies. Which
parties are involved and to what extent, not only depends on the company size, but also on their level(s) of
experience with advanced technological systems. While some companies develop in-house solutions, others
use third-party suppliers and customise those robots to their needs. It is therefore not goal-oriented to try and
create a standard list of implementation steps to follow for every company. It can be more helpful to gather
factors that helped companies along the implementation process that are not limited to their company size,
sector or level of experience.
Questions regarding the introduction process and the assessment of worker attitudes towards robotics
systems need to be asked to facilitate a successful introduction and human-centred long-term working
conditions. All case studies acknowledged the workers’ fear of job loss due to continuous automation. The
case studies have developed ways to address these concerns with their workers and successfully reduce fear
of job loss. It can be approached from two sides: first by intervention, and second by prevention. When a
company becomes aware that their workers experience fear of job loss due to automation through advanced
robotics or AI-based systems in their workplace, certain interventions can be made. Most companies decide
to improve their workers’ understanding of the technology and the implications automation has for
them and their work. All companies in the case studies stressed that the intention of their automation is not
to eliminate jobs but to improve working conditions. Educating workers on how the technology impacts
and benefits them, their routine, and occupational safety and health is a separate intervention from training
them on how to use the system. Workers have begun to recognise that being technologically literate is essential
in an increasingly digital workplace. 7 Providing them with the skill sets needed in the long term could reduce
their adjustment period to a new digital work environment and provide them with a subjective sense of job
security. 8 Preventing subjective fear of job loss entirely is likely to be impossible. However, companies can
take preventive measures independently from the introduction of a singular system. Firstly, in cases where
workers were the initiator of the automation process, acceptance was higher. This falls under the larger
umbrella of worker involvement in the implementation process. Early worker involvement provides them with

7
Smith, C. L. (2015). Technology literacy skills needed in further education and/or work: A Delphi study of high school graduates’
perspectives (Publication No. 5776) [Doctoral dissertation, University of South Florida]. USF Tampa Graduate Theses and Dissertations.
http://scholarcommons.usf.edu/etd/5776
8
Kozak, M., Kozak, S., Kozakova, A., & Martinak, D. (2020). Is fear of robots stealing jobs haunting European workers? A multilevel study
of automation insecurity in the EU. IFAC-PapersOnLine, 53(2), 17493-17498. https://doi.org/10.1016/j.ifacol.2020.12.2160

3
 Automating physical tasks using AI-based systems in the workplace:
cases and recommendations

the opportunity to exercise some form of influence and to communicate their wishes, needs and concerns early
on. While not all companies may have the opportunity to involve workers in the design process of a new
system, informing them about upcoming changes as early as possible, and creating a way for them to voice
concerns, is good practice and might reduce fear of job loss along the way. Having managed the subjective
fear of job loss for the introduction of one kind of robotic or AI-based system does not mean that a company
should neglect this topic in any other/future automation projects.
Early worker involvement goes hand in hand with a functional communication strategy. Empirical research
supports the companies’ experience that having a formal communication avenue while introducing a change
initiative reduced uncertainty and enhanced commitment. 9,10 Communicating future changes to workers can
reduce feelings of uncertainty towards the rationale behind the change. Furthermore, clear and direct
communication has been found to promote change-supportive behaviour from workers. 11 These findings in the
literature were reconfirmed in the experiences outlined in the case studies. Having both personal (for example,
team lead) and anonymous (for example, feedback box, or workers council representative) communication
systems has been described as helpful to receive worker feedback and create conversation around relevant
topics.
The relative novelty of advanced robotics in the workplace is accompanied by a workforce inexperienced in
how to interact with them. A lack of familiarity can influence workers’ attitudes towards robotic systems and
colour the initial experience. Initial attitudes are perhaps even more so shaped by external sources like
newspaper reports, which can be negatively biased regarding robotic systems at the workplace. 12 In one case
study, the company specifically stressed how crucial it is to differentiate between fictional representation of
robotic systems and the actual technology. Researchers found that negative attitudes towards robots
decreased as experiences of interacting with them increased. 13 This is confirmed by companies, like in
the second case study that faced initial hesitance towards the system but saw a significant reduction once
workers gained experience with the system. Companies interested in reducing this initial hesitancy could start
by offering early education on robotic systems before implementing them. Furthermore, to reduce unfamiliarity
in the interaction, systems designers should orient themselves with established interaction design principles,
one of them being the EN ISO 9241-110. This standard contains seven interaction principles for human–
technology interaction called ‘Suitability for the user’s tasks’, ‘Self-descriptiveness’, ‘Conformity with user
expectations’, ‘Learnability’, ‘Controllability’, ‘Use error robustness’ and ‘User engagement’. They can be used
to evaluate HRI.
The fact that the world of work is rapidly changing, in part due to advancements in robotic systems, also
increases the expectation that workers adapt and change with it. Working with advanced robotics often
requires a new skill set and more detailed understanding of technology than most workers’ previous tasks did.
Some case studies have described difficulties of workers in keeping up with this demand for cognitive change.
New work environments can challenge workers more cognitively, such as the requirement of increased
decision-making, in contrast to the mostly physical tasks they were previously performing. 14 For this reason,
cognitive and sensorial aids should be provided to workers to prevent information overload and
potential negative effects on the operators.
Lastly, several companies within the project reported a change in task structure, in relation to the
implementation of a robotic system. This is mostly described as a positive change, as workers now have more
time to perform their main task or are described as using the additional time to help other co-workers or to
perform secondary tasks. Task design needs to be considered during the implementation process. Other
companies raised concerns that workers now possibly perform more disjointed tasks, decreasing task
completeness. Others raised the concern that automation can lead to task consolidation and work

9
Bordia, P., Hobman, E., Jones, E., Gallois, C., & Callan, V. J. (2004). Uncertainty during organizational change: Types, consequences,
and management strategies. Journal of Business and Psychology, 18, 507-532. https://doi.org/10.1023/B:JOBU.0000028449.99127.f7
10 Hobman, E. V., Bordia, P., & Gallois, C. (2004). Perceived dissimilarity and work group involvement: The moderating effects of group
openness to diversity. Group & Organization Management, 29(5), 560-587. https://doi.org/10.1177/1059601103254269
11
Kozak, M., Kozak, S., Kozakova, A., & Martinak, D. (2020). Is fear of robots stealing jobs haunting European workers? A multilevel
study of automation insecurity in the EU. IFAC-PapersOnLine, 53(2), 17493-17498. https://doi.org/10.1016/j.ifacol.2020.12.2160
12
Riemer, J., & Wischniewski, S. (2019). Robotics at work - News headline analysis 2016. In 2019 IEEE International Conference on
Advanced Robotics and its Social Impacts (ARSO) (pp. 202-207). IEEE. https://doi.org/10.1109/ARSO46408.2019.8948759
13 Nomura, T., Suzuki, T., Kanda, T., Yamada, S., & Kato, K. (2011). Attitudes toward robots and factors influencing them. In K.
Dautenhahn & J. Saunders (Eds), New frontiers in human–robot interaction (pp. 73-88). John Benjamins Publishing Company.
https://benjamins.com/catalog/ais.2.06nom
14
Gualtieri, L., Rauch, E., Vidoni, R., & Matt, D. T. (2020). Safety, ergonomics and efficiency in human-robot collaborative assembly:
Design guidelines and requirements. Procedia CIRP, 91, 367-372. https://doi.org/10.1016/j.procir.2020.02.188

4
 Automating physical tasks using AI-based systems in the workplace:
cases and recommendations

intensification. This risk is also described in the literature relating to task design. 15 So, before implementing
an advanced robotics system, companies need to consider what job content and tasks are left for the human
worker. Tight technological coupling should be avoided, and the human should still perform meaningful tasks,
not only tasks that cannot be currently automated.

Figure 2: Recommendations for successful introduction

Internal
Early worker Increase of
Task design communication
involvement familiarity
and support

Technology-based factors, like choosing the best-suited robotic system for the task one wants to automate,
are highly important for the successful implementation of advanced robotics. However, it is within their
workforce and internal structures where companies report the most important steps to take to facilitate a
successful, long-term implementation. Figure 2 presents the four most often named antecedents, from the
companies’ perspective, for the long-term successful implementation of advanced robotics. In their collective
experience, human-centred design and communication reduce or prevent hurdles along the way.

Authors: Eva Heinold, Federal Institute for Occupational Safety and Health (BAuA), Patricia Helen Rosen,
Federal Institute for Occupational Safety and Health (BAuA), Linus Sioland, Milieu Law & Policy Consulting
Brussels, Dr Sascha Wischniewski, Federal Institute for Occupational Safety and Health (BAuA).
Project management: Ioannis Anyfantis, Annick Starren (EU-OSHA).

This policy brief was commissioned by the European Agency for Safety and Health at Work (EU-OSHA).
Its contents, including any opinions and/or conclusions expressed, are those of the authors alone and do not
necessarily reflect the views of EU-OSHA.

European Agency for Safety and Health at Work, EU-OSHA, 2023

©Reproduction is authorised provided the source is acknowledged

15Karasek, R. (1979). Job demands, job decision latitude, and mental strain: Implications for job redesign. Administrative Science
Quarterly, 24(2), 285-308. https://doi.org/10.2307/2392498