Generative AI, the new colleague? 
– a sociomateriality perspective on the  
human-AI relationship 
 
 
 
 
 
 
Authors: 
Maya Drossel, 2000/04/02 
Felix Löfgren Hallbeck, 1997/08/20 
Supervisor: 
Sara Brorström 
Department: 
Graduate School 
Course: 
GM0861 V24 Master Degree Project in Management 
Graduate School The School of Business, Economics, and Law, University of Gothenburg 
 
Spring 2024  
Generative AI, the new colleague? 
– a sociomateriality perspective on the human-AI relationship 
 
Maya Drossel 
Master of Science in Management, Graduate School The School of Business, Economics, and 
Law, University of Gothenburg 
 
Felix Löfgren Hallbeck 
Master of Science in Management, Graduate School The School of Business, Economics, and 
Law, University of Gothenburg 
 
 
 
Abstract 
With the increasing human-like capabilities of generative AI, management studies have 
acknowledged to reconsider the traditional roles of AI, moving beyond a tool-based view 
towards AI as an active participant in human collaboration. Thereby, a shift towards a relational 
perspective when studying human-AI interaction has been recognized. This study aims to 
overcome the existing scarcity of empirical research on the human-generative AI relationship. 
Using a qualitative approach, 28 semi-structured interviews were conducted with managers in 
the tech field, focusing on their daily practices with generative AI. Applying the theoretical 
framework of sociomateriality enables an understanding of how managers and generative AI 
form a relationship while highlighting changing interactions within four stages of the 
relationship. As the interactions between managers and generative AI become more complex 
the collaboration within the relationship is observable, ultimately leading to interactions of a 
partnership. This research contributes to the field of management studies by observing the 
enactments of human and material practices and by identifying the transitional process from a 
tool perspective into a collaboration perspective in which AI becomes seen as a colleague at 
the final stage. Trust and solidarity remain limiting aspects of the relationship emphasizing the 
need for future research. 
 
Keywords: Generative AI, Collaboration, Sociomateriality, Interaction, Managerial Practices, 
Human-AI Relationship, Affordances and Constraints 
 
 
  
1 
Introduction 
The rise of generative AI compels with its humanlike responses making it almost impossible 
for people to distinguish a 500-word text that is created by generative AI from a human-written 
one (Brown et al., 2020). Beyond this simple text generation, generative AI has become a 
human-like discussion opponent (Korzynski et al., 2023) with its ability to create unique and 
creative content that mirrors human intelligence (Banh & Strobel, 2023). Thereby AI can 
perform both administrative and strategic tasks, entering the full range of managerial practices 
(Korzynski et al., 2023). This creates a new context in the research of management scholars of 
seeing AI as a counterpart of the daily work field of the manager (Anthony et al., 2023; Dwivedi 
et al., 2023; Raisch & Krakowski, 2021) calling into question the ideas of how managers work 
with generative AI, a technology that is much like a human itself. 
Previous research on managerial interactions with AI has focused either on AI as a tool 
to conduct certain tasks, such as diagnosing radiology scans (Lebovitz et al., 2021), or as a 
medium to communicate between different positions in an organization (e.g. Bechky, 2003). 
Yet, recent studies have highlighted that these perspectives are no longer suitable for generative 
AI due to the invisibility, inscrutability, and constant change of AI (Anthony et al., 2023). 
Therefore, new technologies like ChatGPT should be viewed from a counterpart perspective 
enabling to identify AI as an active participant in interactions (Anthony et al., 2023; Bailey et 
al., 2022). Researchers have acknowledged the needed shift towards a more relational research 
approach identifying collaboration to be a suitable term for the interaction (Anthony et al., 
2023; Raisch & Krakowski, 2021; Wang et al., 2023), yet few have done empirical studies on 
human-generative AI collaboration (Chowdhury et al., 2022). This study follows the 
perspective on human-AI collaboration that managers and AI co-exist but also enable one 
another (Chowdhury et al., 2022). At the same time, studies of AI and humans in knowledge-
intense positions, have shown that the opacity of learning algorithms can hinder the adoption 
of AI into their work practices (Faraj et al., 2018). The black box characteristics of AI create 
uncertainty about how AI derives a certain output (Anthony, 2021). As a consequence trust in 
AI is perceived to be lower than in humans both internally by the person interacting with AI 
(Lebovitz et al., 2021) and externally by people observing the interaction (Kellogg et al., 2020; 
Waardenburg et al., 2022). Thereby studies have shown that knowledge workers substitute the 
work of AI by disregarding suggestions from AI and following their human judgment (Lebovitz 
et al., 2022; Waardenburg et al., 2022). This leads to a contrasting view in the literature where 
research calls for collaboration studies of generative AI, yet empirical studies on algorithmic 
learning demonstrate that uncertainties and mistrust can hinder the adoption of such 
technologies. 
The lack of certainty over human-AI collaboration causes us to reconsider its nature by 
trying to understand the relationship in which AI and managers engage and identify whether 
they create a collaboration in the specific case of generative AI. Aiming to study the 
relationship, we draw on the concept of agency from sociomateriality as a framework that 
acknowledges that not only humans but also technologies can become agents of managerial 
practices (Floridi & Sanders, 2004). By focusing on a sociomateriality perspective, this 
research provides a new understanding of how managers and generative AI both can have 
agency in a managerial position by being constantly intertwined with one another (Leonardi, 
2 
2013a). To understand the formation of the relationship the study will follow the concept of 
affordance and constraint (Leonardi, 2011) and the practice lens on emergent agency 
(Orlikowski, 2000). 
Due to the complexity of generative AI and its early stage of organizational 
implementation, we focus on a micro-level perspective investigating the daily actions that can 
shape the construction of technology on a macro-level (Leonardi, 2013a, 2013b) and close the 
empirical gap of generative AI in managerial practice. Managers within the tech field show 
strong expertise and high adaptability to new technologies making them especially suitable as 
study objects (Siltanen, 1993). These findings can provide a guideline for future in-depth 
studies of the organizational embedding of human-generative AI collaboration, guiding 
subsequent research that employs a zooming-in zooming-out approach to the intra-
organizational enactment (Nicolini, 2009). Insights gained from this study will not only 
contribute to academic knowledge but also offer practical guidance for technology managers 
navigating the evolving landscape of generative AI, trying to overcome the practical relevance 
gap of academic papers (Baldridge et al., 2004). The human resemblance of generative AI calls 
for a new approach to human-AI interaction where a relationship is created through human and 
material interactions (Leonardi, 2011). The inscrutability of generative AI makes the 
interactions with generative AI and the manager in an organizational context unclear. 
Therefore, we propose the research question: How are managers and generative AI forming a 
relationship through interactions of everyday work practices? 
The research question aims to understand the human-AI relationship by studying the 
interactions in the case of managers in the tech field and text-to-text generative, making us 
believe that qualitative studies will generate the most valuable data to study this process. We 
propose an initial empirical study with a thematic analysis of generative AI to provide a general 
real-life perspective. By pursuing a qualitative study with 28 semi-structured interviews this 
paper focuses on the interactions on the ground from the human perspective upon which more 
in-depth empirical studies can be conducted in the future.  
This thesis is structured to provide a comprehensive exploration of managers’ 
relationship with generative AI, guided by a sociomaterial perspective. Following this 
introduction, the thesis is organized by a literature review on generative AI, human-AI 
collaboration, responsibility over AI, and sociomateriality. Subsequently, we will explain the 
methodology thoroughly and delve into the analysis of the empirical data. Lastly, in the 
discussion, the research will follow a sociomateriality framework for the findings and finalize 
with a conclusion. 
 
Literature review 
With the rise of generative AI, the distinctions between the capabilities of managers and 
learning algorithm technologies become vague. Generative AI falls under the category of 
algorithmic learning and represents the forefront of technologies that largely mimic the 
cognitive abilities of humans. Algorithmic learning describes technologies that can generate 
responses, predictions, or classifications (Faraj et al., 2018). These tasks require technologies 
to understand a context and perform situational tasks by extracting knowledge from large data 
sets (Duan et al., 2019). Thereby algorithmic learning enters knowledge work, such as the 
3 
critical thinking of managers, that has historically been categorized as unique human domains 
(Faraj et al., 2018), making a clear distinction between pure human and pure technological 
tasks difficult. 
 
Generative AI in managerial practices 
Generative AI possesses the capacity to conduct knowledge work, characterized by its ability 
to create content that is novel and human-like (Banh & Strobel, 2023). This technology is 
primarily based on semi-supervised learning, which requires only a small amount of labeled 
data to facilitate ongoing learning on extensive unlabeled data (Kingma et al., 2014). Thereby 
it learns to create meaningful outputs from the data it is exposed to without any additional form 
of external supervision (Prasad Agrawal, 2023). This research focuses on text-to-text 
generative AI, such as ChatGPT, that can interact with the user through written language.  
These tools are based on an algorithm, that responds with a coherent human-like text based on 
predicting a response to a request of the user (Brown et al., 2020). Consequently, the underlying 
model can create a conversational AI tool, where the user creates a request following a text 
response of the AI application. These initial requests are an interaction technique called 
prompting, a unique generative AI property (Dang et al., 2022; Teubner et al., 2023). The user 
can write a short text called a prompt to engage with the algorithm and generate a text response.  
Korzynski et al. (2023) identify current research trends in categorizing managerial 
applications of generative AI on an administrative and strategic level. On an administrative 
level, generative AI can be used for translation (Brown et al., 2020), answering simple 
questions in research (Dwivedi et al., 2023) but also to automate repetitive tasks such as 
scheduling appointments or writing e-mails (Korzynski et al., 2023). On a strategic level, 
decision-making and knowledge management can be supported through applications such as 
ChatGPT (Korzynski et al., 2023). The benefits in decision-making are based on providing 
recommendations for specific situations. These recommendations can be based on data analysis 
for example in the risk-management field, generative AI can evaluate companies’ public 
financial statements and can provide an assessment of investment uncertainties (Chen et al., 
2023). Additionally, interacting with AI can provide inspiration and feedback on your thought 
process (Dang et al., 2022; Dwivedi et al., 2023). The benefits in knowledge management are 
realized if the companies connect their generative AI model with their internal data system, 
enabling a direct search in non-public knowledge (Davenport & Alavi, 2023). The advantages 
of generative AI can also be found across task levels by improving learning or time 
management. For example, generative AI can step into the more sophisticated role of a coach 
to improve skill development in the management area of diverse tasks (Terblanche & Cilliers, 
2020). In another example, the automation of administrative tasks can free up time for 
managers to focus on more creative tasks (Dwivedi et al., 2023). 
By showing opportunities in managerial tasks, generative AI enters the context of 
managerial practices. Through their human-like performances, a new class of agents in 
organizations is implied (Floridi & Sanders, 2004). Recent research calls for an 
acknowledgment of managers and AI as being both interdependent agents of management 
positions (Raisch & Krakowski, 2021). Until now, AI interactions with managers have been 
oftentimes overlooked by management scholars (Raisch & Krakowski, 2021). However, 
4 
management studies can view human-AI interactions in real-world settings, where the 
technological and social worlds are merging (Orlikowski, 2010), and are therefore able to make 
highly needed contributions to the development of AI.  In interacting with AI, managers shape 
and reshape algorithms and practices (Deng et al., 2017). A balanced analysis of the 
interactions should thereby be the focus of the study.  
 
Human-AI collaboration in managerial practices 
In the information system and management field there exist different hypotheses of how 
humans and AI co-exist. AI can either be viewed as a replacement for employees or a new 
partner for collaboration. Through studies of robotization and computerization researchers are 
reflecting on the potential of AI replacing humans (e.g. Czarniawska & Joerges, 2018). Whilst 
early-stage publications have evaluated the potential of job losses due to robotization (e.g., 
Ford, 2014), most recent perspectives on robotization show that jobs have not disappeared 
significantly since the introduction of AI (Fleming, 2019; Willcocks, 2020). Rather it is argued 
that low-skilled jobs proliferate while high-skilled jobs, such as those of the management, 
become more difficult to acquire as fewer managers are required to supervise work (Dixon et 
al., 2021; Fleming, 2019).  
Recent studies of robotization follow the view that humans will continue to co-exist with 
AI (e.g. Fabri et al., 2023). The concept of how they will co-exist has been a much-debated 
topic within management and information systems concepts, with collaboration at the center 
of its research encompassing three main research perspectives.  
The first perspective examines the task areas amongst which AI and humans will exist 
either conjointly or independently. In the context of managerial practices, tasks within this 
position can be either augmented or automated (Raisch & Krakowski, 2021). As elaborated 
before, AI can support human actors to perform both their administrative and strategic tasks 
within knowledge work. A strong focus has been on the ability of learning algorithms to 
perform repetitive and routine tasks (Frey & Osborne, 2017; Raisch & Krakowski, 2021). 
Whilst administrative tasks are likely to be automated, decision tasks turn into a process of 
augmentation, improving the decision by integrating AI into the process. The possibility of 
augmenting managerial capabilities through AI has been highly discussed in the field of 
decision-making. By working with AI, human skills can be complemented through the 
analytical work of AI (Brynjolfsson & Mitchell, 2017). Thereon researchers have taken the 
stance that the augmentation of managers can be seen as “joining forces” (Wilson & Daughtery, 
2018, p.1) highlighting the collaboration between humans and AI.  
The second perspective concentrates on exploring the interactions between humans and 
AI. Initially emerging in the 1980s, the concept of these interactions focused on humans using 
AI (Suchman, 2006). This focus has evolved into a newer research perspective that neglects 
the wording “usage”. The new narrative of humans and AI develops towards a collaborating 
perspective, in which AI is not acted upon but becomes an individual actor of a partnership 
(Bankins et al., 2023; Chowdhury et al., 2022). Instead of viewing AI simply as a passive tool 
to be used or a medium between humans, AI is viewed as a counterpart that engages in work 
practices (Anthony et al., 2023). Within this collaboration managers and AI are so closely 
5 
intertwined that their actions are determined by an interpretation of the relation (Floridi & 
Taddeo, 2016) and not inherent in the technology (Orlikowski & Scott, 2008). 
The third perspective is focused on relational studies of humans and AI. Technologies 
are viewed as a set of evolving relations within an organization (Bailey et al., 2022), for 
example a constellation of the developer, the manager and the employees (Anthony et al., 
2023). Thereby relational studies on technologies oftentimes focus on the organizational 
context in which AI is embedded, emphasizing that the social construct influences AI’s 
development and interaction (MacKenzie & Wajcman, 2011). The human-AI relationship is 
influenced by individual and organizational factors, yet individual relations with AI remain a 
less focused research domain (Bankins et al., 2023). 
 
Responsibility over managerial practices 
Issues of uncertainty and trust, due to the opaque nature of algorithmic learning in a managerial 
context, have increasingly been the discussion object of research (e.g., Wang et al., 2023). 
Opacity describes the opposite of transparency emphasizing the difficulty of understanding the 
reasons for a particular result because its derivation is unclear (Stohl et al., 2016). Although 
initial research argued that algorithmic learning increases confidence and reliability in AI-
supported decisions, recent writings demonstrate that a lack of understanding and missing trust 
impede the adaptation of AI systems (Bieda, 2020; Glikson & Woolley, 2020). Various reasons 
contribute to the uncertainty and reduced trust in algorithmic learning. Within generative AI, 
data security concerns arise due to the sensibility and confidentiality of data shared in prompts 
with software like ChatGPT and its potential to leak data to third parties (Chen et al., 2023). 
Others highlight the notion of biases and their reproduction in AI-led decisions to hinder trust 
(Heyder et al., 2023). This is a critical factor for a relationship because difficulties in 
understanding generative AI oppose an obstacle to human-AI collaboration (Chowdhury et al., 
2022). The research within the field broadly covers two areas. 
The first area investigates the effects of algorithmic transparency on managerial 
acceptance of AI output (Lebovitz et al., 2022). This work is based on the reasoning that 
managers change their initial position based on their ability to understand the reasoning of the 
AI (Carlile, 2004). The uncertainty of AI results leads either to a reflective engagement with 
AI or an ignoration of the AI (Lebovitz et al., 2022). However, an opposing view is that 
managers solely rely on the generated output of AI disregarding their knowledge or capability 
to understand the decision (Balasubramanian et al., 2022; Murray et al., 2020). The possibility 
of not interacting with AI either through ignoration or blind acceptance becomes especially 
difficult with the purpose to study collaboration between managers and AI. 
The second area investigates the distributions of accountability and responsibility over 
conjoint managerial practices from AI and humans (Scott & Orlikowski, 2012). Since 
generative AI can function on low complexity levels of routine-based tasks and high 
complexity levels of decision-making tasks, it can engage in the full variety of managerial 
practices (Korzynski et al., 2023). This creates uncertainty over the accountability of AI-based 
actions and questions the actor of managerial practices (Mittelstadt et al., 2016; Pakarinen & 
Huising, 2023). The fact that even low complexity tasks can be used as a decision basis leads 
to a shift from solely a human responsibility to a potentially shared responsibility of humans 
6 
and AI, leading to debates on managerial responsibility especially in decision-making 
(Lindebaum et al., 2020). While some researchers highlight managerial fear of making the 
wrong decision based on AI output (Lebovitz et al., 2021), other research shows that managers 
might not feel the same sense of achievement if AI has supported them (Dwivedi et al., 2023). 
Related questions also arise regarding the employee acceptance of conjoint manager-AI 
decisions as employees are unaware of the exact evaluation of their work (Kellogg et al., 2020). 
The different areas indicate that a successful collaboration or even a collaboration in itself 
is not guaranteed between managers and algorithmic learning. Building on this notion, some 
research focuses on what is needed to create a collaboration (e.g. Bankins et al., 2023). Despite 
the challenges, research demonstrates that companies are already enhancing their managerial 
activities through generative AI (Dwivedi et al., 2023), thus paving the way for potential 
collaboration. Consequently, this study takes a step back trying to understand the relationship 
between AI and how responsibility is balanced across a diverse and complex task set. 
 
Human-AI relationships within a sociomaterial framework 
To understand the relationship between generative AI and managers this research focuses on 
the sociomaterial practices and interactions. The fact that algorithmic learning such as 
generative AI can possess characteristics that traditionally have been connected to human 
agents creates the need to discuss agential practices within the field of technologies from a new 
perspective (Johri, 2022). Previous research shows that sociomateriality can provide an 
insightful lens to study human-AI relationships (Bailey et al., 2022). It allows us to study the 
change in relations of emerging technologies that are inscrutable and under constant change. 
Yet, only a few have studied sociomaterial relations empirically (Jones, 2014).  
The theory of sociomateriality has emerged as a way to see the relationship between the 
social and the material (Leonardi, 2013b). Highlighting the relationships, sociomaterial studies 
focus on the practice (Scott & Orlikowski, 2012) and interaction (Bailey et al., 2022; Barrett et 
al., 2012) between the social and material to understand its entwinement (Anthony et al., 2023). 
According to Leonardi (2013b), the social could be described as “abstract concepts such as 
norms, policies, communication patterns, etc,” while the material could be explained as “the 
arrangement of an artifact's physical and/or digital materials into particular forms that endure 
across differences in place and time” (p. 74). By analyzing the entwinement of the social and 
material it can be understood how generative AI and managers form a relationship. While some 
research discusses potential asymmetries between the human and the AI within sociomateriality 
(Mutch, 2013; Suchman, 2006), other research envisions both entities as equal and 
simultaneously constructed (Leonardi, 2013b; Scott & Orlikowski, 2012). The latter is the 
focus of this research allowing for a balanced perspective on the relationship in which AI gains 
resemblance of a human.  
According to Leonardi (2011), human and material agency are constructed by their 
capability to act and can only be distinguished by nuances in which human agency concerns 
how people have the capacity to achieve their goals, while material agency concerns how 
technology does specific aspects that the people are not entirely in control over. Consequently, 
people need to accept and work around the material agency in order to reach their goals 
(Leonardi, 2011). Agency does not only need to be understood as something that humans obtain 
7 
but also in relation to materiality. Due to the perceptual element of technology, technologies 
are not to be seen as independent and definite as materiality is able to gain agency and capable 
of both being acted upon and be acting (Styre, 2017). The perceived ability of the technology 
to enable the person interacting with it to achieve their goals does also affect the material 
agency. In other words, the perceived affordances or constraints of the person interacting with 
the technology together with the materiality itself, constitutes the way technology is acting and 
being acted upon (Leonardi, 2013b). As stated by Orlikowski and Scott (2008), work practices 
are constituted not only of social practices but also of materiality, and sociomaterial practices 
are not able to be reduced to solely the technology itself or the social aspects. Leonardi (2011) 
further explains how the affordances and constraints can appear in the relationship between 
people and materiality, as people perceive whether the technology can help them accomplish 
their goals or not. Observing the affordances and constraints that are related to the materiality 
of the technology is beneficial when aiming to understand how the social and material practices 
are being shaped (Leonardi, 2013b). Depending on these perceptions regarding affordances 
and constraints, new intentions can be realized that could be accomplished through the material 
features which consequently might end up in new routines (Leonardi, 2011). 
Although the concept of affordances and constraints have found high resonance in 
empirical studies (Cranefield et al., 2022), it shows difficulties in presenting the entanglement 
between humans and technology within a relationship (Faraj & Azad, 2012). This is because, 
within empirical research, the material would need to be at least implicitly defined, which 
would decouple the material from the social and contradict the constitutively entangled stance 
in sociomateriality (Leonardi, 2013b). An attempt to overcome these issues appears in a study 
on the opacity of AI that considers sociomateriality as enacted in practices rather than treating 
characteristics of AI as inherently possessed (Lebovitz et al, 2022). This is described as 
constitutive entanglement, meaning that there are entities that only exist dependently without 
inherent characteristics (Orlikowski, 2007). Consequently, the choice is made to define the 
characteristics of AI through sociomaterial practices and follow a more nuanced perspective on 
relations (Faraj & Azad, 2012). Therefore, viewing the relationship as mutually constitutive 
from the beginning could lead to the researcher overlooking how the relationship is being 
shaped just as well as the reasons behind it (Scott & Orlikowski, 2014).  
Building upon Orlikowski (2007), in this thesis the material and the social are seen as 
mutually shaping one another yet not independently existing individuals. The relationship 
between the social and the material is constituted through each other’s interactions (Scott & 
Orlikowski, 2014). This makes us understand agency as something that is created within the 
relationship of generative AI and the manager. Through the interactions of AI with humans, 
the agency is identified in relationships created through non-human and human practices (Fabri 
et al., 2023; Latour, 1987). The theoretical perspective of sociomateriality supports the study 
by providing a lens that makes us view interactions equally shaped by the manager and 
generative AI. Sociomateriality as a framework sees the agential relationship as a whole and 
not a framework for human and technology-separated (Cecez-Kecmanovic et al., 2014). 
Thereby the material and the social are seen as one entity while the agency is not located in 
technology or humans but rather lies in the intra-actions of the sociomaterial (Orlikowski, 
1992).  
8 
Building upon the theoretical framework of sociomateriality when understanding how 
materiality is an integral part of everyday work practices, the concept of agency supports the 
observation of the relationship between managers and generative AI (Orlikowski, 2000). As 
stated by Barad (2007): “It is important to note that the ‘distinct’ agencies are only distinct in 
a relational, not an absolute sense (…) they don’t exist as individual elements” (p. 33). 
Furthermore, agency is not a concept that someone has inherently but instead comes out of 
sociomaterial enactments (Orlikowski, 2010). Moreover, the iterative process of performing 
social practices and work routines requires an understanding based on what Orlikowski and 
Scott (2008) call sociomaterial reconfigurations in order to perform organizational realities. 
The continuous reconfiguration of the sociomaterial work practices, referred to as practice lens, 
is therefore what constitutes agencies and not inherent attributes of humans or technologies. 
Applying a practice lens is beneficial due to the performative element being central in 
sociomateriality and the relationship between humans and technology is not pre-determined 
but instead understood through the enactments of technologies in practices (Orlikowski, 2000). 
According to Scott and Orlikowski (2013) a theoretical plurality in studies should be 
maintained to allow for new perspectives and experimentations. Thereby the framework of this 
study combines the complementing studies from Leonardi (2011) on affordances and 
constraints and the practice lens on emergent agency of Orlikowski (2000). 
 
Methodology 
As outlined by Silverman (2022) qualitative research is especially suitable for being able to 
analyze the in-between, how situations evolve, and how people communicate. In an attempt to 
study the AI-managerial relationship, we pursued qualitative research through interviews and 
gained access to managers with experience in daily interactions with AI. Focusing on managers 
across different organizations enabled us to gain a larger data set of the individual activities of 
managers and AI, exceeding the number of managers in one organization. This cross-
organizational study allowed us to focus on the relationship rather than the social embedding 
and surroundings that constitute the context shaping sociomaterial practices (Scott & 
Orlikowski, 2012). Thereby we observed relational practices between managers and AI, instead 
of organizational practices, highlighting the formation of a relationship. Each manager engaged 
in a relationship with generative AI which was differently formed and shaped on an individual 
level and which could be aggregated across the interviewees into the evolving process of a 
relationship. 
The research followed a micro perspective (Leonardi, 2013b) by focusing on the daily 
practices of managers and generative AI that shape a relationship. Analyzing micro-actions, 
such as managerial practices, rather than macro-level patterns has proven to be a suitable 
method for relational studies by showing variations in the enactment of social and material 
entanglement as something processual instead of stable (Jones, 2014). The thematic analysis 
of the collected data focused on relational practices enabled us to understand both the 
manager’s and generative AI’s engagement. Thereby we could explore the dynamic and co-
constitutive nature of social relations and material artifacts highlighting the interplay between 
social and material elements (Orlikowski & Scott, 2008). 
 
9 
Setting of the study 
To understand the relational entanglement of managers and generative AI we conducted 28 
semi-structured interviews of managers in the tech field (see Table 1). The managers were 
selected based on responding positive to using text-to-text generative AI within their daily work 
practices through a request via companies’ open inquiry e-mail addresses. Due to the similar 
user interface, ChatGPT and Microsoft Copilot could be utilized. The managerial positions had 
to incorporate both administrative and strategic tasks such as Chief Product Officer or Head of 
Digital Innovation. This allowed us to observe interactions with AI across all potential task 
application levels. Given the preliminary stage of this generative AI study and the fact that few 
managers currently employ generative AI (KPMG, 2023), we wanted to ensure that the 
interviewees had sufficient expertise and experience in interacting with generative AI to 
potentially be entangled in a relationship. Research has proven that employees with 
intermediate expertise in technologies, such as managers in the tech field, are usually more 
prone to use AI and identify its benefits (Allen & Choudhury, 2022). Moreover, the work 
context, such as transport, medicine, etc. is crucial for people’s willingness to interact with AI 
(Schwesig et al., 2023). Therefore, focusing on a technical work context among the 
interviewees supported the acceptance rate of generative AI and ensured a high willingness to 
interact with AI.  
The managerial tech field focus was ensured by interviewing managers who either have 
a managerial position focused on technologies or who work in a technology-related company. 
Managers with a technology-focused position have responsibility over technologies whether it 
is the programming process, the technical products, or technology implementations. This can 
be found in positions such as Manager for Digitalization or Lead Data Scientist. Contrary, 
managers within a technology company contribute to the production of technical products. A 
technical product can range from software development to IT service providers. Their 
managerial responsibility does not necessarily have to be tech-related and can include various 
departments, or projects, such as a Director of Marketing & Sales. Nevertheless, the position 
can directly be linked to the technical product such as Chief Product Officer. Studying both 
these two types of managers enabled a large set of data consisting of interviews with 24 male 
and 4 female managers within the tech field.  
In the following sections, both types of managers within the tech field are collectively 
referred to as tech managers. Although their work practices showed differences, we could 
identify administrative and strategic tasks that create a similar interaction potential with AI. 
The daily work of both manager types in the tech field consisted of writing company updates, 
reading articles, participating in meetings, and negotiating with different stakeholders. Work 
practices of a manager in a tech position included additional evaluations on software 
investments and policies, as well as occasional coding whilst managers in a tech company 
evaluated product developments, developed organizational strategies, and connected with 
clients.  
 
 
 
 
10 
No. Gender Tech Focus Position Country Application 
1 Male Company Chief Product Officer Germany ChatGPT 
2 Male Company Chief Product Officer Germany ChatGPT 
3 Female Manager Manager of Digitalization Sweden ChatGPT / Copilot 
4 Male Manager Product Owner Sweden ChatGPT / Copilot 
5 Male Company Manager for Strategy  Sweden Copilot 
6 Female Manager Head of Digital & Innovation Sweden ChatGPT 
7 Male Company Senior Manager Sweden ChatGPT  
8 Male Manager Head of AI Governance Sweden ChatGPT / Copilot 
9 Male Manager Lead Data Scientist Sweden ChatGPT / Copilot 
10 Male Company Director Marketing & Sales Sweden ChatGPT 
11 Male Company Senior Manager Sweden ChatGPT 
12 Male Company Chief Product Officer Sweden ChatGPT 
13 Male Company Lead Data Scientist Sweden ChatGPT / Copilot 
14 Male Manager Manager of Digitalization  Sweden ChatGPT 
15 Male Company Chief Product Officer Germany ChatGPT 
16 Male Company Chief Executive Officer Sweden ChatGPT / Copilot 
17 Male Manager Manager of Digitalization Sweden ChatGPT 
18 Male Company Chief Executive Officer Sweden Copilot 
19 Male Company Product Owner Germany ChatGPT 
20 Male Manager Chief Technology Officer Sweden ChatGPT 
21 Male Manager Lead Data Scientist Sweden ChatGPT 
22 Male Company Head of Development Sweden ChatGPT 
23 Male Company Chief Executive Officer Sweden ChatGPT 
24 Male Company Chief Executive Officer Sweden ChatGPT 
25 Female Company Lead Customer Care Sweden ChatGPT 
26 Male Manager Senior Manager Digital Germany ChatGPT 
27 Female Company Manager for Strategy Sweden ChatGPT 
28 Male Manager Manager of Digitalization Germany Copilot 
Table 1. Characteristics of interviewed managers in the tech field  
The participants were based in Sweden and Germany, to leverage the advanced AI integration 
in these countries’ tech industries. According to Hofstede Insights (2023), both cultures show 
a pragmatic orientation towards changes making them capable of adapting easily to the 
emergence of new technologies. While acknowledging that organizational cultures may differ, 
this study focuses on the relational dynamics with AI across similar managerial roles. 
11 
Therefore, the potential variability in organizational culture is not a primary concern for this 
research, as the emphasis is on understanding the interaction between managers and AI 
technologies, which transcends local cultural specifics. 
 
Data collection method 
By following a semi-structured interview, the managers in the tech field were able to speak 
freely about their own perceptions and ways in which they interacted with AI (Bryman & Bell, 
2011; Silverman, 2022). To gain an understanding reduced in preconceptions we followed a 
grounded approach with the interviews (Anthony et al., 2023). In alignment with Charmaz’s 
(1995) understanding of grounded studies, we developed our interview guide through initial 
broad questions allowing for open data collection, such as “How do you integrate generative 
AI in your daily work life?”. This allowed us to identify more prominent topics and narratives 
that were brought up by the interviewees themselves and identify similarities across different 
managers. Moreover, the broader perspective of the questionnaire and managers across 
multiple organizations enabled us to ensure the transferability and applicability of our findings 
(Carminati, 2018).  
To provide an open atmosphere, the interviewees could choose the language they felt 
most comfortable with (Chen, 2011), resulting in conducting the interviews in the manager’s 
respective first languages. In the initial interview, both researchers were present to ensure 
consistency in the interview style. Subsequent interviews were conducted individually to 
mitigate the power dynamics of the interviewer (Kvale, 2006). The interviews were conducted 
online, via Microsoft Teams, to overcome physical distance in a cost-effective and easy-to-use 
way, while being aware of the risk of technical problems (Archibald et al., 2019). The majority 
of the interviewees were already familiar with Microsoft Teams within their organization, 
leading to almost no technical problems. The duration of the interviews ranged from 32 minutes 
to 71 minutes. Through an iterative data collection and data analysis process, we could identify 
data saturation at 28 interviews (Bowen, 2008; Glaser & Strauss, 1967).  
 
Data analysis 
The interviews were recorded and transcribed by the interviewer. The transcription was 
automated through Microsoft Word Online and then examined by the researcher while 
simultaneously listening to the recorded audio file. Thereby, not only errors could be detected 
but also previously unnoted features, reoccurring wordings and underlying meanings of the 
managers on the relationship (Silverman, 2022). The interviews were then translated by deepL 
and discussed with the other researcher to overcome misinterpretations of wordings (Winchatz, 
2006).  
To identify themes of managers’ relationship with AI, without being influenced by the 
strong public narrative of AI (Anthony et al., 2023), we perused an open coding approach where 
no codes were initially generated but derived from the empirical data (Corbin & Strauss, 2008). 
The coding process was conducted on NVivo to facilitate a collaborative qualitative data 
analysis. The initial phase of the coding process was to identify important codes related to the 
relationship between managers and generative AI. By coding meanings instead of sentences or 
words, we could overcome the decontextualization of the responses (Garrison et al., 2006). 
12 
Since two researchers have conducted this study, securing intercoder reliability was an 
important issue (Campbell et al., 2013). To ensure intercoder reliability we followed the 
guidelines of Campbell et al. (2013) and each interviewer initially coded the first interviews 
individually and compared them to agree on the codes and coding approach. To strengthen 
reliability, a codebook was developed following the coding of the first set of transcripts 
together, including codes that were well-defined and commonly understood by the researchers. 
The codebook supported getting an overview of all of the codes as well as enabled both 
researchers to track where every single code stemmed from within each of the transcripts 
(Campbell et al., 2013). The following interviews were coded separately. The additional codes 
that were derived during the coding process were collectively discussed and subsequently 
incorporated into the codebook. After our initial study of the data, we refined the coding 
scheme and sorted the codes topic-related into axial coding groups (Corbin & Strauss, 2008). 
Generating overarching themes out of codes was beneficial to understanding the codes more 
thoroughly (Braun & Clarke, 2006). 
 
1st Order Concept 2nd Order Themes Aggregated Dimension 
Routine based tasks 
Time-demanding tasks  AI is performing tasks 
Inspiration 
Support 
More time for strategic tasks 
More time spent with partners  Managers can do more tasks 
More time to increase work quality 
Creating data access 
Third party information 
AI gains data availability 
Private information 
Strategic company information Hesitation 
Adaptation of practices  
Unclear usage Managers change work habits 
Time-consuming prompting 
Confirmation on uncertain topics AI gives personal reassurance 
Potential of wrong results from AI 
Sensibility for hallucination 
Managers fact check 
Sensibility for ethics 
Controlling sources Validation 
The manager is responsible for the used 
AI questions responsibility of work 
results 
Rewriting generative AI results 
Ensuring copyrights Managers personalize 
Appreciation of effort 
Responsive discussion 
Easy user interface AI communicates human-like 
Providing context of the manager and AI 
Nicknames 
Managers replace human 
Approaching AI instead of colleagues 
interactions with AI 
Unimaginable to be without AI Partnership 
Low understanding of context 
AI is no full human-being 
Missing initiative for follow-up question 
Presenting the results to humans 
Managers seek group cohesion 
Adding human experience 
Table 2. Thematic analysis of interactions following the Gioia method (2013) 
13 
The data analysis of this study was conducted through the thematic analysis provided by Gioia 
(2021), which comprised three steps. Firstly, the open codes generated during the initial coding 
were sorted into categories with the objective of creating first-order concepts. The initial aim 
was to maintain a substantial number of categories for the researcher to subsequently identify 
and delineate the key categories. The second step was the creation of second-order themes, 
where the first-order concepts were reviewed from a theoretical standpoint. Thereby we 
followed Gioia’s (2021) key question: “Can I see something the informants themselves do not 
see?” (p. 25). Aiming to study the relationship of managers and AI we identified practices from 
the manager and AI that formed the relationship, such as “AI is performing tasks” (see Table 
2). By going over the concepts with more of a theoretical perspective in mind, the key 
categories that alluded to a similar topic were structured under the same theme. The last step 
was that of creating aggregated dimensions, where the second-order themes are distilled even 
further from a theory-centric point of view. Concerning this step, four major aggregated 
dimensions were created based on the interactions between managers and AI that are created 
out of the practices. This method permitted the presentation of the data in an overview of 
concepts, themes, and aggregated dimensions as presented in Table 2 above.  
 
Limitations and Ethical Reflections 
The data collected in this study was anonymized and handled following the privacy regulations 
of the University of Gothenburg (2023). The participants were informed about the process as 
well as the purpose of the study and their role in it, aiming to reduce their risk of withholding 
information. Moreover, all interviewees were informed of the measures taken to ensure the 
confidentiality of their data and could withdraw their participation and request the cessation of 
recording at any point.  
Reflecting on our method from an ethical perspective, we need to acknowledge that this 
study has been conducted by two researchers of different sexes. Although we aimed to align 
our interview style there is an inhabited difference in conducting interviews by male or female 
that might have affected the responses of the interviewees (Matteson & Lincoln, 2009).  
The methodological limitations of our study are primarily based on the setting. While the 
study selected managers with experience in technologies, it is primarily limited to those who 
responded positively about their AI usage. This could create a selection bias, as it may exclude 
those who no longer interact with AI based on negative experiences and limit the transferability 
to all tech managers (Clougherty et al., 2016). Moreover, the interviewees could identify 
generative AI as taken for granted meaning that they will struggle to articulate how they interact 
with it (Orlikowski, 2000), reducing the accuracy of the interviews. However, due to the 
novelty of text-based generative AI, we believe that the taken-for-granted perspective is still 
limited. Furthermore, despite aiming for transferable findings and focusing on practices with 
AI, interviewing managers of different positions can remove the contextuality thus our findings 
need to account for that. 
 
 
 
 
14 
Empirical findings 
The focus of the analysis is to understand the relationship between the manager and generative 
AI and how they engage in it. The aggregated dimensions, described subsequently, are 
individual parts of the process of forming a relationship (see Table 2). They are presented below 
in the initial order of forming a relationship: First, the interest in a relationship is created due 
to a supporting function; second, a hesitation is opposed due to accommodating activities; third, 
mutual validation is shaping the work of the relationship; and fourth, managers and AI engage 
in a human-like partnership. Nevertheless, it is not always the case that these dimensions are 
sequential, they can also be iteratively and only partially exerted. Each process step can 
influence another through interacting in the combined process of a relationship, without 
necessarily following each dimension. In the following section tech managers will be referred 
to as M with their respective numbers. Their detailed characteristics can be found in Table 1. 
 
Dimension of support 
The manager and generative AI engage in a supporting function through their interactions. 
Managers can identify the benefits of gaining support in their daily work life and based on that 
decide to engage in a relationship. The support that is gained through the interaction of 
managers and generative AI is based on two aspects.  
The first aspect is that generative AI can perform certain tasks in the managerial position. 
Many tech managers describe how ChatGPT and other applications can provide support in 
routine-based tasks such as structuring e-mails, rephrasing LinkedIn posts, or translating 
messages and technical manuals in an international work context. By creating text content with 
AI, managers can improve their writing styles. Whilst some managers explain that they provide 
AI simply with their notes and ideas others already have a pre-written text and expect AI to 
rewrite it. Thereby the tech managers highlight the improved grammar, text structure, and word 
variety. Those that emphasize the text quality increase, highlight that formulations are one of 
their weaknesses. One manager asserts that their writing style oftentimes reflects the inner 
thought process making it difficult for others to understand, but through AI the clarity and 
coherence improved significantly. Furthermore, those managers who do not feel comfortable 
with a second language show strong interest in ChatGPT’s business vocabulary and 
formulations. One interviewee describes AI’s routine-based supporting role as: 
[It can] respond, summarize documents and everything like that which we 
otherwise do manually, and even with translations we have previously done 
manually. Some assistant or some junior lawyer has had to sit and translate 
contracts manually from page A to B and now of course you use these tools with 
the features and you get a translation - M6 
Additionally, time-demanding tasks that are described as “not great fun” (M16) can also be 
supported by generative AI. These tasks can range from finding an important Excel file, that 
the manager knows exists but not where, to summarizing company statements, meeting notes, 
or e-mails that otherwise take a long time to read and understand. The summaries generated 
with AI appeal to the managers due to the low time requirements and the clear structure. For 
example, it can extract the root cause and solution of a long e-mail threat that managers receive 
15 
and would have worked through manually before. In addition to the summary, tech managers 
can ask questions about the provided documents or e-mails so that they no longer have to scan 
the whole file but can quickly read the important section, like a “little intelligent FAQ” (M12).  
Moreover, generative AI supports the inspiration process of managers. Instead of starting 
on a blank page, AI generates ideas and creates that first draft – “you do not need to start with 
an empty canvas” (M8). This draft is the support the manager needs to further develop 
PowerPoints, e-mails, or LinkedIn posts by skipping the time-consuming initial phase of idea 
generation. The draft creation usually contains two steps. First, there is an initial phase of 
understanding a topic which is often an individual thought process and has been supported by 
search engines in the past. Unlike other technologies such as Google, ChatGPT can combine 
multiple sources reducing the time managers would have to spend on screening multiple 
articles provided by Google. One interviewee explains the difference like this:  
In Google you only get lots of examples and then you have to sit and read and it 
takes so long to read all this. Whereas on ChatGPT I get a summary view of it 
anyway. Then I kind of do not have to read through four articles or five articles to 
get a summary – M13 
Second, generative AI can support the actual idea generation and draft creation. The topics 
managers want inspiration on can significantly vary such as workshop invitations, information 
on security measures, or writing a contract. The questions managers then ask are content-based 
such as: "How could I conduct such a workshop? How could I structure it? What points should 
I have in the invitation?” (M11). Thereby applications like ChatGPT provide an overview and 
creative suggestions on the topic. Leveraging these ideas, managers can skip their initial 
brainstorming phase and further develop AI’s ideas based on whether they like, dislike it, or 
identify missing aspects in AI’s list. The process of understanding a topic with its potential 
breadth and depth is described as highly shortened due to the support of AI. 
The second aspect of support shows in the analytical parts of a manager’s daily work that 
generative AI cannot help them with but indirectly supports them by providing more time for 
it. The support from AI is recognized as time efficient as it enables an acceleration of routine-
based, time-demanding, and inspirational tasks. Therefore, tech managers “can do more in the 
same time” (M2). The areas in which the managers use their additional time can vary. Some 
report that they spend more time on strategic tasks and others state that they spend more time 
creating analytical output for customers instead of being stuck on repetitive tasks. One manager 
describes the synergies between AI’s support and the manager’s own work as such:  
I can become more effective as a leader to spend less time on routine tasks, be more 
present in the organization, be closer to the people by getting that support and that 
assistance - M17 
The strategic tasks are seen by the manager as “fun” (M17) as they are significant for the 
company’s success and require their managerial expertise. One interviewee summarizes the 
time synergies with AI as such: “It probably saves me half an hour every time, which allows 
me to focus on what actually requires the human brain in a different way” (M27). Both strategic 
and analytical tasks are viewed as more valued by the interviewees since AI cannot conduct 
them. Moreover, the acceleration of simple tasks shows an increased speed in managerial 
decision-making as managers reach the point of option evaluation more quickly. Furthermore, 
16 
many highlight how the work content quality has increased resulting from having more time to 
investigate details, gain a broader perspective on the topic and develop ideas more thoroughly. 
One manager called this effect “a performance booster” (M8).  
Additionally, the time-efficiency is used by managers to spend more time with clients, 
partners, or colleagues. Due to e-mails being generated faster and simple questions being 
quickly answered by AI, managers spend less time planning and more time with human 
interactions to discuss complex topics. Those soft values are summarized by one interviewee 
as “[having] more time to spend on personal conversations with people or just walk around the 
office landscape” (M27). Thereby, managers feel that they can use their time more effectively. 
A direct synergy between AI and the managers’ attendance in the physical world is identified 
by one manager. The manager describes that generative AI could provide a summary for a large 
part of meetings that they currently attend yet often feel unnecessary: 
It will be fewer people in meetings and then a summary goes out that the AI has 
generated; ‘this was what was being said during the meeting, these are the key 
points, these were the tasks that were delivered’. … instead of that people have to 
spend time in meetings that could have been an e-mail, you just receive your 
summary – M8 
It is even possible to identify a non-productive benefit of a manager getting coffee while 
ChatGPT is solving a prompt: “I'll do something different at that moment. Get myself a coffee 
or prepare the website for a colleague” (M2). The tasks AI can do and those that the manager 
can do more in-depth both show how the relationship between both parties consists of support.  
 
Dimension of hesitation 
Although managers and AI engage in a supporting manner it can also be noticed that hesitation 
is shaping their relationship. To create the relationship preconditions for the managers and the 
generative AI applications need to be met that can oppose hesitation.  
First, the manager needs to change its existing work habits so that applications like 
ChatGPT can be integrated. This step consists of a continuous adaptation and learning process. 
Although many interviewees explain how interacting with generative AI is straightforward, 
they still admit that their working habits have to be changed. These habits can be rather simple 
as in switching from Google to ChatGPT or can be more complex as in understanding how 
ChatGPT structures its output. One manager explains: 
It's still a bit like this with the power of habit, so it's very easy to, you know, google 
a question or google when gathering information. But then I've also started to 
realize that I can ask the question to the chat, but so far I can say I'd rather google 
questions because it just feels like it's a habit – M7 
The time this adaptation process takes can vary, however, tech managers highlight that it not 
only takes time, but they have to actively make time to foster adaptation and learning. Managers 
need to acknowledge the potential support an interaction with AI will provide and then spend 
time to familiarize themselves with AI by understanding the underlying mechanisms, the 
interaction techniques called prompting, and the possibilities to extract the most benefits from 
interacting. To engage in a relationship, managers need to spare time to gain knowledge and 
17 
change their work practices. This trade-off between time and benefits from the support is 
elaborated by a tech manager:  
Time is money in this industry. We sell time and everyone is always under pressure 
to work a lot on weekends, evenings, nights and then to take the time to kind of 
stop and 'no, but now I'm going to do this in a new way', it's always like a threshold 
before you are up and running with something new – M6 
Both interest and age are identified as key determinants of the adaptation and learning process. 
The majority of the interviewees share their perspectives on the early stage of adaptation. Many 
interviewees employ similar phrases like “But it is also, again, very early at the stage” (M5). 
Generative AI has recently been introduced into their work practices. Consequently, the 
interactions are still explorative and mainly driven by their interest in technologies. The novelty 
of the new AI technique leads to uncertainties in adaptation. Some report that understanding 
how AI works is difficult while others underline the missing guidelines on how to benefit from 
generative AI the most. In some cases, the adaptation to ChatGPT or Copilot is discontinued 
due to the manager’s belief that the tasks can be done faster or better themselves. One manager 
describes the uncertainties as managing a jungle of AI: “Which applications and which models 
are good and what suits us and navigate the jungle because right now it's a jungle” (M11).  
Furthermore, a strong difficulty in the relationship between managers and generative AI 
is the unique prompting process that does not exist with other AI algorithms. Managers describe 
the prompting process as iterative. The first results from ChatGPT etc. are mostly not used, but 
new, more specific prompts are created thereafter to refine the results: 
You rarely get the perfect answer the first time, so then it's important to ask a 
follow-up question and ask for clarifications and improvements to the answer of 
course, so that with one, two or three iterations you can usually get a pretty good 
answer – M17 
While this iterative process is described by many interviewees, some believe that their 
formulation of prompts was the cause of the problem of vague results. Consequently, 
prompting has been identified as time-consuming, and in some cases, managers give up on the 
iterative process due to time constraints or dissatisfaction with the quality: “I have also had 
cases where I gave up because it took too much time for me to ask and formulate things over 
and over again. And it's not fast either” (M2). 
As a second precondition of the interaction, generative AI has to gain access to 
information and data. On the one hand, managers provide AI with information in the prompt 
formulation on their customer, company and personality thus enabling AI to generate insightful 
results. On the other hand, it can be beneficial if technologies like ChatGPT have direct access 
to internal documents of the organization, providing internal sources for AI. While interacting 
with the manager, AI necessarily gains availability on some data. However, many of the 
interviewees indicate hesitation in revealing information. Generative AI is perceived as an 
external server making every data available to AI a potential obstacle for security. One manager 
described the external server: “You do not know where the information is processed and ends 
up” (M22). Consequently, tech managers discuss data security on an external, internal and 
private level. The interviewees agree that external, third-party data, such as those from clients 
18 
have to be anonymized as the managers have no rights on them. The opinions on company 
internal data are torn between the managers. Some argue that technologies such as ChatGPT 
along with their developers are incapable of accessing the strategic data that will be revealed 
in a prompt. Other managers view ChatGPT and Copilot as full external actors and restrict the 
usage for strategic and company internal content to limit the risk of exposure. This can be 
observed in the following exclamation: 
ChatGPT is an external service and has to be managed like that. I can’t explain that 
'I’m a manager that works at *company name* and do all of these things', that 
leaves too much information to the AI that is too detailed simply put - M8 
By reducing AI’s data availability managers notice that the interaction capabilities are limited 
and inspiration in complex contexts is not possible. Especially, strategic projects that have not 
been open to the public are oftentimes made unavailable to AI. Although managers interact 
with generative AI, in some cases they prevent AI from some data availability. One manager 
describes their way of avoiding company internal data access as: “We write the press release 
in Word beforehand not using any generative AI” (M16).  
Moreover, managers also fear to reveal private information about themselves. Despite 
recognizing the need to reveal certain private information, like position or personal preferences, 
they highlight that it is a strange feeling to tell AI about themselves and they will not reveal 
everything. At the same time, they acknowledge that due to registrations of the residence and 
websites like Facebook, there is already a lot of information on the managers online. Still, they 
explain the step of giving AI personal data: “It’s more a bit scary the first time you do it” (M8). 
To engage in the relationship on the one hand managers need to accommodate their existing 
work practices to incorporate AI and on the other hand generative AI needs to have access to 
information to provide precise results. Although managers are aware of the prerequisites that 
need to be met for the relationship, they still hesitate to reveal information and learn about the 
usage of AI.  
 
Dimension of validation 
Even when hesitations on the relationship are overcome, one can identify a consistent step of 
validation. This validation in the relationship comes from both the manager and the generative 
AI. On the one hand, generative AI provides a source of validation for the manager. By double 
checking if ChatGPT would have given them a similar approach to a problem, managers gain 
assurance especially when either the topic or the task is new to them. One interviewee even 
describes that if something is created without interacting with generative AI, the manager is 
doubting the quality of the work. Another manager highlights the capability of AI to not only 
reassure but also identify overlooked areas of the work: 
I think idea generation is good (…) as a completion check. If I have a topic, it can 
tell me what else there is. For me, this is a great check to see whether I'm mutually 
exclusive and have covered everything. – M26 
On the other hand, managers feel obliged to validate AI generated content. This is a 
consequence of tech managers noticing that AI can make mistakes: “When you start writing 
specific things, it can become very wrong and then you have to know that” (M8). Therefore, 
19 
applications like ChatGPT call into question who is actually responsible for the final output of 
the work that is created within the relationship between tech managers and generative AI: The 
generative AI? The developers of AI? The manager? The responsibility is frequently discussed 
among the managers. Concluding, they agree on viewing themselves as responsible for the 
generative AI both for the data included in the prompting and the quality of the result. One 
interviewee compares ChatGPT with a knowledgeable person who supports in daily tasks but 
for whom the manager is responsible.  
The responsibility of the manager over AI is exerted by including fact-checking as a 
practice of the relationship. Managers oversee whether the answers of generative AI are 
relevant and double-check sources provided by AI for correctness. This is motivated by their 
preliminary assumption that answers from generative AI could be wrong:  
But my point of departure is always that they [AI] are not right. It has to do with 
the fact that the language model is such that even though it may be right in 
substance, I may not think that it is right. (…) So I always assume that it could be 
flawed and I've noticed that it is a lot – M24 
Not only are the managers critical about the correctness of generative AI but they are also 
sensible about ethics and potential hallucinations. Ethical sensibility is caused by potential 
biases of generative AI results. Some managers discuss the possibility of biased results of 
generative AI that can potentially be negatively abused – “for evil intentions” (M27). Another 
one adds that only a few developers are in control of the results of ChatGPT etc and can 
alternate the truth. Moreover, interviewees highlight the caution of unselected sources that AI 
draws from. One manager underlines that if one opinion is highly represented on the internet it 
will also replicate in AI’s content. These concerns of ethics and abuse make the interactions 
not fully reliable and highlight the importance for the manager to reflect on AI content. 
Concluding, one manager says: “I will always endeavor to avoid extreme viewpoints and 
positions” (M15). Concerns about hallucinations are more focused on ChatGPT creating 
information that does not exist. Therefore, the interviewees are alerted when noticing “What 
kind of strange thing does (AI) write?” (M6) and rewrite responses to ensure credibility. 
However, the managers need to have a certain understanding of the topic to identify incorrect 
statements from generative AI. If the managers’ knowledge of the area is too limited, they 
either do not interact with AI or they ask other people to proofread. One interviewee explains 
that they have to consult an expert on the topic to identify the mistakes AI has made. 
Apart from fact-checking managers identify personalization as an important step in the 
interactions. Despite AI being able to generate texts that support the managers, many report 
that the texts are missing personality and the responses are rather general.  
Often it is that it sounds far too formal. It simply sounds too good. 'But my 
colleague didn't write that. (…) I've never seen him write like that. But now, all of 
a sudden, he writes very correctly and formally' – M11 
Consequently, managers only take the ideas from generative AI and rewrite them afterwards. 
This rewriting process is motivated by three reasons. First, they want to make it sound personal 
and more human-like, believing that texts generated from AI are identifiable. This can be 
summarized by: “I kind of just write so that I feel like I'm the one who sent it.” (M24). Second, 
20 
they want to avoid the label “generated by AI”. Either because it might be perceived negatively 
by others or because it might reproduce content that is protected under copyrights. Third, they 
believe that rewriting texts from generative AI represents the appreciation of the task. Changing 
the phrasings or even the content of ChatGPT and Copilot shows that the task is important and 
valued. The opposite, if it has not been changed, shows “laziness” (M24) or that “it’s not worth 
it” (M2). The interviewed managers describe that they are not proud of writing texts with AI 
and others have a lower interest in reading texts generated by AI. Both aspects of fact-checking 
and personalization can be based on the fear of letting generative AI act on its own, being 
mentioned multiple times. The autonomous actions of generative AI is described as “the point 
[where] you lose control” (M7) and “a nightmare” (M8). Validation is a significant step in the 
relationship, for reassurance on the one side but on the other side to highlight the responsibility 
of the manager through fact-checking and personalization. 
 
Dimension of partnership 
The relationship between managers and generative AI is identified through actions of 
partnering. The partnership is embodied in the communication between managers and 
generative AI having both internal and external effects.  
First, generative AI and managers communicate internally in a human-like manner. The 
interactions resemble those of a colleague: “How you ask questions and which questions you 
ask is probably just like the dialogue with colleagues at work” (M14). The discussions are 
described as responsive and interactive. Instead of gaining generic answers, the managers can 
provide context both about themselves and about AI. Through details on the manager’s 
preferences and work situation AI becomes aware of the environment of the manager and can 
tailor the output accordingly. Additionally, managers choose which position ChatGPT or 
Copilot should fill and how it should behave. This enables the manager to choose the discussion 
opponent, such as “from now on, act as you are a compliance officer” (M4) and due to the 
context provided the communication is filled with background information and responses that 
are based on individual knowledge, making it feel like one is talking to a human. One manager 
describes this process as “being a good storyteller just as for humans” (M4). The sharing of 
context is perceived to enhance the quality of the results. Consequently, they aim to have an 
individual chat for each context within the application, so that AI can remember like a human, 
and they avoid starting over each time.  
The interface of text-based generative AI allows managers to have a responsive 
interaction and they cite that they are having “discussions” (M3) and “conversations” (M2) 
with generative AI. Due to the chat field within the applications, it is not only referred to as 
easy to interact with, but managers also start engaging in an interactive conversation with AI. 
This interaction is shaped by recursive back-and-forth questions with AI. Thereby generative 
AI is referred to either as a “dialogue partner” (M10) or a “professional discussion partner” 
(M12). Within this dialogue, managers incorporate filling words like “I need a …” (M19) and 
curtesy words like “thank you” (M12) making it sound like human conversations. One manager 
summarizes it by saying: “It's tempting to talk to the thing like a person. It's more comfortable 
21 
and it's easier to think” (M19). Another interviewee fears that discourteous communication 
with AI would transfer into the external world and chooses to remain polite with AI:   
When you start to be careless with politeness or anything like that, you will start to 
be careless in context with real people as well, so I try to keep it with both greetings 
and thanks – M14 
Moreover, managers compare the characteristics of AI with those of a human. Thereby 
hallucination is seen as an obstacle, but at the same time something that all humans do. The 
following interactions between managers and AI are compared to either verbalizing an internal 
thought process or to a discussion within a team. One manager highlights: “The process is very 
similar to working with someone else” (M1). In particular, those within managerial roles are 
aware that their workday consists of many external meetings making their internal meeting 
times limited. Consequently, managers appreciate the opportunity to engage in dialogue with 
AI, rather than being left isolated at the end of the day: ”I use it because I don't have a colleague 
to ask or collaborate with” (M12). 
Second, the external effects of communication between managers and AI can be 
identified by changes in human communication. On the one hand, managers replace human 
interactions with generative AI. Many interviewees report that in some situations they approach 
AI instead of their colleagues. Some do not want to disturb their human colleagues while others 
appreciate the anonymity of generative AI to discuss sensitive topics. One interviewee recounts 
how a colleague had experienced the death of a family member and was uncertain about the 
most appropriate means of providing support. Consequently, the manager seeks anonymity of 
AI to pose the following question:  
'How do you support someone when you know they're having a little bit of trouble 
privately without intruding?' These are things that typically take a lot of time for 
me to find the right words. – M3  
Other managers highlight the constant availability and unlimited time capacity of AI, making 
it appealing to approach AI instead of colleagues. Consequently, one manager explains the 
benefit of never disturbing AI: “Otherwise I would have to abuse those around me. For this 
ping pong, for sharpening your thoughts” (M2). Another manager draws attention to the fact 
that the colleagues do not have the same availability on the weekend as AI does:  
My communicator, she usually helps me write the CEO's speech, and often we sit 
and spend a while together (…). So you take a few turns before it's ready. This 
time, it almost happened instantaneously via AI on a Saturday and she didn't have 
to come in on the weekend and write this – M18 
Moreover, department-specific but rather simple questions that would have been answered by 
colleagues of the IT or Research and Development departments, among others, are answered 
quicker by AI and do not disrupt anyone’s workflow: “It can explain a lot of obvious things in 
their [colleagues] world to me then” (M14). On the other hand, the interactions with AI 
influence the communication style between humans. The managers report how they believe 
that their content is less misunderstood, and the communication is more thought through since 
they interact with AI. Engaging professionals with different backgrounds such as a manager 
22 
and a developer used to be difficult due to specific language and reasoning. Through AI, their 
communication gains clarity and neutral explanations: “It helps a lot in bridging these 
miscommunications” (M4). Moreover, the partnership becomes visible through how managers 
frequently give generative AI nicknames, such as “live coach” (M2) or their “own little tutor” 
(M3). Additionally, in some cases, managers bring AI into the physical world during meetings 
by discussing in groups with AI either on a shared Microsoft Teams screen or on a physical 
screen in a meeting room. One manager describes their group work as such: 
We were actually programming yesterday, me and my team, and I said like this: 
'What should I do now? What should I ask Copilot?' And then three other 
developers were sitting there, and I said: 'What am I even going to do with you?' 
The Copilot had the right answer right away – M13 
The partnership takes over a role where many describe that they can no longer imagine a life 
without AI. It is incorporated into the managers’ everyday life and interactions with AI turn 
into an automatism. One interviewee summarizes the partnership as: “I almost see AI as a 
colleague” (M12). At the same time, managers identify differences between a human being and 
the human resemblance of AI. On the one hand, AI does not fulfill all human characteristics. 
Unlike human colleagues, generative AI does not understand the detailed company’s internal 
context. Some report that the quality is not always sufficient as generative AI does not 
understand implied context like the managers’ age or the company the manager works at, that 
contributes to answering a question. One interviewee describes AI’s missing context:  
It doesn't see how I'm sitting here; it doesn't know me; it doesn't see what stuff is 
on the wall behind me. So it only has the information it gets from me – M19  
Others underline the company’s internal language that AI does not understand. 
Counterquestions would support AI’s understanding of the context, however, contrary to 
humans, AI does not ask questions by itself. The respondents wish for an independent 
engagement of AI in identifying that prompts might be misleading and need further 
refinements.  Moreover, if managers provide AI with context, some report that it also forgets 
its persona and context over time. Unlike human coworkers, AI shows a limited capacity to 
remember thereby basing interactions with the managers only on the most recent prompts 
despite having a long-running conversation. On the other hand, managers seek group cohesion 
that cannot be replaced by generative AI. While some report that the physical component such 
as personal contact and body language is missing, others underline the lack of “social 
interactions” (M4) and “real communications” (M15) which are still needed. Human 
interactions show to create unique feelings of comrades and social belonging. One interviewee 
highlights different excitements between interacting with AI and colleagues: 
If you talk to a colleague, it becomes more of a conversation and not as concise and 
boring. (…) If you only had an AI assistant compared to having colleagues to talk 
to, it would be boring – M11 
Finally, human experience remains highly valued, and trust in colleagues is shown to be higher 
than in AI. After all, final presentations of the results, despite being conjointly created with AI, 
remain a task of the manager.  
23 
Discussion 
In exploring the interactions of generative AI and managers one can identify an evolving 
process of forming a relationship. The key dimensions that have been developed in the data 
analysis: support, hesitation, validation and partnership, prove to be interactions that form 
relation stages between AI and the manager (Barad, 2003). Thereby, the relationship is created 
in four subsequential stages moving from managers perceiving AI’s role as a junior assistant 
to AI becoming a colleague for the manager (see Table 3). These roles result from direct 
quotations from the interviewees. Orlikowski and Scott (2008) refer to the relationship as a 
creation of practices: “Relations are enacted in recurrent activities” (p. 462). Consequently, 
different practices also lead to different stages within this relationship process.  
 
Perspective Interaction AI Role Affordances Constraints 
AI can take over routine-
Trust on AI is highly 
based and time-
limited and tasks that 
“Junior consuming tasks that 
Support include decision-making 
Assistant” require low responsibility 
are excluded from the 
and low knowledge on 
relationship 
the task 
Tool 
The relationship is 
Trust and responsibility 
considered to be worthy 
“External are sensitive topics in 
Hesitation of adapting and altering 
Freelancer" terms of data security and 
routines to gain better 
slow down the adaptation 
results from AI 
AI shows to have high Trust for AI is still 
knowledge and can help limited through 
the manager in hallucination and 
overcoming barriers of misunderstanding. 
“Knowledge-
Validation insecurity, and context Managers remain 
able Person” 
understanding. The responsible by including 
responsibility is extended strong reflections on the 
to providing drafts and sources and wording of 
proposals AI 
Collaboration 
The relationship is 
comparable to a human 
one, including high trust Trust in human 
“Buddy” and and low reflection. It colleagues is higher 
Partnership 
“Colleague” extends to managers valued because AI cannot 
feeling more comfortable provide solidarity  
with AI than humans 
over certain areas 
Table 3. Evolving relationship between managers and generative AI  
The evolving process of forming a relationship is presented in Table 3 above. Based on the 
analysis, the managers in the tech field do not always follow the sequential process. Individual 
24 
forms of the relationship may be omitted as the interactions are not embodied in the technology 
but are an enactment of the human and material (Leonardi, 2013b). However, for the majority 
of the interviewees, there is a linear sequence in which the managers sometimes remained in 
one relational stage, influenced by the previous steps, and are not formalized up to the final 
relationship of the colleague. This aligns with Leonardi’s (2011) understanding of imbrication, 
meaning that past interactions accumulate over time causing human and material agencies to 
change in the process. Consequently, the discussion is based on the process of forming a 
relationship constituted by the interactions that influence managers’ perception of the AI role, 
which ultimately result in affordances and constraints affecting every stage of the relationship. 
 
Tool perspective 
When entering the relationship, the managers reveal to view their AI interactions from a tool 
perspective. Through a phase of trial-and-error managers identify AI as a tool in their 
managerial practices. Thereby the perception of a tool is not inherent in the attributes of AI but 
comes out of sociomaterial enactments (Orlikowski, 2000). The interactions are shaped by the 
manager’s realization of benefits in their work life and limitations in AI’s capabilities driven 
by an explorative interpretation of generative AI.  
The first stage that the manager and AI engage in, is that of a junior assistant. In this stage 
of the relationship, the manager sets the goal of reducing the daily workload by interacting with 
AI. The managerial aspiration to accelerate their pace of work can also be found in previous 
managerial AI studies (Lifshitz-Assaf et al., 2021) and transfers into supportive interactions 
with generative AI, which the managers equate with a junior assistant. Managers respond to AI 
conducting managerial tasks by implementing new creative practices and by feeling relief from 
unnecessary duties. The material and human agency are not possessed by either the manager 
or the AI but come out of the interactions (Leonardi, 2011) causing each manager to conceive 
different support functions.   
The human and material interactions result in both affordance and constraint based on 
the manager’s perception of whether AI is able to achieve the intended goal (Leonardi, 2011). 
The manager has set the objective of reducing the daily workload through AI interaction 
leading to the affordance that AI assists in routine tasks that require low knowledge and 
responsibility and allows the manager to be more time efficient. Thereby AI can take over 
mundane tasks and allow the manager to focus on creative practices (Jarrahi, 2018). 
Nevertheless, it is also possible to recognize constraints on the relation in the limitation of AI 
to practices of low responsibility and the exclusion of strategic practices such as decision-
making. Managers keep taking on complex tasks and cannot feel a workload reduction in all 
managerial practices (Korzynski et al., 2023). This is based on the manager’s limited trust in 
AI as this relationship shows to be still at an early and explorative stage. Previous research on 
technology constraints has highlighted accuracy and transparency as barriers to human-AI 
interactions (Cranefield et al., 2022) that have also been mentioned by the interviewees and 
may contribute to their low trust in Al.  
As the managers aim for support in work practices, they allow for the technology to 
perform parts of the managerial practices and consequently respond with an efficient change 
in time allocation for tasks. This resonates with Leonardi’s (2011) view of a catalyst that is the 
25 
actor that initiates the change in routines and practices. Thereby, in this stage, the data suggests 
that managers initiate the formation by identifying the benefits of support in AI.  
The second stage is that of AI’s role as an external freelancer, which corresponds to the 
dimension of hesitation from the data analysis. To engage in the relationship, generative AI 
needs to be incorporated into the daily practices of the manager and gain access to data. 
Consequently, managers change their work habits by learning how to prompt or change their 
standard search engine from Google to ChatGPT resulting in new routines (Leonardi, 2011). 
In the prompts, AI gains data availability of internal, external, or private data that it uses to 
refine its results (Chen et al., 2023). The potential data leakage has been acknowledged as a 
risk of interacting with AI by the managers. Most of the interviewees reported how AI collects 
information through the written prompts. Consequently, managers do not proactively provide 
data availability but the AI acts upon the manager (Styre, 2017) by making the manager reveal 
to some degree sensible information. Thereon managers feel a loss of control resonating with 
an opaque tool that cannot be acted upon (Kellogg et al., 2020). In this relation, managers 
identify AI as an external freelancer who does not show to be in close relation to the manager 
but rather is seen as an external actor. This opposes obstacles to the manager to engage in this 
relation. Since the work practices regard data access and changed work habits, the interaction 
is characterized by hesitation and AI is still not regarded in a liaising manner. Thereby 
managers do not set a direct goal for this relation but try to accommodate the obstacles. 
The observable affordance is based on the manager’s evaluation of whether it is worth 
adapting their practices to interact with AI. The manager must invest time and effort to learn. 
If human agency can be observed through practices of changed work habits, the technology 
affords the manager to consider the relation worthy enough to accommodate (Leonardi, 2011). 
It is imperative that managers need to be aware of the needs for a relation and willingly adapt 
to it (Zhang et al., 2021). The constraint is however that both trust and responsibility are 
subjects that slow down the adaptation and the learning process for the managers (Cranefield 
et al., 2022). These human and material agencies need to be enacted, meaning managers and 
AI need to engage in the practices, (Jones, 2014) otherwise the relationship cannot form, and 
managers cannot benefit from AI  (Cranefield et al., 2022). This shows in some managers 
deciding to no longer engage with AI and exit the relationship. The hesitation in AI comes out 
of the low trust when providing AI with data and not wanting to risk potential data leakage. 
The fear of data leakage might be connected to the opaque characteristics of algorithmic 
learning causing low trust if managers cannot understand the AI system and who can access 
the data (Chowdhury et al., 2022). The affordances and constraints are evaluated against one 
another if it is worth the time and risk to learn and adapt.  
While Leonardi (2011) based the catalyst on human perceptions, one can identify that 
machines can also initiate the formation of a relation through changes in practice. In the second 
stage, AI can be seen as the catalyst that initiates the new routines and practices becathe 
uncertainty that AI inhabits leads to hesitations that interact through adaptations of the manager 
and accommodations of AI. The manager does not intentionally set a goal for this interaction 
but rather adapts to the AI to continue the relationship process (Zhang et al., 2021).  
 
 
 
26 
Collaboration perspective 
Moving from a tool perspective to a collaboration perspective, managers acknowledge that a 
relationship with AI consists of adjusting practices to overcome risks of data privacy and 
trustworthiness. This change in practices leads to a reconfiguration of the technological 
attributes (Orlikowski & Scott, 2008) resonating with a human-like collaboration. The 
collaboration is shown to be an intentional interaction between managers and AI in which 
managers knowingly adjust their practices while responding to the identified characteristics of 
AI (Orlikowski, 2000). This willingness to respond to technologies in practice is based on a 
recognition of the affordances and constraints of the manager (Leonardi, 2011).  
The third stage of the relationship between managers and AI is that of a knowledgeable 
person, in which the interaction is more complex. The relation is formed by a two-sided 
validation helping the manager to overcome the hesitation of the second relation leading to an 
entanglement of the human and material so that they are perceived as inseparable (Orlikowski 
& Scott, 2008). Managers are not only validated by AI, but they also validate AI. On the one 
hand, the manager sets the goal of being validated through AI. Thereby AI provides reassurance 
by either confirming or contradicting managerial opinions on topics. On the other hand, 
managers reflect on the provided results from AI and create the new practice of fact-checking 
against AI. In order to do so managers need to have confidence in their validation capabilities 
to avoid false interpretations from AI (Lebovitz et al., 2021). At the same time, managers do 
not want to publish texts to their colleagues that are fully generated by AI. Due to the manager’s 
perception that AI results are distinguishable from the manager’s own work and do not fulfil 
the managerial expectations, managers personalize the provided texts from AI. Because AI 
provides the basis for managerial work through their interactions, AI participates in managerial 
practices (Korzynski et al., 2023). This entwinement of managers and AI into managerial 
practices leads to AI questioning who becomes responsible for their co-creative work. 
Managerial practices are no longer solely done by human workers but are co-created by 
managers and AI (Wilson & Daughtery, 2018). Consequently, the entanglement of 
sociomaterial practices becomes evident, as a clear distinction on whether AI or managers have 
done the duties becomes challenging (Scott & Orlikowski, 2014).  
Within this stage, managers realize the knowledge of AI and its ability to both help with 
uncertain questions for the managers while also understanding context better. Thereby they 
identify the benefit of AI as a validation and see an augmentation of the managerial work 
through generative AI (Dwivedi et al., 2023; Raisch & Krakowski, 2021). Consequently, 
affordance is extended by AI’s ability to do more complex practices, such as the creation of 
proposals and increase AI’s responsibility (Leonardi, 2011). Managers feel comfortable with 
letting AI generate early base material such as proposals and drafts and can help the managers 
overcome insecurities when they feel uncertain over a topic. However, managers still notice 
the constraints of AI hallucinating and misunderstanding context and thereby keep a limited 
trust in AI. Although managers validate AI’s results the risk of overlooking vague or false 
statements remains. Unlike the second relation, managers show the goal of actively maintaining 
the relationship and changing their practices to benefit from AI’s validation (Leonardi, 2007). 
By fact-checking and personalization, the texts generated through managers and AI become 
less noticeable and seem to be invisible to their human colleagues. This might be a response to 
27 
the accountability gap that is caused because the co-creation of work makes it difficult to 
attribute a specific outcome to one actor (Mittelstadt et al., 2016). 
Within the interaction of validation, one can observe that the practices are initiated by 
both the manager and the AI. Since the manager aims for reassurance and AI shows constraints 
in accountability initiating practices of validation for both actors.  Therefore, the data suggests 
that both managers and AI become the catalysts for changing the practices of the relationship 
(Leonardi, 2011). Due to the entanglement of the validation practices the necessity for one 
initiation to precede the other is negated (Scott & Orlikowski, 2014).  
The fourth and last observable stage is that of a buddy and a colleague. In this stage of 
the relationship, human and AI interactions become visibly entangled in a partnership. On the 
one hand, the manager’s goal is to have a communication partner that is twenty-four seven 
available, an expectation set by generative AI’s human-like communication. Because managers 
can communicate with AI as if it were human, managers change their routines of approaching 
AI instead of their co-workers and thereby replace their colleagues to some extent, indicating 
that AI might have an influence on employment in the future (Dixon et al., 2021). The 
entanglement of communication practices is revealed in the two-sided human-like discussions 
between the manager and AI. Orlikowski (2010) describes the sociomaterial entanglement in 
practice as the “dynamic sociomaterial configuration performed in practice” and further adds 
that “the social and material are inseparable” (p. 136). On the other hand, generative AI cannot 
provide solidarity which leads to managers seeking group cohesion. Consequently, AI is not 
perceived as a full human being and lacks contextuality, initiative and human belonging.  
The perceived affordance by the manager concerns how both trust and responsibility 
increase in this relationship, as the manager at times feels more comfortable interacting with 
AI than with colleagues. Thereby, AI affords the manager to be less dependent on the co-
workers as managers trust AI with sensible topics such as cases of death and gain comfort 
because they have an interaction partner during weekends and late working hours. 
Nevertheless, the perceived constraint is that the manager never trusts AI to such an extent as 
they do with a colleague because AI does not provide human solidarity and contextuality. AI 
cannot provide the joy managers have in human interactions and managers need to explain 
everything AI needs to know. This is because AI cannot act upon itself. The material agency 
is based on sociomaterial interactions (Leonardi, 2011) and is not something the material 
inherently possesses (Barad, 2003).  
This stage of the relationship marks the last observable process in forming a relationship 
between managers and AI in which AI is most closely knit to the manager. Whilst in the 
validation the managers are aiming to maintain the relationship, now moving on to the 
partnership they acknowledge that it is maintained and the relationship becomes a trustworthy 
partnership. According to Jarrahi (2018), a synergic partnership is articulated in humans and 
machines where the “strengths of one compensate the limitations of the other” (p. 579). 
Managers have identified that AI becomes a discussant partner due to its strength in availability 
when their colleagues are unavailable. They set the intention to form a human partnership and 
form human-like practices of AI while AI initiates a reflection on group cohesion (Leonardi, 
2011). When referring to AI as a buddy or colleague managers make the partnership visible 
and demonstrate that they are not concerned about their interactions with AI becoming 
observable to their colleagues. Rather they are proud that through the partnership they no longer 
28 
need to disturb their colleagues. Acknowledging the partnership can be transferred to the 
sociomaterial context as acknowledging the sociomaterial entanglement (Scott & Orlikowski, 
2014). Managers indicate that they cannot envisage their managerial practices without AI. 
Nevertheless, the manager does not perceive AI as an equal to a human being, and managers 
compensate for it through strong valuations of group cohesion. Perceiving AI with its human 
and non-human characteristics is a recursive process that is different for each manager and is 
constantly recreated (Orlikowski, 2000). 
 
Evolving process of forming a collaboration 
The evolving relationship between generative AI and managers describes the process of 
forming a collaboration. Previous research has called for a move from a tool perspective to a 
collaboration perspective (Anthony et al., 2023). However, in this research, one can identify 
that these perspectives are not contradictory but part of a processual transformation. By 
describing generative AI as a junior assistant and external freelancer managers engage in a one-
sided relation. Either a managerial or material intention is shaping the relation (Leonardi, 
2011). This resonates with a tool perspective where managers simply use AI to gain more time 
and AI needs to be integrated into the daily workflow (Faraj et al., 2018; Kellogg et al., 2020; 
Rosenblat, 2019). The first two relation stages are perceived as beneficial but also obstacle-
driven relations by the managers. This can be seen by affordances in simple tasks and 
constraints in trust and responsibility.  
Moving from hesitation to validation interactions one can understand that a new narrative 
is driving the relationship. Narratives are interpretations of life (Czarniawska, 2004) and show 
that managers perceive the relationship differently along the formation of a colleague. Within 
the process, managers start to discuss responsibility over tasks and AI gains trust not only in 
its performance but also in guiding the manager through their daily work. This narrative shift 
can be theoretically found in the entanglement of the social and material (Orlikowski & Scott, 
2008). In both interactions of validation and partnership, the entanglement between human and 
material agency, initiated by the manager and AI, is identified. This entanglement of the agency 
is forming a collaboration. Managers and generative AI are perceived in these relations as 
agents conjointly engaging in the interaction process (Murray et al., 2020). Collective agency 
of managers and generative AI is a key criterion in forming a collaboration and moving from 
using AI to collaborating with AI (Lebovitz et al., 2022; Michel, 2020). In AI’s role of a 
knowledgeable person, it can be acknowledged that the entanglement of agency becomes a 
two-sided intention of validation. The intentionality of the manager to change routines and 
maintain the relationship (Leonardi, 2011) leads to a switch from a tool to a collaboration 
perspective. However, this relation is formed by a personalization practice that aims to hide the 
interaction practices. Thereby this relation stage is identified as a form of invisible 
collaboration. The fourth stage of the relationship, described as a colleague, is shaped by 
managers replacing interactions with partners through interactions with AI. Trust is created 
between managers and AI to a degree of trusting AI with sensitive topics and valuing its opinion 
and all-time availability. In the final stage of the relationship, managers acknowledge the role 
generative AI has taken on in their lives and thereby form a visible collaboration. Thereby we 
can illustrate not only a narrative shift around AI in academia (Anthony et al., 2023), but also 
29 
demonstrate that managers undergo a change in narratives through words like buddy and 
colleague and make the collaboration visible for themselves.  
 
Partnership as the final stage of the collaboration 
The evolving relationship between managers and AI ends in the stage of a partner based on the 
collected data. Nevertheless, generative AI is a fast-changing technology, constantly updating 
its underlying algorithm (Tse et al., 2024) and thereby confirming the constant state of 
fluctuations of technologies (Baldwin & Clark, 2000). Fast-paced developments of AI could 
enable the relationship to reach new stages as time progresses. At the moment of the research 
and on this micro perspective between managers and AI one can identify stable characteristics 
that form the relation (Bailey et al., 2022). Generative AI is shown to be missing significant 
human characteristics especially based on tacit knowledge, meaning the embodied knowledge 
through experience. Thereby, tacit knowledge remains an advantage managers have over 
algorithmic learning (Shestakofsky, 2017). Although a visible collaboration with AI is 
identifiable, human colleagues are still considered to be more trustworthy and valued. The 
entangled agencies are not evolving further as interpretations are stabilized for the moment 
(Orlikowski, 1992) and thereby create an interagency relationship (Leonardi, 2011). 
Consequently, trust and solidarity are currently limiting a possible fifth stage of the 
relationship. Yet, technologies are not stabilized for long (Bailey et al., 2022). Through 
adaptations in the relational interactions either by the engagement of developers in altering the 
capabilities of AI or by changes in managerial practices or objectives, the relationship might 
evolve further (Orlikowski, 1992). 
Although management and organization journals have discussed the potential of 
decision-making with algorithmic learning, especially in the case of generative AI (Anthony et 
al., 2023; von Krogh, 2018), managers have not recalled interactions in which AI had agency 
over decision-making. Despite that the interviewees work in different organizations and 
represent a broader field of managers, no one explains a decision based on AI but rather 
highlights how they would not trust AI with decision-making. Although seeking support in AI 
with inspiration is categorized under the decision-making process, it is only the initial step of 
a decision (Doya, 2008). The relationship of validation contradicts existing studies on 
algorithmic learning based decisions in which managers solely rely on the AI output instead of 
reflecting on the suggestions (Lindebaum et al., 2020; Murray et al., 2020). This indicates that 
generative AI is currently not an example of delegated decision-making authority (von Krogh, 
2018). This research suggests that the reason for the limited engagement in decision-making is 
the reduced trust from managers in generative AI within the relationship causing constraints in 
the relation. Not only is trust an important criterion for forming a collaboration but it is also 
crucial for joint decision-making (Henderson & Smith-King, 2015). While trust between 
managers and AI increases within the evolving relationship and finally is strong enough for a 
collaboration it never reaches a sufficient level for decision-making. 
As managers and AI are forming a relationship, there is not always a full development of 
all four stages since some managers for example only perceive the AI relation as a junior 
assistant and it never evolves into a buddy or a colleague. This could be because individual 
managers do not feel that AI is helping them to achieve their objectives. Affordances and 
30 
constraints are based on perceptions of objectives that can affect choices being made for new 
work practices and routines (Leonardi, 2011). The manager’s functional need for AI is an 
individual perception that can vary (Bailey et al., 2022). Through this thesis, we can say that 
the changes in AI relations can come out of reinterpretations of affordances and constraints 
(Leonardi, 2011) but cannot draw any conclusions about how the emergent process occurs in 
specific cases (Orlikowski, 2000). This research implies that the evolving trust is driven by an 
interpretation of fears and risks rather than by harmful experiences. In fact, no interviewee 
recalls a situation with actual negative consequences. Rather trust might be increased through 
positive interactions such as AI providing one with insightful information resulting in increased 
responsibility for the relation. Future research is required to conclude how trust and 
responsibility emerge in specific cases.  
 
Limitations and Directions for Future Research 
The limitations of this study can be found in the research sample. The source of the analysis 
was based on interviews with managers in the tech field looking at one side of the human-AI 
relationship. Although the aim was to identify AI practices equally and to let AI speak through 
managers' reflections, the method showed limitations when it came to foregrounding the 
technology. In the interviews, managers emphasized their personal practices, and it was up to 
us researchers to identify AI practices. This might have led to incompleteness. Research 
suggestions to make AI speak for itself have been identified in the past by analyzing the 
collected data from AI (Cecez-Kecmanovic et al., 2014). However, it was unfeasible for us to 
gain access from Microsoft and OpenAI. Moreover, out of our open interview requests, only 
four female managers in the tech field responded positively. Male and female engagement 
within work relations differs (Merluzzi, 2017) leading to this relationship study representing 
rather the male relationship with generative AI. A more female-focused study would be 
interesting in the future as the entangled relationship might differ.  
This research examines the relational spectrum of managers in the tech field and text-
based generative AI. Building on this relationship study it will be interesting to ascertain to 
what extent the process of forming a collaboration is applicable with image-based generative 
AI or other human actors. This study focuses on the human-AI relation from a micro 
perspective with the potential of further extension to the macro level of organizations 
(Leonardi, 2013b). Thereby, it enables the understanding of whether a partnership is also 
formed within a larger team or even an entire organization, and if the process also shows a 
transition from a tool perspective into a collaboration perspective. According to Leonardi 
(2013b) “micro-level action [that] aggregate[d] into macro-level organizational structure” (p. 
64), suggesting that the formed relationship between managers and AI might transfer into the 
organizational context. Some interviewees highlight that they incorporate generative AI into 
their external team meetings providing an interesting ground for further studies on 
organizational relations. This can be done through a zooming-in zooming-out perspective 
allowing for a dynamic emerging relation study (Nicolini, 2009) that allows for a 
reinterpretation of the results into one organizational context across one team over a longer 
time period. The data suggests that generative AI could extend to organizational relations and 
could become an additional colleague in a team of multiple people. Nevertheless, emerging 
31 
research on leadership suggests that employees react differently to human managers than to AI 
researchers (de Cremer, 2020), making it valuable to review how human-AI collaboration, like 
those in this research, is perceived by both employees and colleague managers.  
 
Conclusion 
Technologies like ChatGPT and Copilot engage with managers in a relational process that 
moves beyond a tool and forms into a collaboration. Allowing the possibility of a human-AI 
partnership indicates a shift in organizational dynamics where colleagues are no longer 
exclusively human but can also include AI-driven entities. For these collaborations to be 
effective, organizations should cultivate trust and responsibility with AI systems. This involves 
paying careful attention to the user interface design, result quality, ethics, and data security. 
Companies should strive to accommodate these factors to support seamless integration. When 
preparing to interact with AI, it is crucial for the manager to reflect on the individual goals of 
the interaction to identify the most suitable type of relationship. Expectation management 
becomes important as it influences perceptions of technology’s usefulness and satisfaction 
(Buschmeyer et al., 2022). A collaborative partnership requires more practices and effort than 
assigning tasks to a digital assistant in the form of a tool, so being intentional about how to 
integrate AI can lead to more productive interactions. Generative AI is not one-sided but 
involves both material and human interaction that have to be managed. Managers need to be 
aware of the emerging trust within the relationship both in terms of developing and mediating 
trust to avoid a blind-sided interaction where AI’s capabilities could be either valued too high 
in the collaboration or too low in the tool perspective. Unlike other technologies with 
predictable usage patterns, generative AI demands nuanced management due to the intertwined 
influence of algorithms and managerial practices.  
The purpose of this research was to understand the formation of a relationship between 
managers and generative AI. Following the literature on collaboration we wanted to gain 
insights into the relationship that is shaped by the interactions and practices. It showed that 
managers and AI initiate a process of four stages to form a relationship that can be seen in 
changing affordances and constraints and evolves into a partnership. Based on a sociomaterial 
framework we investigated the interactions within the relationship. Material and human 
practices, that create an interaction, were present at each stage of the relationship but were seen 
as sequential in the first two and as entangled in the last two. Concluding, generative AI can 
become a colleague in a relationship with the manager shaped by an interaction of agencies. 
Thereby, we have contributed to management studies in three ways. 
First, generative AI proves to be a suitable example of the interwoven human and 
material interactions (Leonardi, 2011). Whilst relational sociomateriality in algorithmic 
learning has been more of a theoretical concept in the past (Bailey et al., 2022; van Rijmenam 
& Logue, 2021) we can identify its relevance in empirical data. Previous empirical research on 
sociomateriality has been focused on affordances and constraints (Cranefield et al., 2022). 
Combining this concept with Orlikowski’s (2000) practice lens allows us to acknowledge that 
affordances and constraints differ not only across context but also in the relationship process 
of an individual manager. This proves that attributes of a tool or collaboration are not inherent 
in the technology but emerging in the interactions (Orlikowski & Scott, 2008). Moreover, the 
32 
initiations of practices which Leonardi defines as a catalyst (2011), are not only seen as being 
human but the study shows that the catalyst for practices can also be in technology. 
Second, this paper acknowledges that managers and AI can engage in a collaboration. 
Contrary to previous research on collaboration (Anthony et al., 2023), this study demonstrates 
that a change in perspective from tool to collaboration is not simply a theoretical construct but 
rather the result of a process of interactions. The manager’s perspective on AI is influenced by 
their intention to actively engage in the relationship. Therefore, research on collaboration needs 
to account for the transition from a tool to a collaborative entity and become more flexible. 
Third, managerial practices can be performed through joint interactions between 
managers and AI in which AI becomes a colleague of the manager. However, this role of AI 
requires a process that the managers actively need to engage in and overcome opacity. This 
stage does not simply happen to the manager, but an intentional entanglement of human and 
material agency leads to this position. Nevertheless, AI continues to show differences between 
a human colleague and cannot fully replace the manager’s colleagues in the “real world”.  
 
  
33 
References:  
Allen, R., & Choudhury, P. (2022). Algorithm-Augmented Work and Domain Experience: The 
Countervailing Forces of Ability and Aversion. Organization Science, 33(1), 149–169. 
https://doi.org/10.1287/orsc.2021.1554 
Anthony, C. (2021). When Knowledge Work and Analytical Technologies Collide: The 
Practices and Consequences of Black Boxing Algorithmic Technologies. Administrative 
Science Quarterly, 66(4), 1173–1212. https://doi.org/10.1177/00018392211016755 
Anthony, C., Bechky, B. A., & Fayard, A.-L. (2023). “Collaborating” with AI: Taking a System 
View to Explore the Future of Work. Organization Science, 34(5), 1672–1694. 
https://doi.org/10.1287/orsc.2022.1651 
Archibald, M. M., Ambagtsheer, R. C., Casey, M. G., & Lawless, M. (2019). Using Zoom 
Videoconferencing for Qualitative Data Collection: Perceptions and Experiences of 
Researchers and Participants. International Journal of Qualitative Methods, 18(1), 1–8. 
https://doi.org/10.1177/1609406919874596 
Bailey, D. E., Faraj, S., Hinds, P. J., Leonardi, P., & von Krogh, G. (2022). We Are All 
Theorists of Technology Now: A Relational Perspective on Emerging Technology and 
Organizing. Organization Science, 33(1), 1–18. https://doi.org/10.1287/orsc.2021.1562 
Balasubramanian, N., Ye, Y., & Xu, M. (2022). Substituting Human Decision-Making with 
Machine Learning: Implications for Organizational Learning. Academy of Management 
Review, 47(3), 448–465. https://doi.org/10.5465/amr.2019.0470 
Baldridge, D. C., Floyd, S. W., & Markóczy, L. (2004). Are Managers from Mars and 
Academicians from Venus? Toward an Understanding of the Relationship between 
Academic Quality and Practical Relevance. Strategic Management Journal, 25(11), 
1063–1074. https://doi.org/10.1002/smj.406 
Baldwin, C. Y., & Clark, K. B. (2000). Design Rules: The Power of Modularity. The MIT 
Press. https://direct.mit.edu/books/book/1856/Design-RulesThe-Power-of-Modularity 
Banh, L., & Strobel, G. (2023). Generative artificial intelligence. Electronic Markets, 33(1), 
1–17. https://doi.org/10.1007/s12525-023-00680-1 
Bankins, S., Ocampo, A. C., Marrone, M., Restubog, S. L. D., & Woo, S. E. (2023). A 
multilevel review of artificial intelligence in organizations: Implications for 
organizational behavior research and practice. Journal of Organizational Behavior, 
45(2), 159–182. https://doi.org/10.1002/job.2735 
Barad, K. (2003). Posthumanist Performativity: Toward an Understanding of How Matter                     
Comes to Matter. Signs, 28(3), 801–831. https://doi.org/10.1086/345321 
Barad, K. (2007). Meeting the Universe Halfway: Quantum Physics and the Entanglement of 
Matter and Meaning. Duke University Press. 
Barrett, M., Oborn, E., Orlikowski, W. J., & Yates, J. (2012). Reconfiguring Boundary 
Relations: Robotic Innovations in Pharmacy Work. Organization Science, 23(5), 1448–
1466. https://doi.org/10.1287/orsc.1100.0639 
Bieda, L. C. (2020, October 13). How Organizations Can Build Analytics Agility. MIT Sloan 
Management Review. https://sloanreview.mit.edu/article/how-organizations-can-build-
analytics-agility/ 
34 
Bowen, G. A. (2008). Naturalistic inquiry and the saturation concept: A research note. 
Qualitative Research, 8(1), 137–152. https://doi.org/10.1177/1468794107085301 
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research 
in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa 
Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., 
Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, 
T., Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter, C., … Amodei, D. (2020). 
Language Models are Few-Shot Learners. 34th International Conference on Neural 
Information ProcessingSystems (NIPS’20), 33, 1877–1901. https://doi.org/10.48 
550/arXiv.2005.14165 
Bryman, A., & Bell, E. (2011). Business research methods. Oxford University Press. 
Brynjolfsson, E., & Mitchell, T. (2017). What can machine learning do? Workforce 
implications. Science, 358(6370), 1530–1534. https://doi.org/10.1126/science.aap8062 
Buschmeyer, K., Hatfield, S., Heine, I., Jahn, S., & Markus, A. L. (2022). Expectation 
management in AI implementation projects: A case study. EuroMed Journal of Business, 
18(3), 441–451. https://doi.org/10.1108/EMJB-10-2021-0161 
Campbell, J. L., Quincy, C., Osserman, J., & Pedersen, O. K. (2013). Coding In-depth 
Semistructured Interviews: Problems of Unitization and Intercoder Reliability and 
Agreement. Sociological Methods & Research, 42(3), 294–320. 
https://doi.org/10.1177/0049124113500475 
Carlile, P. R. (2004). Transferring, Translating, and Transforming: An Integrative Framework 
for Managing Knowledge across Boundaries. Organization Science, 15(5), 555–568. 
https://doi.org/10.1287/orsc.1040.0094 
Carminati, L. (2018). Generalizability in Qualitative Research: A Tale of Two Traditions. 
Qualitative Health Research, 28(13), 2094–2101. https://doi.org/10.1177/104973231 
8788379 
Cecez-Kecmanovic, D., Galliers, R., Henfridsson, O., Newell, S., & Vidgen, R. (2014). The 
sociomateriality of information systems: Current status, future directions. Management 
Information Systems Quarterly, 38(3), 809–830. https://doi.org/10.25300/ 
MISQ/2014/38:3.3 
Charmaz, K. (1995). Grounded Theory. In J. Smith, R. Harré, & L. Langenhove (Eds.), 
Rethinking Methods in Psychology (pp. 27–65). Sage. http://dx.doi.org/10.4135/ 
9781446221792.n3 
Chen, S.-H. (2011). Power Relations Between the Researcher and the Researched: An Analysis 
of Native and Nonnative Ethnographic Interviews. Field Methods, 23(2), 119–135. 
https://doi.org/10.1177/1525822X10387575 
Chen, S.-H., Wu, Z., & Zhao, R. (2023). From fiction to fact: The growing role of generative 
AI in business and finance. Journal of Chinese Economic and Business Studies, 21(4), 
471–496. https://doi.org/10.1080/14765284.2023.2245279 
Chowdhury, S., Budhwar, P., Dey, P. K., Joel-Edgar, S., & Abadie, A. (2022). AI-employee 
collaboration and business performance: Integrating knowledge-based view, socio-
technical systems and organisational socialisation framework. Journal of Business 
Research, 144(1), 31–49. https://doi.org/10.1016/j.jbusres.2022.01.069 
35 
Clougherty, J. A., Duso, T., & Muck, J. (2016). Correcting for Self-selection Based 
Endogeneity in Management Research: Review, Recommendations and Simulations. 
Organizational Research Methods, 19(2), 286–347. https://doi.org/10.1177/ 
1094428115619013 
Corbin, J., & Strauss, A. (2008). Basics of Qualitative Research: Techniques and Procedures 
for Developing Grounded Theory (3rd ed.). SAGE Publications, Inc. 
Cranefield, J., Winikoff, M., Chiu, Y.-T., Li, Y., Doyle, C., & Richter, A. (2022). Partnering 
with AI: The case of digital productivity assistants. 53(1), 95–118. 
https://doi.org/1080/03036758.2022.2114507 
Czarniawska, B. (2004). Narratives in Social Science Research. SAGE Publications, Ltd. 
https://methods.sagepub.com/book/narratives-in-social-science-research 
Czarniawska, B., & Joerges, B. (2018). Robotization—Then and Now (GRI-Rapport 1; 
Research Report). Gothenburg Research Institute. https://gupea.ub.gu.se/handle/ 
2077/56200 
Dang, H., Mecke, L., Lehmann, F., Goller, S., & Buschek, D. (2022). How to Prompt? 
Opportunities and Challenges of Zero- and Few-Shot Learning for Human-AI 
Interaction in Creative Applications of Generative Models (arXiv:2209.01390). arXiv. 
https://doi.org/10.48550/arXiv.2209.01390 
Davenport, T., & Alavi, M. (2023, July 6). How to Train Generative AI Using Your Company’s 
Data. Harvard Business Review. https://hbr.org/2023/07/how-to-train-generative-ai-
using-your-companys-data 
de Cremer, D. (2020). Leadership by Algorithm: Who Leads and Who Follows in the AI Era? 
Harriman House. 
Deng, Y., Bao, F., Kong, Y., Ren, Z., & Dai, Q. (2017). Deep Direct Reinforcement Learning 
for Financial Signal Representation and Trading. IEEE Transactions on Neural Networks 
and Learning Systems, 28(3), 653–664. https://doi.org/10.1109/TNNLS.2016.2522401 
Dixon, J., Hong, B., & Wu, L. (2021). The Robot Revolution: Managerial and Employment 
Consequences for Firms. Management Science, 67(9), 5586–5605. 
https://doi.org/10.1287/mnsc.2020.3812 
Doya, K. (2008). Modulators of decision making. Nature Neuroscience, 11(4), 410–416. 
https://doi.org/10.1038/nn2077 
Duan, Y., Edwards, J. S., & Dwivedi, Y. K. (2019). Artificial intelligence for decision making 
in the era of Big Data – evolution, challenges and research agenda. International Journal 
of Information Management, 48(1), 63–71. https://doi.org/10.1016/ 
j.ijinfomgt.2019.01.021 
Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., Baabdullah, A. 
M., Koohang, A., Raghavan, V., Ahuja, M., Albanna, H., Albashrawi, M. A., Al-Busaidi, 
A. S., Balakrishnan, J., Barlette, Y., Basu, S., Bose, I., Brooks, L., Buhalis, D., … Wright, 
R. (2023). “So what if ChatGPT wrote it” Multidisciplinary perspectives on 
opportunities, challenges and implications of generative conversational AI for research, 
practice and policy. International Journal of Information Management, 71(1), 1–63. 
https://doi.org/10.1016/j.ijinfomgt.2023.102642 
36 
Fabri, L., Häckel, B., Oberländer, A. M., Rieg, M., & Stohr, A. (2023). Disentangling Human-
AI Hybrids. Business & Information Systems Engineering, 65(6), 623–641. 
https://doi.org/10.1007/s12599-023-00810-1 
Faraj, S., & Azad, B. (2012). The Materiality of Technology: An Affordance Perspective. In 
P. Leonardi, B. A. Nardi, & J. Kallinikos (Eds.), Materiality and Organizing: Social 
Interaction in a Technological World (pp. 237–258). Oxford University Press. 
https://doi.org/10.1093/acprof:oso/9780199664054.003.0012 
Faraj, S., Pachidi, S., & Sayegh, K. (2018). Working and organizing in the age of the learning 
algorithm. Information and Organization, 28(1), 62–70. https://doi.org/10.1016/ 
j.infoandorg.2018.02.005 
Fleming, P. (2019). Robots and Organization Studies: Why Robots Might Not Want to Steal 
Your Job. Organization Studies, 40(1), 23–38. https://doi.org/10.1177/ 
0170840618765568 
Floridi, L., & Sanders, J. W. (2004). On the Morality of Artificial Agents. Minds and Machines, 
14(3), 349–379. https://doi.org/10.1023/B:MIND.0000035461.63578.9d 
Floridi, L., & Taddeo, M. (2016). What is data ethics? Philosophical Transactions of the Royal 
Society A: Mathematical, Physical and Engineering Sciences, 374(1), 1–4. 
https://doi.org/10.1098/rsta.2016.0360 
Ford, M. (2014). The Rise of the Robots: Technology and the Threat of Mass Unemployment. 
Oneworld Publications. 
Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to 
computerisation? Technological Forecasting and Social Change, 114(1), 254–280. 
https://doi.org/10.1016/j.techfore.2016.08.019 
Garrison, D. R., Cleveland-Innes, M., Koole, M., & Kappelman, J. (2006). Revisiting 
methodological issues in transcript analysis: Negotiated coding and reliability. The 
Internet and Higher Education, 9(1), 1–8. https://doi.org/10.1016/j.iheduc.2005.11.001 
Gioia, D. (2021). A Systematic Methodology for Doing Qualitative Research. The Journal of 
Applied Behavioral Science, 57(1), 20–29. https://doi.org/10.1177/0021886320982715 
Gioia, D. A., Corley, K. G., & Hamilton, A. L. (2013). Seeking Qualitative Rigor in Inductive 
Research: Notes on the Gioia Methodology. Organizational Research Methods, 16(1), 
15–31. https://doi.org/10.1177/1094428112452151 
Glaser, B. G., & Strauss, A. L. (1967). The Discovery of Grounded Theory: Strategies for 
Qualitative Research. Aldine. 
Glikson, E., & Woolley, A. W. (2020). Human Trust in Artificial Intelligence: Review of 
Empirical Research. Academy of Management Annals, 14(2), 627–660. 
https://doi.org/10.5465/annals.2018.0057 
Henderson, S. S., & Smith-King, E. J. (2015). Sectoral decision making: Structures, processes 
and trust. Management Decision, 53(7), 1545–1559. https://doi.org/10.1108/MD-04-
2015-0128 
Heyder, T., Passlack, N., & Posegga, O. (2023). Ethical management of human-AI interaction: 
Theory development review. The Journal of Strategic Information Systems, 32(3), 1–50. 
https://doi.org/10.1016/j.jsis.2023.101772 
Hofstede Insights. (2023, October 16). Country comparison tool. https://www.hofstede-
insights.com/country-comparison-tool 
37 
Jarrahi, M. H. (2018). Artificial intelligence and the future of work: Human-AI symbiosis in 
organizational decision making. Business Horizons, 61(4), 577–586. 
https://doi.org/10.1016/j.bushor.2018.03.007 
Johri, A. (2022). Augmented sociomateriality: Implications of artificial intelligence for the 
field of learning technology. Research in Learning Technology, 30(1), 2642. 
https://doi.org/10.25304/rlt.v30.2642 
Jones, M. (2014). A Matter of Life and Death: Exploring Conceptualizations of 
Sociomateriality in the Context of Critical Care. MIS Quarterly, 38(3), 895-A6. 
https://doi.org/10.25300/MISQ/2014/38.3.12 
Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at Work: The New 
Contested Terrain of Control. Academy of Management Annals, 14(1), 366–410. 
https://doi.org/10.5465/annals.2018.0174 
Kingma, D. P., Mohamed, S., Rezende, D. J., & Welling, M. (2014, June 20). Semi-supervised 
Learning with Deep Generative Models. ArXiv. International Conference on Learning 
Representations 2014 (ICLR). https://doi.org/10.48550/arXiv.1406.5298 
Korzynski, P., Mazurek, G., Altmann, A., Ejdys, J., Kazlauskaite, R., Paliszkiewicz, J., Wach, 
K., & Ziemba, E. (2023). Generative artificial intelligence as a new context for 
management theories: Analysis of ChatGPT. Central European Management Journal, 
31(1), 3–13. https://doi.org/10.1108/CEMJ-02-2023-0091 
KPMG. (2023, June). Generative AI: From buzz to business value. https://info.kpmg.us/news-
perspectives/technology-innovation/kpmg-generative-ai-2023.html 
Kvale, S. (2006). Dominance through interviews and dialogues. Qualitative Inquiry, 12(3), 
480–500. https://doi.org/10.1177/1077800406286235 
Latour, B. (1987). Science in action: How to follow scientists and engineers through society. 
Harvard University Press. 
Lebovitz, S., Levina, N., & Lifshitz-Assaf, H. (2021). Is AI Ground Truth Really True? The 
Dangers of Training and Evaluating AI Tools Based on Experts’ Know-What. MIS 
Quarterly, 45(1), 1501–1526. https://doi.org/10.25300/MISQ/2021/16564 
Lebovitz, S., Lifshitz-Assaf, H., & Levina, N. (2022). To Engage or Not to Engage with AI for 
Critical Judgments: How Professionals Deal with Opacity When Using AI for Medical 
Diagnosis. Organization Science, 33(1), 126–148. https://doi.org/10.1287/ 
orsc.2021.1549 
Leonardi, P. (2011). When Flexible Routines Meet Flexible Technologies: Affordance, 
Constraint, and the Imbrication of Human and Material Agencies. MIS Quarterly, 35(1), 
147–167. https://doi.org/10.2307/23043493 
Leonardi, P. (2013a). 7 The Emergence of Materiality within Formal Organizations. In P. R. 
Carlile, D. Nicolini, A. Langley, & H. Tsoukas (Eds.), How Matter Matters: Objects, 
Artifacts, and Materiality in Organization Studies (pp. 142–170). Oxford University 
Press. https://doi.org/10.1093/acprof:oso/9780199671533.003.0007 
Leonardi, P. (2013b). Theoretical foundations for the study of sociomateriality. Information 
and Organization, 23(2), 59–76. https://doi.org/10.1016/j.infoandorg.2013.02.002 
Lifshitz-Assaf, H., Lebovitz, S., & Zalmanson, L. (2021). Minimal and Adaptive Coordination: 
How Hackathons’ Projects Accelerate Innovation without Killing it. Academy of 
Management Journal, 64(3), 684–715. https://doi.org/10.5465/amj.2017.0712 
38 
Lindebaum, D., Vesa, M., & Hond, F. (2020). Insights From “The Machine Stops ” to Better 
Understand Rational Assumptions in Algorithmic Decision Making and Its Implications 
for Organizations. Academy of Management Review, 45(1), 247–263. 
https://doi.org/10.5465/amr.2018.0181 
MacKenzie, D. A., & Wajcman, J. (Eds.). (2011). The social shaping of technology (2. ed., 
reprinted). Open University Press. 
Matteson, S. M., & Lincoln, Y. S. (2009). Using Multiple Interviewers in Qualitative Research 
Studies The Influence of Ethic of Care Behaviors in Research Interview Settings. 
Qualitative Inquiry, 15(4), 659–674. https://doi.org/10.1177/1077800408330233 
Merluzzi, J. (2017). Gender and Negative Network Ties: Difficult Work Relationships Within 
and Across Gender. Organization Studies, 28(4), 636–652. https://doi.org/10.1287/ 
orsc.2017.1137 
Michel, S. (2020). Collaborative institutional work to generate alternative food systems. 
Organization, 27(2), 314–336. https://doi.org/10.1177/1350508419883385 
Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016). The ethics of 
algorithms: Mapping the debate. Big Data & Society, 3(2), 1–21. https://doi.org/10.1177/ 
2053951716679679 
Murray, A., Rhymer, J., & Sirmon, D. (2020). Humans and Technology: Forms of Conjoined 
Agency in Organizations. The Academy of Management Review, 46(3), 1–44. 
https://doi.org/10.5465/amr.2019.0186 
Mutch, A. (2013). Sociomateriality—Taking the wrong turning? Information and 
Organization, 23(1), 28–40. https://doi.org/10.1016/j.infoandorg.2013.02.001 
Nicolini, D. (2009). Zooming In and Out: Studying Practices by Switching Theoretical Lenses 
and Trailing Connections. Organization Studies, 30(12), 1391–1418. 
https://doi.org/10.1177/0170840609349875 
Orlikowski, W. J. (1992). The Duality of Technology: Rethinking the Concept of Technology 
in Organizations. Organization Science, 3(3), 398–427. https://doi.org/10.1287/ 
orsc.3.3.398 
Orlikowski, W. J. (2000). Using Technology and Constituting Structures: A Practice Lens for 
Studying Technology in Organizations. Organization Science, 11(4), 404–428. 
https://doi.org/10.1287/orsc.11.4.404.14600 
Orlikowski, W. J. (2007). Sociomaterial Practices: Exploring Technology at Work. 
Organization Studies, 28(9), 1435–1448. https://doi.org/10.1177/0170840607081138 
Orlikowski, W. J. (2010). The sociomateriality of organisational life: Considering technology 
in management research. Cambridge Journal of Economics, 34(1), 125–141. 
https://doi.org/10.1093/cje/bep058 
Orlikowski, W. J., & Scott, S. V. (2008). Sociomateriality: Challenging the Separation of 
Technology, Work and Organization. Academy of Management Annals, 2(1), 433–474. 
https://doi.org/10.1080/19416520802211644 
Pakarinen, P., & Huising, R. (2023). Relational Expertise: What Machines Can’t Know. 
Journal of Management Studies, Preprint, 1-30. https://doi.org/10.1111/joms.12915 
Prasad Agrawal, K. (2023). Towards Adoption of Generative AI in Organizational Settings. 
Journal of Computer Information Systems, Preprint, 1–16. https://doi.org/10.1080/ 
08874417.2023.2240744 
39 
Raisch, S., & Krakowski, S. (2021). Artificial Intelligence and Management: The Automation–
Augmentation Paradox. Academy of Management Review, 46(1), 192–210. 
https://doi.org/10.5465/amr.2018.0072 
Rosenblat, A. (2019). Uberland: How Algorithms Are Rewriting the Rules of Work. University 
of California Press. 
Schwesig, R., Brich, I., Buder, J., Huff, M., & Said, N. (2023). Using artificial intelligence 
(AI)? Risk and opportunity perception of AI predict people’s willingness to use AI. 
Journal of Risk Research, 26(10), 1053–1084. https://doi.org/10.1080/ 
13669877.2023.2249927 
Scott, S., & Orlikowski, W. J. (2012). Reconfiguring relations of accountability: 
Materialization of social media in the travel sector. Accounting, Organizations and 
Society, 37(1), 26–40. https://doi.org/10.1016/j.aos.2011.11.005 
Scott, S., & Orlikowski, W. J. (2013). Sociomateriality — taking the wrong turning? A 
response to Mutch. Information and Organization, 23(2), 77–80. https://doi.org/10.1016/ 
j.infoandorg.2013.02.003 
Scott, S., & Orlikowski, W. J. (2014). Entanglements in Practice: Performing Anonymity 
Through Social Media. MIS Quarterly, 38(3), 873–893. https://doi.org/10.25300/ 
MISQ/2014/38.3.11 
Shestakofsky, B. (2017). Working Algorithms: Software Automation and the Future of Work. 
Work and Occupations, 44(4), 376–423. https://doi.org/10.1177/0730888417726119 
Siltanen, A. (1993). Case technology adaptation and managerial maturity. Information and 
Software Technology, 35(5), 312–316. https://doi.org/10.1016/0950-5849(93)90064-A 
Silverman, D. (2022). Doing qualitative research (Sixth edition). SAGE. 
Stohl, C., Stohl, M., & Leonardi, P. (2016). Managing Opacity: Information Visibility and the 
Paradox of Transparency in the Digital Age. International Journal of Communication, 
10(1), 123-137. https://doi.org/1932–8036/20160005 
Styre, A. (2017). Thinking about materiality: The value of a construction management and 
engineering view. Construction Management and Economics, 35(1–2), 35–44. 
https://doi.org/10.1080/01446193.2016.1272760 
Suchman, L. (2006). Human-Machine Reconfigurations: Plans and Situated Actions. 
Cambridge University Press. 
Terblanche, N., & Cilliers, D. (2020). Factors that influence users’ adoption of being coached 
by an Artificial Intelligence Coach. Philosophy of Coaching: An International Journal, 
5(1), 61–70. https://doi.org/10.22316/poc/05.1.06 
Teubner, T., Flath, C., Weinhardt, C., Aalst, W., & Hinz, O. (2023). Welcome to the Era of 
ChatGPT et al.: The Prospects of Large Language Models. Business & Information 
Systems Engineering, 65(2), 95-101. https://doi.org/10.1007/s12599-023-00795-x 
Tse, T., Esposito, M., Goh, D., & Lee, P. (2024, March 8). Why Adopting GenAI Is So 
Difficult. Harvard Business Review. https://hbr.org/2024/03/why-adopting-genai-is-so-
difficult 
University of Gothenburg. (2023, November 10). Personal data in student projects | Student 
Portal. https://student-prod.apps.k8s.gu.se/en/your-studies/rights-and-responsibilities/ 
personal-data-in-student-projects 
40 
van Rijmenam, M., & Logue, D. (2021). Revising the ‘science of the organisation’: Theorising 
AI agency and actorhood. Innovation, 23(1), 127–144. https://doi.org/10.1080/ 
14479338.2020.1816833 
von Krogh, G. (2018). Artificial Intelligence in Organizations: New Opportunities for 
Phenomenon-Based Theorizing. Academy of Management Discoveries, 4(4), 404–409. 
https://doi.org/10.5465/amd.2018.0084 
Waardenburg, L., Huysman, M., & Sergeeva, A. V. (2022). In the Land of the Blind, the One-
Eyed Man Is King: Knowledge Brokerage in the Age of Learning Algorithms. 
Organization Science, 33(1), 59–82. https://doi.org/10.1287/orsc.2021.1544 
Wang, W., Gao, G., & Agarwal, R. (2023). Friend or Foe? Teaming Between Artificial 
Intelligence and Workers with Variation in Experience. Management Science, Preprint, 
1-23. https://doi.org/10.1287/mnsc.2021.00588. 
Willcocks, L. (2020). Robo-Apocalypse cancelled? Reframing the automation and future of 
work debate. Journal of Information Technology, 35(4), 286–302. 
https://doi.org/10.1177/0268396220925830 
Wilson, J., & Daughtery, P. R. (2018, July 1). Collaborative Intelligence: Humans and AI Are 
Joining Forces. Harvard Business Review. https://hbr.org/2018/07/collaborative-
intelligence-humans-and-ai-are-joining-forces 
Winchatz, M. R. (2006). Fieldworker or foreigner? Ethnographic interviewing in nonnative 
languages. Field Methods, 18(1), 83–97. https://doi.org/10.1177/1525822X05279902 
Zhang, R., McNeese, N. J., Freeman, G., & Musick, G. (2021). “An Ideal Human”: 
Expectations of AI Teammates in Human-AI Teaming. Proceedings of the ACM on 
Human-Computer Interaction, 4(1), 1–25. https://doi.org/10.1145/3432945 
 
 
41