Human Resource Management - 2022 - Einola - Best friend or broken tool Exploring the co‐existence of humans and artificial

DOI: 10.1002/hrm.
22147
SPECIAL ISSUE ARTICLE
Best friend or broken tool? Exploring the co-existence of

humans and artificial intelligence in the workplace ecosystem
Katja Einola 1,2 | Violetta Khoreva 3
1
Stockholm School of Economics, Stockholm,
Sweden Abstract
2
Management and Organization, Hanken Artificial intelligence (AI) has become an important topic in business literature and
School of Economics, Helsinki, Finland
3
strategy talk. Yet, much of this literature is normative and conceptual in nature. How
Management and Organization, Hanken
School of Economics, Vaasa, Finland organizational members perceive AI and the job role changes that come with it is, so
far, largely unknown territory for both HR scholars and practitioners. We sought to
Correspondence
Violetta Khoreva, Management and investigate the relationship between humans and AI and conducted an in-depth
Organization, Hanken School of Economics,
exploratory study into the co-existence of humans and two early-stage AI-solutions,
Vaasa, Finland.
Email: violetta.khoreva@hanken.fi one for “low-status” automation and another for “high-status”; augmentation. We
suggest that different organizational groups may engage in distinctly different sense-
Funding information
Liikesivistysrahasto, Decision of 2020 making processes regarding AI, an important insight for successful HRM strategies
when AI is being introduced into the workplace. Moreover, contrary to recent con-
ceptual work, our findings indicate that AI-enabled automation and augmentation
solutions may not be detached from nor exist in tension with each other. They are
deeply embedded in organizational processes and workflows for which people who
co-exist with the technologies must take ownership. Our findings, in part, go against
discussions on AI “taking over” jobs or deskilling humans. We describe a more
nuanced version of reality fluctuating around the various ways different organiza-
tional groups encounter different AI-solutions in their daily work. Finally, our study
warns against unconditional technological enthusiasm, managerial ignorance of the
nature of work that employees undertake in different organizational groups, and a
neglect of the time and effort required to successfully implement AI-solutions that
affect not only the home organization but also members of the broader ecosystem.
KEYWORDS
artificial intelligence, change—organizational, co-existence, HR and technology, organizational
development, paradox theory, research methods and design—qualitative research methodology,
technology
1 | I N T RO DU CT I O N that is, presently transforming our work lives (Haenlein & Kaplan, 2019;
Huang et al., 2019; Snell & Morris, 2021). In the mainstream literature,
Artificial intelligence (AI) is a modern-day buzzword, a topic of ongoing AI is often conceptualized as having a two-fold purpose: to take over
debate, a tagline for best-selling books, and an essential element of simple jobs or routine tasks from humans (commonly referred to as
emerging, sophisticated technologies in the process of digitalization, “automation”), and to assist humans in more complex tasks, such as
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2022 The Authors. Human Resource Management published by Wiley Periodicals LLC.
Hum Resour Manage. 2023;62:117–135. wileyonlinelibrary.com/journal/hrm 117

1099050x, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/hrm.22147 by Indian Institute Of Technology, Wiley Online Library on [09/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
118 EINOLA AND KHOREVA
data analysis and decision-making (commonly referred to as “augmen- (Kriesberg, 1998) relating to the phenomenon of AI in workplaces in
tation”) (Raisch & Krakowski, 2021). its early stages of development. In this study, we specifically explore
Tasks that are performed by humans versus those conducted by how people and early-stage AI-solutions co-exist, and how different orga-
AI-enabled automation or augmentation solutions (referred to in the nizational members experience the changes brought about by this new
text as “AI-solutions”) are a moving target. This is due to the pace of technology. We contribute to HRM and AI literatures by bridging the
technological change and idiosyncrasies in each workplace ecosystem theory-practice divide and by providing insights into human–AI
(for example, digitalizing a car manufacturing plant is different from dig- co-existence in workplace ecosystems. We depict the various ways
italizing investment banking advisory services). Moreover, what is com- members of different organizational groups encounter different
monly accomplished by humans versus by AI has changed considerably AI-solutions in their daily work and seek to better understand nuances
in the past decades (Brynjolfsson & McAfee, 2014; Brynjolfsson & in employee acceptance of technology (Schneider & Sting, 2020), as
McAfee, 2017; Daugherty & Wilson, 2018; Davenport & Kirby, 2016). well as underlying reasons for their reluctance.
The velocity of change seems only to accelerate, expanding the com- Our initial ambition was to broadly explore how dual AI augmenta-
plexity of the workplace ecosystem where humans and AI need to co- tion and automation implementation unfolds in an empirical setting and
exist, being continuously synchronized to accomplish specific tasks. then narrow the focus down. Our early working assumption concerned
There will most likely be a considerable variety of tasks within this eco- not only tension between automation and augmentation, but also
system, which is a source of richness and value (Snell & Morris, 2021). employee resistance, friction, and uncertainty because we expected that
At the same time, this variety is a breeding ground for challenges. For an AI implementation by nature implies dramatic and immediate changes
humans and AI to successfully co-exist, synchronization is required, not in the division of labor, job roles, and workflows (Davenport &
only among members of a given organization, but also with others, such Kirby, 2016). However, our empirical material led us to challenge these
as customers, supply chain partners, and members of the broader eco- working assumptions that unfold in the pages that follow. An important
system to which an organization belongs. early insight was that, when considering an operational or human per-
Determining how work needs to be organized can be challenging spective, there is less interest in technological distinctions, such as “auto-
in a fluid technology-driven workplace with porous boundaries. Yet, if mation” and “augmentation.” Organizational members encounter
organizations want to leverage the benefits of AI, they need to different AI-solutions and other new tools at work and experience them
become acquainted with variety and complexity and face them up- variously. What matters most is how these encounters are experienced,
front. In our view, it is the task of HR scholars to explore how the not the technological specifications or technology roadmaps created and
implementation of specific AI-solutions unfolds in different workplace maintained by those responsible for designing and implementing them.
ecosystems to gain a broader understanding of the phenomenon, and In line with the interpretive tradition of case studies (Alvesson &
to provide business leaders and HR practitioners with insights into the Sköldberg, 2018; Einola & Alvesson, 2019; Ricoeur, 1981; Stake, 2005,
implications for human resource management (HRM). Our study is 2010), we delve into the sensemaking of different organizational
one step in this direction. groups in a technology-savvy media consultancy agency that, at the
Our research takes as its starting point Raisch and Krakows- time of our fieldwork, was implementing several AI-solutions. While
ki's (2021, p. 203) review essay inviting scholars to apply a paradox some HR scholars tend to aggregate—both conceptually and
theory perspective (Schad et al., 2016; Smith & Lewis, 2011) in their empirically—all organizational members into one comprehensive “work-
empirical studies of AI in organizations to explore “human–machine force” and treat variation among cohorts as noise (Snell &
interactions in real-world settings.” According to the paradox perspec- Morris, 2021), our study explores in detail the dynamics within organi-
tive, AI-enabled automation and augmentation solutions are both con- zational groups. Whereas we uncovered markedly diverging sensemak-
tradictory and interdependent, leading to inherent tension between ing patterns toward AI in different groups (too significant to be
the two (Schad et al., 2016). The complexity that characterizes AI is considered “noise”), we detected neither paradoxical tension between
even greater if we consider that humans and AI-solutions of any type different types of AI-solutions nor dramatic job shifts from humans to
do not take place in separate worlds but, of necessity, co-exist. Any AI, at least not in the foreseeable future. Rather, our research empha-
inherent tension, therefore, does not just happen between types of AI sizes a need for continual learning for a harmonious co-existence to
but inevitably involves humans. Raisch and Krakowski (2021) suggest form between humans and locally implemented AI-solutions, and long-
that organizational members must learn to co-exist with AI, by either term development work for people to comprehend the nature of the
adjusting to the situation or solving the tension, for AI to be success- new technology and to find the perfect “fit” between humans and AI.
fully introduced into the workplace. Hence, understanding how differ-
ent organizational members experience this co-existence is key,
especially for organizations embarking on strategic, long-term AI pro- 2 | LI T E RA T U R E RE V I E W
jects involving multiple interlinked solutions.
We use the notion of co-existence in this study, a concept we will 2.1 | Definitions
further develop and apply, because it makes little sense to talk about
relations between thinking, feeling humans, and inert AI. Co-existence To avoid terminology confusion, we build on long-standing discus-
is a neutral, open concept that can capture a variety of relations sions among scholars to clarify our approach to the phenomena of
EINOLA AND KHOREVA 119
digitalization and AI in this study. The term digitalization was coined to based on “mutually shared foundations” because AI has no human
describe a manifold sociotechnical phenomenon. It is a fundamental qualities, such as emotions, sensorial experiences, intentionality,
process of adopting and applying emerging, sophisticated (or digital) morality, and so on (Moser et al., 2022). Even though the co-existence
technologies at different levels (Haenlein & Kaplan, 2019). These of humans and AI is different from relations between co-existing
technologies include, but are not limited to AI, data analytics, robotics, humans, we need a suitable vocabulary to talk about AI and other
digital platforms, digital twin, social media, digital traces, blockchain, emerging technologies as they become more integrated and interac-
and 3D printing. That is, AI is one of these artifacts which together tive in our workplace ecosystems. The otherwise-criticized quality of
constitute the process of digitalization. openness and the lack of direction in the concept “co-existence” are
Raisch and Krakowski (2021, p. 192) following Nilsson (1971), appropriate to describe interaction between humans and AI, a rela-
define AI as machines performing cognitive functions, which have been tively new phenomenon. We thus define the co-existence of humans
traditionally associated with human minds, such as learning, interacting, and AI in the workplace ecosystem as organizational members interact-
and problem solving. Kaplan and Haenlein (2019, p. 15), on the other ing with AI-solutions, including any kind of contact or bond between peo-
hand, refer to AI as a system's ability to interpret external data correctly, ple and AI generating beliefs, attitudes, emotions, and behavioral patterns
to learn from such data, and to use those learnings to achieve specific once the AI-solution is implemented and as it evolves over time.
goals and tasks through flexible adaptation. AI can also be considered a Here, the idea that human AI co-existence generates beliefs, atti-
shorthand language for a set of complex algorithms for data collection tudes, emotions, and, ultimately, behaviors in employees is central, as
and analysis that makes possible sophisticated predictions and evalua- is the temporal nature of this co-existence in that it keeps evolving.
tion. It is capable of interacting with the environment and simulates or Moreover, members of different organizational groups may encounter
even exceeds human intelligence (Glikson & Woolley, 2020). different AI-solutions and other new tools at work and experiences
Taken together, these definitions imply (rather ambitiously) that them variously. For example, executive and management groups may
AI is flexible, self-adapting, human-like (or exceeding human capabili- encourage frontline employees to use AI-solutions as extensively as
ties); it takes over tasks, engages in learning, can achieve specific possible without themselves being users, and the use of AI-solutions
goals, is capable of making predictions and evaluations, and interacts may mean different things to different people (Amershi et al., 2014;
with the environment. While these statements are broadly true, they Kaplan & Haenlein, 2019). Older employees are often categorized as
apply only to specific instances, tasks, and functionalities. There is no generally change resistant, although this could be negative stereotyp-
generic AI capable of performing all these functions simultaneously. ing and other factors, such as tenure and occupational status, may be
AI-solutions are local, task-specific, and captured by algorithms that more important (Kunze et al., 2013). Schneider and Sting's (2020)
are defined and coded by humans into a language the AI can “under- study on employee perspectives of digitalization-induced change
stand” and follow. These definitions of AI are broad and static and identified five distinct cognitive frames (utilitarian, functional, anthro-
may represent a poor fit in today's fast-pace, manifold workplace eco- pocentric, traditional, and playful) driving employee attitudes toward
systems, thus having limited practical relevance for organizations digitalization. However, employee role-based differences remain
embarking on AI journeys. largely unexplored.
These human dilemmas require trade-offs such that no course of
action is considered universally supreme (Telkamp & Anderson, 2022).
2.2 | The co-existence of humans and artificial However, organizational members at all levels need to trust AI-solu-
intelligence tions, feel empowered to make decisions, and experiment for a suc-
cessful co-existence to form that requires abandoning traditional top-
Co-existence refers to conditions that are the fundamental prerequi- down approaches (Glikson & Woolley, 2020).
sites for the evolution of advanced harmonious relations (Bar-Tal &
Bennink, 2004; Rothstein, 1999; Whittaker, 1999). It denotes recogni-
tion of the right of other groups to exist peacefully with their differ- 2.3 | The paradox theory perspective
ences and to accept others with whom differences need to be
resolved in nonviolent ways. Co-existence promises to provide a The paradox perspective offers a vantage point for HR scholars to
springboard into stronger, more respectful intergroup relations. Co- observe the dynamic interplay and complexity when studying the co-
existence is not a commonly used concept among HR scholars and is existence of humans and AI “in the wild” (Raisch & Krakowski, 2021,
seldom used in comparison with other concepts describing positive p. 204). The essence of this perspective incorporates moving from
relations (Weiner, 1998). The vagueness and indistinctiveness of the simple either/or perspectives to ones that encompass both/and. Rec-
term may be a reason for this, or it may pertain to minimal positive ognizing that AI-solutions, both automation and augmentation, are
intergroup relations only (Weiner, 1998). Hence, co-existence is an contradictory and interdependent means accepting and embracing
open concept “that leaves a great deal of room for various forms of this complexity forming a persistent tension, a situation where the
relations” (Kriesberg, 1998, p.183). authors suggest that organizational members may co-exist with AI
We find it relevant to use the term co-existence between humans only by either adjusting or solving the tension (Raisch &
to also include AI, even though these relations are not (and cannot be) Krakowski, 2021; Schad et al., 2016; Smith & Lewis, 2011).
By working through this complexity, it becomes apparent that site for technology trials. At the time of the study (2019–2020), it had
humans are no longer the sole agents in workplace ecosystems. Using approximately 170 employees and 150 customers in the local market.
AI for various tasks implies that it is not a simple artifact, but a new In the past decades, Wizz, not unlike other companies in the industry,
class of organizational member (Floridi & Sanders, 2004). While AI has has been facing stiffer competition, shrinking margins, and radical
fundamental limitations, its actions nevertheless imply far-reaching changes in customer demand and business operations due to a grad-
autonomy, because humans delegate knowledge tasks to AI and allow ual digitalization of media and convergence of IT and media industries.
it to act on their behalf (Rai et al., 2019). If we consider that automa- It had to rapidly adjust to find new ways to add value for its customers
tion and augmentation are interdependent, this inevitably extends to with revolutionary, advertising-based business models as a conse-
humans working with AI applications. The paradox perspective thus quence of the entry into the industry of social media giants, such as
suggests that humans and AI-solutions do not simply co-exist in sepa- Google and Facebook.
rate worlds working on different tasks, but are interdependent, inter-
acting on the same or closely related tasks. Humans shape AI through
their daily choices, actions, and interactions by defining objectives, 3.2 | Artificial intelligence at Wizz
setting constraints, generating, and choosing the training data, and
providing AI with feedback (Deng et al., 2017). Simultaneously, AI At the time of the study, Wizz was implementing several AI-enabled
shapes human behavior by informing, guiding, and steering human automation solutions. Among them was the company's first rule-based
judgment (Lindebaum et al., 2020; Moser et al., 2022). software robot integrated with a machine-learning module, designed
The paradox perspective thus entails the co-existence of humans to book TV advertising slots. This, at first sight, rather simple automa-
and AI is an iterative or recursive process. Through both automation tion solution, aimed to reduce costs. Employees in the company's tra-
and augmentation, humans and AI become so closely intertwined that ditional media unit2 who specialized in media planning, booking of
they collectively exhibit entirely new, emergent behaviors, which nei- advertising slots, campaign follow-up, and reporting activities origi-
ther entity displays individually (Amershi et al., 2014; Beer, 2017; nally undertook the robot's tasks. At the same time, Wizz was running
Floridi & Taddeo, 2016). Hence, the co-existence of humans and AI several AI-augmentation projects, including a frontstage customer
leads to emerging hybrid workplace ecosystems where it may be diffi- pilot, aimed at helping the sales teams to better plan and execute cus-
cult, or even impossible, to distinguish between humans and AI or tomer advertising campaigns. This solution was technically more com-
their respective learnings and actions. plex than the automation solution. It included machine learning and
This discussion leads to the assumption that the co-existence of advanced live scenario modeling to evaluate the estimated impact and
humans and AI causes important changes in human job roles return on investment of alternative campaign plans. Business man-
(Brynjolfsson & McAfee, 2014; Brynjolfsson & McAfee, 2017; Faraj agers ran alternative campaign scenarios using the algorithm and
et al., 2018). Although domain expertise remains relevant for organi- interpreted the results to customers in real time. Company account
zational members regarding educating and challenging AI, the co- managers and customer representatives undertook these tasks mainly
existence of humans and AI leads to institutionalized knowledge—for on a consultative basis during business meetings. Initially, these two
instance, in the form of AI-solutions—which is often superior to or dif- solutions seemed to be two separate implementations (based on dif-
ferent from individual experts' knowledge (Raisch & Krakowski, 2021). ferent types of AI and run as separate projects by different people).
Simultaneously, general human skills that complement AI, such as cre- Over time it became clear that they were interlinked. Both solutions
ativity, common sense, and advanced communication (Davenport & were part of the same workflow that accrued value for effective
Kirby, 2016), as well as integration skills, such as AI literacy, become advertising campaigns for companies that hired Wizz to design and
important (Daugherty & Wilson, 2018). The co-existence of humans execute.
and AI changes, thereby, (re)-creates and replaces human job roles.
Over time, as organizational members learn to use AI-solutions and, as
the technology is developed further, organizational members learn 3.3 | Research approach
more about AI's limits and possibilities, triggering further changes in
job roles. We were inspired by the same popular business books
(Brynjolfsson & McAfee, 2014; Daugherty & Wilson, 2018;
Davenport & Kirby, 2016) as Raisch and Krakowski (2021) that under-
3 | RESEARCH DESIGN lined the promises of AI, and the dystopian novel Machine Stops
(Forster, 1909) that Lindebaum et al. (2020) relied on, to warn us of
3.1 | Case company substantial differences between human and AI rationalities that, if not
taken seriously, could lead to an apocalyptic system crash. The tone
Our study was conducted in a media consultancy agency1 we shall call of much of this conceptual literature is, however, normative. As a
Wizz, a Finnish-based subsidiary of a leading multinational corpora- more empirically grounded, organizational, practice-based comple-
tion. Wizz is relatively autonomous in the conduct of its business, and ment to these influential studies, we sought to get to know one com-
is acknowledged for pioneering digital advances and for serving as a pany well, developing close ties with its management and employees
to understand the underlying workplace ecosystem from multiple convenient off-the-shelf explanation or a sensemaking shortcut that
points of view, including the AI-solutions being implemented. both scholars and practitioners alike easily espouse when interpreting
Our study follows the interpretive tradition of case studies, seek- an organizational change phenomenon. The traditional media
ing deep understanding of human experience rather than generating employees were not opposed to AI as a matter of principle but felt
rigid cause and effect explanations, as in positivist epistemology neutral or skeptical about yet another new technology that worked
(Stake, 2005, 2010). In the interpretive tradition, research subjects poorly in their pressured work lives. For these employees, AI was not
ascribe meaning to their own behavior. Hence, rather than working a major issue or a disruptive technology, but rather a piece of mal-
with protocolized templates appropriate in realist qualitative research functioning software; an algorithm written by a coder not sufficiently
following positivistic research traditions, this study relies on familiar with their work to get it “right.”
researchers' reasoning processes and reflexivity to enhance the rigor It may be misleading, therefore, to treat all interview statements
of our work and theorizing (Alvesson & Sköldberg, 2018; Einola & as robust data, and simple, unreflective response codification may
Alvesson, 2019; Harley & Cornelissen, 2020). Two inevitably inter- miss the need to make a broader assessment of the situation and
twined criteria apply here: the coherence of the argumentation and the carefully source-check interviews (as well as documents and superfi-
process through which researchers have arrived at what they regard as cial observations based on interviewee impression management
the best explanation for their findings (Harley & Cornelissen, 2020). We behavior targeting the researcher). A hermeneutic, reflective
deemed this choice of research approach the most suitable due to our approach is needed here, instead of a narrow focus on codification
interest in a close, long-term engagement with the case company. In techniques and a straightforward, mechanical sorting and processing
this type of study, a researcher's first-hand experience and pre- of data. We thus actively searched for coherence and worked pro-
understandings (Alvesson & Sandberg, 2021; Ricoeur, 1981) of the cessually, in that, once we heard an argument, we did not assume it
phenomenon studied can be a generative resource for research to be a correct interpretation. Instead, we were inquiring and
design, fieldwork, and interpretation of results, and there is “no need engaged in source critique; we kept asking questions of subsequent
to hide” (Kump, 2022) behind a more rigid and templated approach respondents to verify the coherence of the argument and ensure
that would limit the potential of uncovering surprises and unexpected that statements were universal and coherent. In Appendix B, we give
findings. In this case, an important source of such pre-understandings examples of how we ensured the coherence of our arguments by
was the practical work experience of one of the authors as an IT pro- actively engaging with respondents on an ongoing basis (see also
ject manager in a multinational corporation. Gjerde & Alvesson, 2020) and how we engaged with an interpreta-
Let us give an example of how we used the method criteria in tive approach in general.
practice in our search for coherence of argumentation and the best
possible explanation. In our early talks with executives, it arose that,
while organizational members in customer-facing roles involved in the 3.4 | Interviews and field observations
augmentation solution were enthusiastic, there was considerable dis-
content among traditional media employees regarding the automation We first contacted the CEO for an initial interview to invite Wizz to
solution being implemented in the company. Reluctance to change participate in our study. After several face-to-face meetings to align
and resistance to technology because of a fear of job loss were the our interests and to delimit the scope of the research, the CEO and
reasons executives gave for the slow progress of AI implementation the CTO responsible for the AI implementation gave us full support.
among traditional media employees. Although we found this reason- We were given access to company premises and were free to engage
ing plausible, we were also skeptical. We both formally interviewed with personnel who wanted to participate voluntarily in our study.
and informally interacted with traditional media employees to under- The study was conducted primarily through 37 in-depth semi-
stand their feelings about the ongoing AI implementation. A better structured interviews, which were recorded and transcribed verbatim
explanation for resistance appeared to be employee reluctance to (see details in Appendix A). Anonymity of the interviewees was guar-
take on yet another new technology that worked poorly in its present anteed. Following Welch and Piekkari's (2006) recommendation, inter-
form combined with overall change fatigue following a long career in views were matched, so that the respondent's preferred language was
the field. This reluctance was exacerbated by mounting frustration used in all cases. Hence, 29 interviews were conducted in Finnish and
with having to work with a malfunctioning robot supposedly designed eight in English. This way of working yielded richer and more accurate
to automate their tasks but which, instead, caused employees more data than would have been possible using English only, and helped to
work. We detected no fear of job loss; there was plenty of work at establish a better rapport between interviewers and interviewees, and
that moment, employee turnover in the industry had always been maximized the authenticity of interview accounts. Both of the
high, and no one could remember anyone who had retired from the research team members, academically qualified speakers using English
field. as their working language (see Harzing et al., 2011 for a similar proce-
We, therefore, had to abandon what, at first, seemed to be the dure), conducted the interviews in Finnish which were then translated
likeliest explanation for the slow progress and opt for an alternative, into English. Such translations tend to be of a higher quality than
more comprehensive reason (Harley & Cornelissen, 2020). “Reluc- those of external translators who are not experts in either the topic or
tance to change” and “technology resistance” are examples of a the organizational context.
FIGURE 1 Data collection and analysis
The interviews were conducted face-to-face and lasted between 3.6 | Data collection
24 and 100 min, and totaling 36 h of conversation. In addition to
the interviews, numerous field observations were made that guided Our on-site fieldwork started with two interviews with the CEO and
the research process. Since both researchers spent a significant the CTO. These organizational members were driving the AI imple-
amount of time at Wizz, 150 h each over a period of 6 months, mentation as a strategic initiative, an important part of the study con-
informal encounters, electronic correspondence, conversations, and text to understand before digging deeper into the fieldwork. During
casual contacts with different organizational members were frequent. these interviews, the executives discussed issues such as business
Off-site data (from e-mails, chats, phone calls, and following Wizz in challenges, the company background, corporate technology strategy,
the news) were also gathered, mainly from executive management and its relation to business goals, the emergence and overall relevance
before and after our formal fieldwork. Our methods of data collection of AI for the industry and the company, and key HR issues. We con-
and analysis are illustrated in Figure 1. sidered these initial encounters that took the form of informal discus-
sions as valuable background information and a helpful starting point
to structure our later, more formal interviews with these executives
3.5 | Organizational groups and discussions with other organizational members.
The AI implementation began shortly after these interviews. We
The sample was initially divided into three groups: (top) executives, conducted 35 interviews, adapting our questions and interview style
(business) managers, and employees. Executives were the company to each interviewee to show that we were interested in how they
CEO, CTO, and CFO. Managers were individuals with business and/or made sense of the ongoing process. We asked questions (repeatedly
overall profit and loss responsibility for customer accounts. Employees on many occasions to make sure our understanding was correct)
were experts responsible for planning and executing customer media about the nature, the need, the scope, the purpose, the benefits, the
campaigns. drawbacks, and the doubts regarding the newly implemented AI-solu-
The employee group was subsequently divided into two sub- tions. We inquired about job changes the interviewees were
groups: digital media and traditional media employees. This separation experiencing and whether they had concerns about their competence
occurred because, based on our early interviews and observations dur- profiles or future job prospects.
ing the first fieldtrips, we noticed that sensemaking around AI differed
significantly between members of these two groups, and we wanted to
understand the reasons behind the differences. Even though not all 3.7 | Data analysis
members in each group made sense of the ongoing AI implementation
in identical ways, perceptions, experiences, and attitudes toward the Our data analysis followed an interpretative method relying on
process converged within each group. A careful analysis of our research (a) inductively generated meanings, (b) further explorations of these
material pointed to differences around the meanings members of dif- using hermeneutic principles, (c) crosschecking reliability of interview
ferent groups gave to AI, metaphors they explicitly or implicitly used to statements via further interviews and field observations, and
characterize the new “AI colleague,” the type of change AI provoked, (d) researcher reflexivity and skepticism toward both our own pre-
and the way they framed the emerging co-existence with AI. These understandings as well as field observations and interview statements
strongly job-role based differences became our main themes that we (Einola & Alvesson, 2019; Einola & Alvesson, 2021; Gjerde &
continued to refine during our fieldwork. Alvesson, 2020). This type of analysis offers an alternative method to
an increasing tendency for qualitative research to systematically code and including the introduction of what executives referred to as AI
data, a procedure that is not always appropriate for all types of quali- “colleagues” into the workforce.
tative data or for all qualitative traditions. Coding can sometimes be a
static, reductionist, mechanistic process that detaches pieces of data
from a broader understanding while downplaying the context 4 | FI ND I NG S
(Alvesson & Sköldberg, 2018; Bansal & Corley, 2011; Potter &
Wetherell, 1987), building conceptual walls between validation and 4.1 | Sensemaking dimensions
discovery (Locke et al., 2015) that can constitute what Maxwell and
Miller (2008) call analytic blinders, preventing the researcher from see- We analyze here the main themes that we call sensemaking dimensions
ing alternative relationships in the data. intended to help us understand the root causes of the perceived dif-
We engaged in a careful hermeneutic reading and re-reading of ferences between the ways organizational members made sense of
our material as it accumulated, gradually conveying a feeling for the their co-existence with AI in terms of: (1) meanings, (2) metaphor,
parts (i.e., interviewee narratives as people made sense of their own (3) type of change, and (4) characterization of co-existence as set out
experience) and the whole (i.e., combining the individual level narra- in the table below (see Table 1).
tives with group narratives as the “organizational group” emerged
within a meaningful analytical category (see, for instance,
Ricoeur, 1981). We made constant efforts to make sense of specific 4.1.1 | Meanings
examples of material that seemed to be representative of significant
parts of the whole (some of these are presented as quotations in the Members of different organizational groups gave different meanings
section that follows), while conveying important clues of something to AI. For executives, the success of the AI implementation was closely
specific, deeper, and less obvious that needed to be discovered, such linked to the long-term viability of the business, and they continually
as the reasons behind the “reluctance to change” and “technology emphasized the overall role AI had in the continued success of the
resistance” in the traditional media group discussed earlier. Thus, our company. Managers responsible for customer accounts, on the other
interpretation is based on a dynamic combination of cues rather than hand, gave meaning to AI through the lens of increased customer sat-
a uniform set of codified material presented as static, clear-cut isfaction and saw business opportunities more in the short term.
categories. According to the managers, AI helped to strengthen the loyalty of
We took notes of our thoughts and observations during and after existing customers and build long-lasting relationships with new
company visits and interviews. These notes were a starting point for clients:
our discussions that occurred directly after each site visit, the purpose
of which was to share our fresh impressions of what we had seen and “The extra value that our customers receive comes
heard and to map topics that we thought were important to under- from the analysis we are able to deliver, which comes
stand but where our interpretations differed or that we had doubts from the cooperation of our people with new AI-solu-
about (Locke et al., 2008). One such topic was the company technol- tions” (Development Manager).
ogy roadmap. Even though this study is not about technology imple-
mentation per se, following our hermeneutic approach, we needed to “Our clients are happier [...] together with these new
understand the overall company technology strategy that encom- tools we do advanced scenarios of return on invest-
passed many solutions and phases, of which the AI-solutions we stud- ments and this helps us to be more strategic [...] Our
ied were part. Throughout our engagement with the company, we clients think we deliver better quality than before”
kept going back to the CTO to refine our understanding of this road- (Managing Director 2).
map. With time, we learned, for instance, that the two AI-solutions
we had decided to focus our study on, were in fact not separate The focus here is implicitly on augmentation, even though organiza-
implementations, but part of one workflow, an important insight to tional members other than the executives never used the word. The
problematize what the literature had characterized as “paradoxical interest is on “cooperation” between AI and humans (or augmenting
tension” between automation and augmentation. Both researchers human capabilities with AI), leading to customer satisfaction and
read and color-coded the interview transcripts for possible themes or increased sales: both important organizational goals. Thus, AI
interesting insights and regularly discussed possible interpretations enhances human intelligence and positive outcomes, which automati-
and ideas as the data collection and analysis progressed. We regularly cally follow when humans use their time more productively. The
cross-checked our findings with some of the participants. absence of automation and any challenges associated with it, in this
Whereas the finding that different organizational groups made rhetoric, indicate a blind spot, especially when the augmentation solu-
sense of the AI-solutions differently may not be surprising per se, the tion itself is based on heavily automated processes that have implica-
extent of these differences was striking for a company embarking on tions for developing “advanced” augmentation processes.
an ambitious AI implementation project involving the whole work- For digital media employees, AI represented work life as it always
place ecosystem, changing competence requirements and job profiles, had been and was perceived as an already well-integrated part of their
TABLE 1 Group member sensemaking of AI
Traditional media
Executives Managers Digital media employees employees
Meaning given to AI is an inevitability AI is an opportunity to AI is an integral part of the AI is yet another technology
AI “These days people consume increase sales by helping work itself “This [the AI-solution] is once
content differently, search customers increase their “I spent most of my time again one of those things
information differently, buy sales dealing with media and that lasts for maybe about
goods and services “With these advanced tools technology [...] AI is a year, and after that a
differently, so we need to be [AI-solutions] we are able naturally present in my new technology will come”
proactive and implement all to boost our customers' work” (Online planner 2). (Operator 1).
these AI tools” (CEO, 1st sales” (Strategist 3).
interview).
Metaphor used to AI as savior, junior colleague AI as horn of plenty AI as friend, closest AI as broken tool, stupid co-
characterize AI “[Name of the robot] is a junior “I find it great that we use colleague worker
and we see a lot of potential these advanced tools [AI- “Technology is naturally “[Name of the robot] cannot
in him” (CTO, 2nd interview). solutions], they [AI- present for me [...] it's my work alone, he is like a
solutions] are like horn of closest colleague” (Social broken tool, he always
plenty for us, they [AI- Media Strategist 2). needs people to boss him
solutions] help our business around” (Printing Planner).
to be more profitable and
effective” (Investment
Director 1).
Type of change Major disruption, Exciting opportunities Work as usual More work/ double work
triggered by AI transformation of business “Our clients are excited and “New solutions are popping “There is even more manual
“With this major change [AI happy with the current up all the time [...] in my work that feels meaningless
implementation] we are able changes [...] we are able to work it is very much work- as I am not yet getting
to explore and take shape our client's needs in-progress” (Online benefits from automation”
advantage of the benefits” better” (Head of Digital Planner 1). (Operator 2).
(CFO). Marketing).
Characterization Co-existence is exciting Co-existence is instrumental Co-existence is natural Co-existence is frictional
of co-existence “I find this tandem [co- “Now when we work so “AI is like the air you breathe “[Name of the robot] is not
with AI existence of humans and AI] closely with these [AI] [...] It is something that well developed yet so he
meaningful and rather solutions, we can increase makes things easier and cannot do what we
thrilling” (CEO, 3rd sales and also bring more helps a person's life” need him to do […]”
interview). value to the customers” (Online Planner 1). (Operator 2).
(Investment Director 2). “I've seen so many new tools,
I had to learn so many of
them” (TV Planner 4).
work. For traditional media employees, however, AI was yet another 4.1.2 | Metaphor
tool waiting to be proven in the long continuum of technology-related
“mandatory changes” (TV Planner 1) they had already undergone in Different organizational groups used different metaphors to portray
the past decade under the label of digitalization: AI. Executives described AI as a “savior of business” when they dis-
cussed the promises of the new solutions in general terms. When they
“Change is an integral part of our industry. Therefore, spoke about specific implementations that were ongoing, the tone
it does not feel like a huge step that oh, wow, now was more affectionate and AI was often referred to as a “junior
things are digitalizing. Fifteen years ago, we were still colleague,” as a way of accepting problems but also excusing this new
using faxes [...] the change is constant [...] it does not colleague's poor performance. Managers thought of AI as a “horn of
feel like now this new AI stuff is something completely plenty” or a way to tap into both human creativity and technology to
ground-breaking” (TV Planner 1). attract customers and increase profits. Digital media employees saw AI
as a friend or close colleague, helpful for doing a better job and with
For these experienced employees, the ongoing change was incremen- whom it was fun to work:
tal at best. The tone was clearly condescending with no excitement or
overall sensation that the industry had suddenly embarked on new AI- “AI is my right hand, my friend […] it automates many
triggered ways of doing work that would revolutionize “everything.” things that I would have to do otherwise by hand. For
The focus of these employees was not on the future potential of AI example, once the offer is placed in Google, I don't
but on lived difficulties with the ongoing project. have to modify it or do anything with it since [name of
the solution] does it for me. I just pilot and keep an eye turn, would motivate them to design more effective advertising cam-
on what's happening” (Search Engine Advertising paigns and allow managers new perspectives for marketing and ser-
Specialist). vice development:
When these employees spoke about their work collaborations, they “We are able to put our hands on the machine-learning
mentioned AI and other tools more often than their human colleagues modules more than ever before. Together with this
working on the same customer deals: technology, we come up with new solutions, for exam-
ple, how to make a campaign better and increase cus-
“The only bad thing with AI is that you cannot go for tomer sales […] This way we are able to bring more
lunch with it” (Digital Designer). value to our customers” (Insight Manager).
In contrast, traditional media employees thought of AI as a tedious, bro- “Technology actually highlights the intellect of our
ken tool, a trouble- or friction-maker, and the least-preferred col- people because not everything can be done by
league. These employees mentioned inconsistency when using the machines. It's only the unique combination of human
newly implemented automation solution and expressed doubts about intellect and technology that can be of value to us”
its reliability and trustworthiness. They criticized the robot for not (Director, Digital Services).
being sufficiently intelligent, making many mistakes, and requiring
more supervision and control than if they did the work manually: AI here is an intelligence-improver and a creativity-booster, and
human/AI co-existence is a source of value that enhances human
“[Name of the robot] is quite stupid. He does not capabilities. The entire focus is on augmentation, and the underlying
understand everything. He bangs his head on the wall. automation that must work if the whole workflow is to function is left
When we use [name of the robot] for our campaign, it unmentioned.
takes so much time and effort to supervise and guide For digital media employees, the ongoing change about which the
him” (TV Planner 2). executives and managers spoke was invisible, and the division
between automation/augmentation and human/AI was meaningless.
Whereas managers and executives focused on the future benefits that These employees talked about a need to autonomously learn the solu-
come with AI, and digital employees emphasized their present, posi- tions. This not only took a significant part of their time but also made
tive co-existence with it, traditional media employees anthropomor- them highly specialized in a narrow business domain. For them, AI
phized AI to reflect their frustration, calling the robot a “he,” using was a natural outcome of digital technology development and a work-
terms such as “stupid,” and seeing the AI as “someone” in need of in-progress and working without AI was not a possibility. Conversely,
better supervision and guidance. traditional media employees seemed either unconcerned with or
unaware of what was happening with technology development at cus-
tomer sites and frontstage, instead raising concerns about ongoing
4.1.3 | Type of change automation efforts that directly affected their work. These employees
did not experience the challenge and thrill of working in tandem with
Executives talked about major change that was going to influence not AI, and the ongoing automation required them to constantly fine-tune
only Wizz but also the industry generally and the local ecosystem tedious process descriptions. The passage from working the old way
within which the company was embedded. Change was necessary to the new one was labor-intensive and frustrating. While traditional
with each customer, content provider (advertising agencies), and media employees acknowledged that the purpose of the newly imple-
media partner because augmentation solutions required an alignment mented automation solution was to reduce manual work, in practice,
of processes between different parties. The executives were entirely not only did the amount of manual work not decrease, but also in
focused on these macro-level changes and the promise of the aug- many cases doubled:
mentation solution being implemented at the time of our research.
They did not talk about automation and augmentation, or the differ- “It almost feels like the amount of manual work has
ent AI-solutions being implemented as part of the same phenomenon, actually increased because new systems need to be
or as inseparable parts of an overall corporate technology strategy. tested and it takes hours to find ways to improve
That these solutions were parts of a bigger whole became evident them” (Operator 2).
only from the technology roadmap, a company artifact we learned to
read over time with the help of the CTO. Managers working with cus- It could be argued that these issues were temporary, but we under-
tomer accounts were almost entirely focused on ongoing customer stood from employee descriptions of their workflow that the number
pilots, man–machine cooperation of a new kind intended to increase of media with which they were working was extensive (multiple TV
sales. According to these managers, AI would provide employees with and radio channels, billboards, etc., each with its own processes and
more room to maneuver with a wider variety of options, which, in legacy systems), all of which were digitalizing at their own pace. The
alignment between parties required significant effort, extending far employees consistently underlined the importance of broad knowledge
into the future. The long-term nature of the change, the difficulty of (of industry, the economy, different media, customers, trends), that they
planning and coordinating with channel members, and the need to felt the more production- and technology-oriented digital media
foresee the future constituted risk factors that did not arise in the rhe- employees lacked. For them, co-existence with AI was high maintenance,
toric of the executive members with whom we spoke. requiring great effort to make AI fit existing work processes. The imple-
mentation of AI for them was part of a long continuum of technology
projects, the transformation of the media industry overall, and just
4.1.4 | Characterization of co-existence with another entry into their ongoing technology-related learning journey:
Artificial intelli
“I don't see this AI implementation as a dramatic
Different organizational groups defined the nature of their co- change, rather little by little things do change [...] This
existence with AI differently. The executives described the co- process [AI implementation] causes stress because it
existence as intimate, close, personally meaningful, romantic even. comes with new information and new knowledge”
They highlighted the importance of human characteristics in this new (Printing Planner).
kind of collaborative tandem. In this type of idealized sensemaking,
harmony and outcomes of alignment and interplay were underlined as These employees were aware that management considered them as
key and expected to happen naturally: change resistant, which clearly angered them. They explained how
they had adjusted to all kinds of new tools since the beginning of their
“Tech itself is kind of useless if you do not have people careers, including a myriad of more or less functional tools they had to
to work with it […] You need the combination of crea- work with daily. For them, AI was no different.
tivity and curiosity to use the tech and enhance your These different lived experiences of co-existing with AI brought
own knowledge through it. Machine plus person is the organizational silos and friction between groups that ultimately under-
best” (CTO, 2nd interview). mined and delayed AI implementation. These “parallel realities” con-
tributed to dissatisfaction between traditional media employees and
For managers, this co-existence was pragmatic, in that they saw the their managers and caused friction in their relations.
need to swiftly integrate AI into employees' work. They portrayed AI
as instrumental; a means to boost employee creativity and firm capa-
bility to deliver what the customer needed. Digital media employees 4.2 | Job role changes
felt they naturally co-existed with AI:
People in all organizational groups experienced job role changes stem-
“It's very natural for me to co-exist with [name of the ming from ongoing AI implementation (see Table 2). These changes
solution]. The only reason I am able to do my job is due were continuous and hard to plan for because their precise impact on
to having [name of the solution]” (Online Planner 1). job roles and skills was uncertain. They occurred in the organizational
silos, further compartmentalizing the sensemaking of organizational
For these employees, AI was not a separate entity but an integral part members and pulling different groups in different directions depend-
of their work routines and an enabler for doing their job. ing on the imperatives of their own jobs. Nothing united them within
Traditional media employees were less enthusiastic and indifferent the workplace ecosystem, not even the technology roadmap pre-
to co-existence. While the goal of executives and managers was to sented to all employees but poorly understood.
allow traditional media employees to spend time on more “exciting,
creative” (TV Planner 4) tasks, these employees were nostalgic for
routine and standard tasks: 4.2.1 | Executives
“I find it nice sometimes to be able to spend a couple As a direct consequence of the ongoing AI implementation, executive
of hours just doing the kind of job which one can run attention and time had shifted from operational and managerial work
without extensive mental capacity. Making reports and to developing and implementing the company's technology strategy,
printing them out is a kind of manual work which is and creating and nurturing the nascent umbrella ecosystem of technol-
nice sometimes. No one has the energy to be creative ogy partners, customers, and others. From our discussions and interac-
the whole working day” (Operator 1). tions with the CTO, we noticed that much of his time was spent off-
site working with technology partners, thus reducing the time he was
Employees in this group raised the importance of “slack,” or empty time, available to support Wizz employees in the transition. The CEO was
to allow them to generate ideas and be creative. This is an important con- portrayed in newspaper articles as an industry pioneer brokering new
sideration in the consulting industry that depends on creativity, an aspect types of deals—and, in this sense, AI was a means of branding the com-
that productivity-maximizing managers appeared to overlook. These pany as an industry leader. These outreach activities were aligned with
TABLE 2 AI-triggered job changes and key focus areas for different organizational groups
Executives Managers Digital media employees Traditional media employees

(1) Development and (1) Communication, interpretation, (1) Vertical job specialization to (1) Testing and de-bugging
implementation of the and “translation” of outputs of master the company AI-solutions automation software (“helping”
company technology strategy new sales-generating AI- and proprietary tools of social the AI-solution to improve)
and roadmap solutions to customers media giants “I end up helping this silly robot
“We see now where technology is “It does not matter how many “It takes a lot of time and effort for myself” (TV Planner 2).
heading and we steer our robots and other AI-solutions we me to learn each new [AI] solution (2) Fixing the work-related
company to this direction” have, machines cannot do [...] I wish I had more time to do problems of the
(CEO, 3rd interview) consulting and communicating, it's something else” (Online Planner 1) underperforming AI-solution
(2) Creation and maintenance of only us who can make the right (2) Autonomous learning to be able “In the future [Name of the robot]
the overall company decisions and become customers' to work with AI-solutions will hopefully learn how to build
ecosystem (headquarters, strategic partners” (Strategist 2) “No one can explain to me in detail better campaigns, but for now he
technology partners, (2) Strengthening the account how these [AI-solutions] tools are is not that skillful […] I have to
customers, content providers, teams' competence to work with working, I have to figure this out overtake his tasks” (Operator 2).
media, etc.) AI by myself” (Social Media
“Now we have even more “We are creating new ways of Strategist 1)
meetings with our technology working to encourage our people
partners, we have to make sure to work with latest technologies”
we bring in the results” (CTO, (HR Director, 2nd interview)
4th interview)
the executives' sensemaking of AI as a means of ensuring company via- difficulties that employees experienced were simply explained away
bility that with which all employees remaining in the company had to as resistance to change.
learn to work. The executives expressed both optimism and pessimism
about employees' reactions to the role changes:
4.2.2 | Managers
“Many people think that after they learn the basics,
they'll just work that way until they retire. But that's not The job changes for managers were the ones related to customer con-
how you would ever have a long career in our business sulting for campaign planning and execution. Managers and others
[…] People at [name of the company] learn all the time involved in sales needed to learn new skills to communicate, interpret,
how to work with these new AI-solutions, how to build and “translate” (Online Planner 1) machine outputs to customer solu-
better solutions for the clients together with these new tions. Working with frontstage AI-solutions required interaction with
non-human junior colleagues” (CEO, 1st interview). the new solutions in live situations in the presence of customers,
which was both exciting and demanding. Managers saw these job
“Some of this ‘busyness’ reminds me of the picture of changes as moving from a simple interpretation of machine data to
the stone age man pushing the cart with rectangular broader strategic planning requiring deeper business insights and cus-
wheels and being too busy to invent circular wheels tomer understanding. Hence, AI implementation highlighted man-
[…] We built this new robotic process [automation agers' customer-facing role:
solution] they [traditional media employees] need to
use for booking advertising slots. But they still book “With the help of advanced live scenario modelling, we
everything in their old way, they fill in an Excel [sheet] are able to add more value to what we produce for our
that calculates numbers which they put in manually [...] customers […] In tandem with the recently implemen-
The new way requires them to use a new Excel [sheet] ted AI-solutions, we say to our customers what will
that [...] produces an algorithm for the robot [...] It's like work and what won't” (Strategist 2).
moving from an old car with a mechanical gear to an
automatic one. I'm amazed that people just won't do These changes represented a move up the value chain and were a
this small thing and we have to go through all this over source of both excitement and worry about whether account teams
and over again” (CTO, 2nd interview). could meet the challenge.
We see here how implementation of the automation solution, ostensi-

bly a small part of the overall technology strategy, was central from 4.2.3 | Digital media employees
the executive perspective. They did not consider possible reasons to
account for the slow progress other than employee resistance in mov- Perhaps counterintuitively, due to the increased technical sophistica-
ing from “using a mechanical gear to using an automatic one.” The tion of the AI-solutions, the job roles of digital media employees
narrowed, becoming more specialized and technology focused. The not see its benefits when it comes to how it can help
pace of technology and channel change was so fast that there was lit- my work […] Quite often we have specialists who
tle time for broadening the digital media employees' compe- come here to develop our technologies and stuff like
tence base: that, but they have no idea what our goals are. They
develop all sorts of systems that would, for example,
“We use AI-solutions that optimize the campaign for book a slot on the TV automatically. That's nice, but
strategies, target groups, audiences, different ad for- often this system doesn't work whatsoever and dealing
mats […] Tools give the numbers and everything else. with the system takes up a lot of time. I think some of
Then I need to double check everything and present those systems are just useless. I think that automating
the results to the customer who does not necessarily this slot booking process is not necessary”
understand this digital stuff […] I need to translate the (Operator 1).
terms into human language and so I don't have much
time for anything else” (Online Planner 1). This employee account reflects a common problem in top-down tech-
nology projects: lack of end-user involvement and poor operational
There is a dilemma here: broader knowledge of the overall business is understanding by those who design the solutions. This lack of knowl-
required to design more sophisticated customer solutions, but the edge is then transferred to employees as extra work causing frustra-
time is scarce and keeping up with technology takes significant tion about working with a “broken tool.”
amounts of employee time.
5 | DI SCU SSION
4.2.4 | Traditional media employees
To bridge the theory-practice divide, this empirical study explored the
For these employees, the shift to AI-solutions in their daily routine fluctuations of human and early-stage AI-solution co-existence in four
was very slow. It started with what executives (and managers) organizational groups within the workplace ecosystem of a
assumed to be a rather straightforward automation of a simple task. technology-savvy company. Our findings remind us that today's AI is
However, when automation came to the employees' desks, the far from the imagined AI portrayed in sci-fi movies and futuristic
mechanical tasks were not simple, nor were they easy to model and novels of a 100 years ago (see Lindebaum et al., 2020). The organiza-
automate. Even a simple-sounding “booking of advertisement space” tional reality is more mundane. Members of different organizational
task was complex because of the many variables and parameters to groups differed significantly when it came to the meanings they gave
consider, and the response needed from other members of the work- to AI, the metaphors they used to describe AI, the type of change AI
place ecosystem. represented, and their characterization of human-AI co-existence.
In the ongoing AI implementation, traditional media employees' These differences slowed AI implementation and increased the mutual
focus shifted to testing, fixing, and working with the “stupid” automa- frustration of managers and employees, rendering the human-AI co-
tion robot (TV Planner 2), instead of handing tasks over to existence as frictional.
it. Traditional media employees' tacit knowledge of colleagues, cus-
tomers, and business cycles was essential to solving deviations, thus
making the automation solution look like a mixed automation- 5.1 | Theoretical contributions
augmentation application requiring employees' regular intervention.
HR representatives, executives, and managers largely ignored this We present our key theoretical contributions and implications in this
type of knowledge. section (see Table 3 below).
The initial phase was lengthy, as traditional media employees and
AI-solutions executed overlapping tasks and learned to co-exist.
Based on user feedback, the AI-solutions needed to be refined by 5.1.1 | No paradoxical tension between automation
coders working for a technology partner inexperienced in the media and augmentation
industry or the types of jobs the employees were doing. Human users
then needed to learn to work with the newly fixed AI-solutions. This Our study only partially supports Raisch and Krakowski's (2021) con-
required time and effort from traditional media employees. Contrary ceptual work based on the paradox perspective that automation and
to executive and management expectations, traditional media augmentation are not essentially separable and conflicting, but funda-
employees did not feel that their jobs would change in the foreseeable mentally interdependent. Our findings reveal that there is no automa-
future: tion here and augmentation there; it is not either/or. Rather, both
automation and augmentation are “everywhere” and constitutive of
“I think that I might be one of those people who each other. The two AI-solutions we studied in detail—initially
doesn't appreciate AI that much and can sometimes described as AI-enabled automation robot for booking advertising
TABLE 3 Theoretical contributions and implications
Theory
domain Authors Key propositions Findings of the present study
AI
Brynjolfsson and McAfee, Automation and Lack of evidence that in HRM related research aiming at building
(2014, 2017); Daugherty augmentation as distinct practically relevant and empirically grounded outcomes this
and Wilson (2018); phenomena distinction should be kept (at least in knowledge-intensive contexts).
Davenport and Kirby
(2016)
Raisch and Krakowski (2021) Automation and No support for the “paradoxical tension” proposition when taking a
augmentation as task/business-driven approach.
complementary phenomena Tension and miscommunication are located between humans, not
but with paradoxical tension between AI-solutions of different kind.
in between
HRM
Brynjolfsson and McAfee AI leads to massive job losses There may not be massive job losses when adopting an industry/
(2014); Glikson and organization-specific approach.
Woolley (2020); Zanzotto AI-triggered changes may occur in the long-term so that they blend
(2019); Zuboff (2019) with the “usual” employee worries about job safety and are absorbed
by the more encompassing changes related to digitalization.
Raisch and Krakowski (2021); Al leads to deskilling of Deskilling may not occur as expected, at least not in the short term,
Salla et al. (2018); Sutton humans due to tacit knowledge not being formally acknowledged before AI
et al. (2018) implementation.
Upskilling (i.e., gaining technology-related knowledge) may come at the
cost of missing other knowledge (i.e., understanding of broader
business issues, social skills, undermining the value of the overall
accumulated human work experience).
Raisch and Krakowski (2021) AI helps in re-humanizing From a well-being perspective, employees may enjoy/need some
work “slack” in terms of some tasks being repetitive; “brainless” time may
be beneficial for employee creativity.
Kim et al. (2021); Snell and Expose the need to highlight A suggestion to expand the HR ecosystem and technology
Morris (2021) the complexity/diversity of conceptualizations to include “AI as a colleague,” blurring the
HR ecosystems and human/machine distinction.
technology Technology partners, suppliers, customers, and other actors in the
conceptualizations broader ecosystem need to be systematically accounted for.
slots and AI-enabled augmentation solution for crafting advertising when re-designing HR strategies, jobs, tasks, and workflows to inte-
campaigns—had elements of both. We present evidence that augmen- grate the work of humans and AI. Based on our study, we describe
tation cannot be neatly separated from automation in a workplace these two types of AI-solutions as equally important puzzle pieces in
ecosystem, at least not in a knowledge-intensive organization, such as a company-specific technology roadmap that can focus on either
our case company, that deployed multiple incremental AI-solutions automation or augmentation or both, depending on the types and tim-
supporting long-term business goals. That is, augmentation is both the ing of problems for which organizations need solutions, or the experi-
driver and the outcome of automation. Automation and augmentation ences of a person or organizational group concerned. Business
are inseparable parts of the same “picture” needed to move the com- leaders, in contrast to technology experts, do not first choose between
pany forward in its long-term AI implementation process. This automation and augmentation (Raisch & Krakowski, 2021) as a first-
included making corresponding changes in human jobs, skills, and hand solution to an overarching problem. Rather, they seek to address
tasks that enmeshed employee groups in different tasks and altered real-life organizational challenges with the help of emerging AI-
daily work routines. solutions (or other means) that are available to them.
Simultaneously, we call for further work and empirically grounded
scholarly debate on whether it is sensible to theorize about the rela-
tionship between automation and augmentation as “paradoxical ten- 5.1.2 | The puzzle-like co-existence of humans and
sion” that HR scholars or organizational members need to either solve artificial intelligence in the workplace ecosystem
or adjust to (Raisch & Krakowski, 2021; Schad et al., 2016; Smith &
Lewis, 2011). The concept implies an inherent dichotomy between The paradox perspective implies that automation and augmentation
the two solution types. This could be misleading and even counterpro- are interdependent, and that this interdependence extends to humans
ductive when imagining the future of work, and, more concretely, working with AI (Raisch & Krakowski, 2021; Schad et al., 2016;
Smith & Lewis, 2011). We contribute to this observation by empiri- Finally, our study problematizes discussions on the role of AI in re-
cally revealing that there are no clear crossovers between humans and humanizing work as the re-humanizing perspective (Armstrong, 1973)
AI and that separate organizational entities do co-exist in the work- partially prescribes, and shifting from repetitive and monotonous tasks
place ecosystems. The co-existence of humans and AI, including both to more creative and fulfilling ones (Daugherty & Wilson, 2018;
types of AI-solutions, is not monolithic (Kaplan & Haenlein, 2019). Zanzotto, 2019). In contrast, we observed a fundamental difference
Rather, it is a constantly changing puzzle, nested in the workplace between executives and managers eager to automate all possible
ecosystem composed of people designing AI-solutions, people work- tasks for employee creativity, and employees evidently were weary of
ing with these solutions, and people in charge of the business. Even constantly being creative and producing innovative ideas.
for a seemingly simple automation solution to work, long periods of The above-mentioned contributions have implications for HRM.
overlap are needed for members in a focal organization to collaborate We suggest that the co-existence of humans and AI is neither an
tightly, not only among themselves but also with its technology part- option nor a recommendation, but rather a necessity. AI implementa-
ners, suppliers, and even customers to improve the solution and oper- tion needs to be foremost a local, collective, iterative, and profoundly
ationalize advanced machine learning. This puzzle inevitably changes human phenomenon. It seems that the importance of human knowl-
over time, and, consequently, transforms the nature of the co- edge and expertise has been side-lined in the debate around the sec-
existence between humans and AI. ond machine age (Brynjolfsson & McAfee, 2014), the fourth industrial
revolution (Schwab, 2017), or the industry 4.0 (Schneider &
Sting, 2020). In our view, the primary focus in management studies
5.1.3 | No linear and straightforward AI-triggered and HRM needs to be on what companies do and why, how they use
job changes AI-solutions and with what timelines, who the key user groups of
these solutions are, and how members think and feel about AI rather
Our findings suggest that changes in job roles following the AI implemen- than AI-solutions themselves. We further recommend that the main-
tation are not necessarily linear or straightforward. Even though AI is stream technology conceptualizations in the HRM domain (see, for
occasionally portrayed as a new agent and rival to humans in the emerg- instance, Kim et al., 2021) need to be rethought to include “AI as a
ing surveillance capitalist system (Zuboff, 2019), it does not necessarily take colleague,” thus blurring the human/machine distinction. Expanding
over tasks previously done by humans, at least not in the foreseeable future. the workplace ecosystem to include not only human colleagues but
In our case company, we saw no evidence of humans fearing the loss of also those running on a machine logic represents a tremendous force
their jobs to AI, seen either as a “friend” or “too stupid” to be threatening. for change.
In contradistinction, many mundane-looking tasks involve creativity and
situation-sensitive, trust-based human interaction, and tacit knowledge
embedded in employee routines—as, in our case, in the work of traditional 5.2 | Practical implications
media employees. Establishing “clear rules” translatable into algorithms
may not be that simple (Raisch & Krakowski, 2021). Our study offers practical advice for business leaders and HR pro-
Our findings in part go against the deskilling theory fessionals interested in implementing AI and other digital technolo-
(Braverman, 1974; Davenport & Kirby, 2016) and the theory of tech- gies. First, by highlighting the fundamentally interdependent
nology dominance (Arnold & Sutton, 1998). Deskilling can happen in relation between automation and augmentation, we emphasize that
unexpected ways and not necessarily as a direct consequence of business leaders should give a balanced consideration to their entire
humans losing touch with work that AI suddenly takes over. In our technology roadmap and all their AI projects. A simple automation
case company, there were concerning signs of job roles narrowing and solution may be more iterative of integrated automation and aug-
verticalizing with more focus on the AI itself and less on the business mentation and employee time than initially thought. Both automa-
and its human side, as people learned skills such as coding and famil- tion and augmentation should be in focus and have the same status
iarity with Google proprietary tools. This was evident in the dichot- on management and employee agendas. If not, there is a danger of
omy between employee accounts in the digital and traditional media over-glorifying technologically advanced front-office augmentation
groups. Digital media employees focused on sharpening their techni- solutions without seeing the practicalities of what is involved in run-
cal skills, thus engaging in AI-induced upskilling. They simultaneously ning long, but equally important automation-related back-office
lacked knowledge of, and often even interest in, the broader media projects.
landscape and business issues: skills that management identified as a Second, business leaders and HR professionals need to acknowl-
“must” to win customer contracts. This contrasted with traditional edge that AI implementation may take much longer than initially
media employees, who, while admitting to a lack of advanced digital thought. Humans and AI need to change slowly, in tandem. Neither
skills at times, had a good oversight of broader business and practical automation nor augmentation can be delegated to the IT experts, nor
issues, mastered many types of traditional media channels, and had can it roll out without all involved members being on board. As a
close ties with people working in partner and customer companies. direct consequence, both management and HR departments need to
These qualities were needed to successfully do their jobs, and were thoroughly plan AI implementation, allocating sufficient resources,
extremely difficult to capture in AI-solutions, as it turned out. including specifically employee time and internal communication.
Crucially, human-to-human interaction around AI needs to be both a companies that sell the products. They design, execute, and measure the
managerial and employee priority. efficiency of media campaigns both in the traditional and digital media,
and work together with other companies, such as advertising agencies
Third, business leaders and HR professionals need to carefully
who provide the content for these campaigns.
map and plan for job role changes and competence shifts, asking 2
With traditional media we refer mainly to radio, TV, newspapers, and bill-
questions such as: What are our people really doing? What drives boards. This department was run as a separate unit, different from the
their motivation (or lack of it)? What skills are needed? How do we digital media—advertising on Google, Instagram, Facebook, TikTok, and
manage the competence shift? What tacit knowledge is involved? so on. AI is naturally integrated into the workflow of the employees in
the digital unit. Hence, advertising booking in this unit did not need to
Have we allocated enough employee time to make AI implementation
be separately automated and the tools used were mainly owned and
possible? Perhaps, the most important question of all is: How do we managed by the big social media companies making it imperative for the
ensure effective communication during AI implementation? employees to learn to use them.
RE FE RE NCE S
5.3 | Limitations and implications for future Alvesson, M., & Sandberg, J. (2021). Pre-understanding: An interpretation-
research enhancer and horizon-expander in research. Organization Studies,
43(3), 1–18.
Alvesson, M., & Sköldberg, K. (2018). Reflexive methodology: New vistas for
Our study has certain limitations, in part the result of our research
qualitative research (3rd ed.). SAGE.
design. These limitations open avenues for future research. Firstly, the Amershi, S., Cakmak, M., Knox, W. B., & Kulesza, T. (2014). Power to the
number of interviews comprising our data is limited since we could people: The role of humans in interactive machine learning. AI Maga-
not undertake a complete ethnography of the time-period forecast for zine, 35(4), 105–120.
Armstrong, R. L. (1973). The rehumanization of work. Social Theory and
the AI implementation. This is partly balanced by our in-depth involve-
Practice, 2(4), 459–473.
ment with the company over 14-months, and our reliance on both Arnold, V., & Sutton, S. G. (1998). The theory of technology dominance:
interviews and field observations and iterative contact with Understanding the impact of intelligent decision aids on decision
informants. makers' judgments. Advances in Accounting Behavioral Research, 1,
We acknowledge that our findings need refinement and expan- 175–194.
Bansal, P., & Corley, K. (2011). The coming of age for qualitative research.
sion through the inclusion of representatives of the broader ecosys-
Embracing the diversity of qualitative methods. Academy of Manage-
tems (for instance, technology and business partners, company global ment Journal, 54, 233–237.
headquarters, customers, government agencies, and competitors) Bar-Tal, D., & Bennink, G. (2004). The nature of reconciliation as an out-
since workforce complexity extends beyond the company, and organi- come and as a process. In Y. Bar-Siman-Tov (Ed.), From conflict resolu-
tion to reconciliation (pp. 11–38). Oxford University Press.
zations rely both on their own organizational workforce and those in
Beer, D. (2017). The social power of algorithms. Information, Communica-
the wider ecosystem (Snell & Morris, 2021). tion & Society, 20, 1–13.
Finally, our study indicates that AI is not one uniform phenome- Braverman, H. (1974). Labor and monopoly capital: The degradation of work
non but many and all at once. Hence, there is a need for more micro- in the twentieth century. Monthly Review Press.
Brynjolfsson, E., & McAfee, A. (2014). The second machine age: Work, pro-
level research using multiple methods and longitudinal designs into
gress, and prosperity in a time of brilliant technologies. W.W. Norton.
what is actually happening in workplace ecosystems from a human Brynjolfsson, E., & McAfee, A. (2017). The business of artificial intelligence.
perspective when organizations introduce digitalized resources and Harvard Business Review, 7(1), 1–2.
other emerging technologies constituting the second machine age. Daugherty, P., & Wilson, H. J. (2018). Human + machine: Reimagining work
in the age of AI. Harvard Business Review Press.
Davenport, T. H., & Kirby, J. (2016). Only humans need apply: Winners and
ACKNOWLEDGEMEN TS losers in the age of smart machines. HarperCollins.
We would like to thank the Associate Editor Corine Boon, three anon- Deng, Y., Bao, F., Kong, Y., Ren, Z., & Dai, Q. (2017). Deep direct rein-
ymous reviewers, as well as Mats Alvesson and Frank den Hond for forcement learning for financial signal representation and trading.
IEEE Transactions on Neural Networks and Learning Systems, 28,
their helpful comments of earlier versions of this article.
653–664.
Einola, K., & Alvesson, M. (2019). The making and unmaking of teams.
DATA AVAI LAB ILITY S TATEMENT Human Relations, 72(12), 1891–1919.
Data available on request from the authors. Einola, K., & Alvesson, M. (2021). When ‘good’ leadership backfires:
Dynamics of the leader/follower relation. Organization Studies, 42(6),
845–865.
ORCID
Faraj, S., Pachidi, S., & Sayegh, K. (2018). Working and organizing in the age
Katja Einola https://orcid.org/0000-0002-1548-9510 of the learning algorithm. Information and Organization, 28(1), 62–70.
Violetta Khoreva https://orcid.org/0000-0002-1118-9983 Floridi, L., & Sanders, J. W. (2004). On the morality of artificial agents.
Minds and Machines, 14, 349–379.
Floridi, L., & Taddeo, M. (2016). What is data ethics. Philosophical Transac-
ENDNOTES
tions of the Royal Society A: Mathematical, Physical and Engineering Sci-
1
Media consultancy agencies are rather invisible to most people; they are ences, 374, 1–4.
a type of back-office of the marketing world. A media consultancy is an Forster, E. M. (1909). The machine stops. In D. Leavitt & M. Mitchell (Eds.),
intermediary between consumers of products and services and the Selected stories. Penguin.
Gjerde, S., & Alvesson, M. (2020). Sandwiched: Exploring role and identity Rai, A., Constantinides, P., & Sarker, S. (2019). Next-generation digital plat-
of middle managers in the genuine middle. Human Relations, 73(1), forms: Toward human-AI hybrids. MIS Quarterly, 43, 3–9.
124–151. Raisch, S., & Krakowski, S. (2021). Artificial intelligence and management:
Glikson, E., & Woolley, A. (2020). Human trust in artificial intelligence: The automation-augmentation paradox. Academy of Management
Review of empirical research. Academy of Management Annals, 4(2), Review, 46(1), 192–210.
627–660. Ricoeur, P. (1981). Hermeneutics and the human sciences: Essays on lan-
Haenlein, M., & Kaplan, A. (2019). A brief history of artificial intelligence: guage, action and interpretation. Cambridge University Press.
On the past, present, and future of artificial intelligence. California Rothstein, R. L. (Ed.). (1999). After the peace: Resistance and reconciliation.
Management Review, 61(4), 5–14. Lynne Rienner Publishers.
Harley, B., & Cornelissen, J. (2020). Rigor with or without templates? The Schad, J., Lewis, M. W., Raisch, S., & Smith, W. K. (2016). Paradox research
pursuit of methodological rigor in qualitative research. Organizational in management science: Looking back to move forward. Academy of
Research Methods, 25(2), 1–23. Management Annals, 10, 5–64.
Harzing, A. W., Köster, K., & Magner, U. (2011). Babel in business: The lan- Schneider, P., & Sting, F. J. (2020). Employees' perspectives on
guage barrier and its solutions in the HQ-subsidiary relationship. Jour- digitalization-induced change: Exploring frames of industry 4.0. Acad-
nal of World Business, 46(3), 279–287. emy of Management Discoveries, 6(3), 406–435.
Huang, M.-H., Rust, R., & Maksimovic, V. (2019). The feeling economy: Schwab, K. (2017). The fourth industrial revolution. World Economic Forum,
Managing in the next generation of artificial intelligence. California Geneva, Switzerland.
Management Review, 61(4), 43–65. Smith, W. K., & Lewis, M. W. (2011). Toward a theory of paradox: A
Kaplan, A., & Haenlein, M. (2019). Siri, Siri, in my hand: Who's the fairest dynamic equilibrium model of organizing. Academy of Management
in the land? On the interpretations, illustrations, and implications of Review, 36, 381–403.
artificial intelligence. Business Horizons, 62(1), 15–25. Snell, S., & Morris, S. (2021). Time for realignment: The HR ecosystem.
Kim, S., Wang, Y., & Boon, C. (2021). Sixty years of research on technology Academy of Management Perspectives, 35(2), 219–236.
and human resource management: Looking back and looking forward. Stake, R. E. (2005). Qualitative case studies. In N. K. Denzin & Y. S. Lincoln
Human Resource Management, 60, 229–247. (Eds.), The Sage handbook of qualitative research (pp. 443–466). Sage
Kriesberg, L. (1998). Coexistence and the reconciliation of communal con- Publications Ltd.
flicts. In E. Weiner (Ed.), The handbook of interethnic existence Stake, R. E. (2010). Qualitative research: Studying how things work. The Guil-
(pp. 182–198). Continuum. ford Press.
Kump, B. (2022). No need to hide: Acknowledging the researcher's intuition Telkamp, J. B., & Anderson, M. H. (2022). The implications of diverse
in empirical organizational research. Human Relations, 75(4), 635–654. human moral foundations for assessing the ethicality of artificial intelli-
Kunze, F., Boehm, S., & Bruch, H. (2013). Age, resistance to change, and gence. Journal of Business Ethics, 178, 961–976. https://doi.org/10.
job performance. Journal of Managerial Psychology, 28, 741–760. 1007/s10551-022-05057-6
Lindebaum, D., Vesa, M., & den Hond, F. (2020). Insights from the machine Weiner, E. (1998). Coexistence work: A new profession. In E. Weiner (Ed.),
stops to better understand rational assumptions in algorithmic The handbook of interethnic existence (pp. 13–24). Continuum.
decision-making and its implications for organizations. Academy of Welch, C., & Piekkari, R. (2006). Crossing language boundaries: Qualitative
Management Review, 45(1), 247–263. interviewing in international business. Management International
Locke, K., Feldman, M. S., & Golden-Biddle, K. (2015). Discovery, validation, Review, 46(4), 417–437.
and live coding. In K. Elsbach & R. Kramer (Eds.), Handbook of qualitative Whittaker, D. J. (1999). Conflict and reconciliation in the contemporary
organizational research: Innovative pathways and methods. Routledge. world. Routledge.
Locke, K., Golden-Biddle, K., & Feldman, M. S. (2008). Perspective— Zanzotto, F. M. (2019). Human-in-the-loop artificial intelligence. Journal of
Making doubt generative: Rethinking the role of doubt in the research Artificial Intelligence Research, 64, 243–252.
process. Organization Science, 19(6), 907–918. Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human
Maxwell, J. A., & Miller, B. A. (2008). Categorizing and connecting strate- future at the new frontier of power. Public Affairs.
gies in qualitative data analysis. In N. S. Hesse-Biber & P. Leavy (Eds.),
Handbook of emergent methods (pp. 461–477). New York: Guilford
Press.
Moser, C., den Hond, F., & Lindebaum, D. (2022). Morality in the age of How to cite this article: Einola, K., & Khoreva, V. (2023). Best
artificially intelligent algorithms. Academy of Management Learning & friend or broken tool? Exploring the co-existence of humans
Education, 21(1), 139–155. and artificial intelligence in the workplace ecosystem. Human
Nilsson, N. J. (1971). Problem-solving methods in artificial intelligence.
Resource Management, 62(1), 117–135. https://doi.org/10.
McGraw-Hill.
Potter, J., & Wetherell, M. (1987). Discourse and social psychology. Beyond 1002/hrm.22147
attitudes and behaviour. Sage.
APPENDIX A
A.1 | Interview schedule (chronological order)
Years of Interview
company length Interview
Job position Organizational group experience Gender Date (minutes) language
CEO (1st interview) Executive 5 M August 29, 2019 92 English
CTO (1st interview) Executive 3 M August 29, 2019 89 English
CTO (2nd interview) October 9, 2019 63 English
CEO (2nd interview) October 15, 2019 67 English
Managing Director 1 Manager 1 F October 29, 2019 36 English
HR Director (1st interview) Manager 3 M November 1, 2019 55 Finnish
Managing Director 2 Manager 4.5 M November 5, 2019 61 English
Head of TV and Radio Planning Manager 4 F November 15, 57 Finnish
2019
TV Planner 1 Employee (traditional 7 F November 15, 57 Finnish
media) 2019
Printing Planner Employee (traditional 9 F November 15, 100 Finnish
media) 2019
Strategist 1 Manager 4 M November 15, 56 Finnish
2019
Senior Manager Insight Manager Not M November 19, 59 Finnish
known 2019
Insight Manager 1 Manager 5 M November 19, 61 Finnish
2019
Investment Director 1 Manager 4 M November 19, 53 Finnish
2019
TV Planner 2 Employee (traditional 15 F December 4, 2019 54 Finnish
media)
Development Manager Manager 7 F December 5, 2019 60 Finnish
Operator 1 Employee (traditional 8 F December 5, 2019 56 Finnish
media)
Operator 2 Employee (traditional 8 F December 5, 2019 57 Finnish
media)
Strategist 2 Manager 3 M December 12, 53 Finnish
2019
Director (Digital) Manager 1.5 M December 12, 59 Finnish
2019
Online Planner 1 Employee (digital media) 2.5 F December 19, 57 Finnish
2019
Search Engine Advertising Employee (digital media) 3.5 M December 19, 50 Finnish
Specialist 2019
Strategist 3 Manager 4 M December 19, 52 Finnish
2019
TV Planner 3 Employee (traditional 10 F December 19, 34 Finnish
media) 2019
Investment Director 2 Manager 9 M December 19, 24 Finnish
2019
Digital Designer Employee (digital media) 3 F January 9, 2019 68 Finnish
CTO (3rd interview) January 14, 2019 35 English
Social Media Strategist 1 Employee (digital media) 2.5 M January 20, 2019 32 Finnish
Head of Digital Marketing Manager 3.5 M January 20, 2019 53 Finnish
(Continues)
Years of Interview
company length Interview
Job position Organizational group experience Gender Date (minutes) language
Strategist 4 Manager 4 F January 20, 2019 31 Finnish
Social Media Strategist 2 Employee (digital media) 4 M January 20, 2019 46 Finnish
TV Planner 4 Employee (traditional 8 F March 3, 2019 54 Finnish
media)
Online Planner 2 Employee (digital media) 4 M March 3, 2019 56 Finnish
CEO (3rd interview) March 6, 2019 86 Finnish
CFO Executive Not F March 6, 2019 90 Finnish
known
HR Director (2nd interview) March 6, 2019 95 Finnish
CTO (4th interview) March 10, 2019 24 English
APPENDIX B
B.1 | Interpretative method, examples of analysis
Inductively generated Application of hermeneutic

Dimension meanings principles Cross-check of statements Researcher reflexivity
Meanings given to AI People across groups sought The CEO and CTO were Other than multiple Our previous work
Executive view: to relate to AI in situated both technology interviews with CEO and experience in IT and in
AI is an inevitability ways that gave this enthusiasts. The initial CTO both together and conducting research in
technology meanings that discussions revolved separately, we acquainted other technological
were highly personal. entirely around the ourselves with the settings made us doubt
technical promises of AI. company technology the executive discourse
Further analysis, however, roadmap and the AI- that “technology project
indicated that, for solutions being is inevitable and there
executives, the change implemented. We is no choice.” Many
was not mainly about conducted an online “inevitable” IT projects
technology leadership. industry analysis and fail. Here, however, with
Innovations were badly researched the history of the changing media
needed to survive in a Wizz up to its acquisition landscape and massive
business in which by the multinational shift of power from
quarterly results mattered corporation in 2012, as traditional to social media
and margins were well as the career history and toward digitalization,
shrinking. of the CEO. we accept the
interpretation of AI as an
inevitability.
Metaphor Employees used many Traditional media employee The use of We were not expecting to
Traditional media metaphors in describing pronounced anthropomorphizing find metaphors when we
employee view: AI and these reflected the anthropomorphizing of AI vocabulary to describe AI started our empirical work
AI as a broken tool, differences among groups. in negative terms was a was present in almost all – but when the
stupid co-worker People in the traditional counter-reaction to the interviews with traditional employees began to talk
media group used affective vocabulary of media employees and about AI in human terms
adjectives implying the executive (who used executive management. (calling it “stupid” and so
dysfunction when they strongly positive words The former used negative on), we started explicitly
spoke about AI, even such as “junior colleague” words, the latter positive. including questions in our
though the degree of and so on). subsequent interviews
hostility varied from and discussions about
tolerant to openly what AI represented to
aggressive. people.
Type of change AI Unlike executives, managers “Managers” was the busiest Some business managers We concluded the research
triggered with business roles did organizational group. We were concerned (“being project accepting that our
Managerial view: not demonstrate any stretched our availability up at night worrying research material from the
passionate feelings about to ensure we achieved about losing key players”) managerial group was the
Inductively generated Application of hermeneutic

Dimension meanings principles Cross-check of statements Researcher reflexivity
Exciting AI per se. Their interest sufficient interview time about employees, others shallowest. This was,
opportunities was business and growing for in-depth interviews were more focused on the however,
the customer base, with with as the least new deals that AI was counterbalanced with
the help of AI. disruptions as possible. expected to facilitate (“I exceptionally rich
This was important for a am expecting significant interview material from
balance between different growth of my business executive managers and
groups. thanks to AI”). In both background information
cases, and even though from the HR Director.
the types of statement
were very different, the
key concern remained
winning and executing
deals.
Characterization of Despite the work of digital We found that, for digital We asked various After becoming acquainted
co-existence with and traditional media media employees, the employees to explain with the work of digital
AI employees being very processes were highly stepwise the process of media employees, we
Digital media similar (design and automated and seamless, “booking advertising realized that AI was
employee view: execution of advertising and “augmentation” took space” on, for instance, a unproblematic because it
Co-existence is campaigns), the employee place on well-functioning local TV channel and at was an inevitable
natural experiences were very automated platforms. Google to find out how presence in their toolbox
different. Traditional media “AI” was involved. and the way digital media
employees needed to functioned from the
spend more time sorting beginning.
out deficient automation
that poorly captured their
workflow.

Human Resource Management - 2022 - Einola - Best friend or broken tool Exploring the co‐existence of humans and artificial

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Human Resource Management - 2022 - Einola - Best friend or broken tool Exploring the co‐existence of humans and artificial

Uploaded by

Copyright:

Available Formats

DOI: 10.1002/hrm.

SPECIAL ISSUE ARTICLE

Best friend or broken tool? Exploring the co-existence of

Katja Einola 1,2 | Violetta Khoreva 3

Hum Resour Manage. 2023;62:117–135. wileyonlinelibrary.com/journal/hrm 117

FIGURE 1 Data collection and analysis

TABLE 1 Group member sensemaking of AI

Executives Managers Digital media employees Traditional media employees

We see here how implementation of the automation solution, ostensi-

TABLE 3 Theoretical contributions and implications

A.1 | Interview schedule (chronological order)

B.1 | Interpretative method, examples of analysis

Inductively generated Application of hermeneutic

Inductively generated Application of hermeneutic

You might also like