Scholarly Article

Journal of Service Research

2024, Vol. 0(0) 1–20
The Future of Work: Understanding the © The Author(s) 2024

Effectiveness of Collaboration Between Human Article reuse guidelines:

sagepub.com/journals-permissions
and Digital Employees in Service DOI: 10.1177/10946705241229419
journals.sagepub.com/home/jsr

Khanh B. Q. Le1 , Laszlo Sajtos2, Werner H. Kunz3 , and Karen V. Fernandez2

Abstract
The use of digital employees (DEs)—chatbots powered by artificial intelligence (AI)—is becoming increasingly common in the
service industry. However, it is unclear whether collaborations between the human employee (HE) and DE can influence customer
outcomes, and what the mechanisms behind such outcomes are. This research proposes and tests a theoretical model that explains
how the communication of HE-DE collaboration in the form of interdependent behavioral cues can influence customer evaluations
of the service they received from such a team. Five experimental studies involving a total of 1403 participants demonstrate that
making HE-DE collaboration visible to customers during the service encounter can reinforce their perception of HE-DE team
cohesiveness and service process fluency, driving satisfaction. The communication of coordination and team goal cues are two
strong stimulants that strengthen such impressions. Further, this research also reveals that the HE-DE collaboration (vs. aug-
mentation or substitution) appeals to customers thanks to their perception of a transparent process, which is induced through
collaborative cues. This research provides theoretical implications for a transparent collaborative process between HE and DE and
practical advice for firms seeking to integrate DE into their organizations’ workflows.

Keywords
human employee, digital employee, collaboration, interdependence, co-presence, coordination cue, supervisory cue, team goal
cue, process fluency, team cohesion, customer satisfaction

Introduction Furthermore, DEs are also being investigated for their po-
tential to share goals with teachers in managing students in
Many companies deploy digital employees (DEs) to coor- education settings (Hew et al., 2022). Microsoft has recently
dinate with human employees (HEs) in service provision. unveiled its new DE, based on a large language model—
Digital employees are robotic software powered by machine Microsoft 365 Co-pilot, which is capable of working si-
learning algorithms and artificial intelligence (AI). They are multaneously with HEs on complex tasks such as analyzing
distinguishable from conventional chatbots such as FAQ spreadsheets, in real time (Spataro, 2023). These examples
bots, which are capable only of simulating conversations so demonstrate companies’ ubiquitous efforts to deploy HE and
that users feel as though they are talking to a human—they DE as the collaborative service frontline, revolutionizing the
have a very limited capability for processing rule-based tasks. future of customer service.
In contrast, DEs (e.g., Chatsonic; Microsoft 365 Co-pilot) are Prior research into DEs has already underlined the im-
characterized by a high-level agency, which enables them to portance of humans collaborating with non-human agents
manage service processes autonomously, perform a wide
range of complex tasks (e.g., scheduling meetings, pro-
cessing incoming data, and coordinating with humans), and 1
Department of Marketing, Management School, Lancaster University,
to make decisions independently. For instance, Stitch Fix—
Lancaster, UK
an online personal styling service—uses their digital agent to 2
Department of Marketing, Business School, University of Auckland, Auckland,
collect information about clothing preferences and give New Zealand
3
initial suggestions, then it transfers those suggestions to a College of Management, University of Massachusetts, Boston, MA, USA
stylist to pick the most suitable ideas before presenting them
Corresponding Author:
to the customer (Lake, 2018). Likewise, the Bank of America Khanh B. Q. Le, Department of Marketing, Management School, Lancaster
uses their virtual assistant, Erica—supervised by HEs—to University, Bailrigg Campus, Lancaster LA1 4YW, UK.
help customers manage their money (Fuscaldo, 2019). Email: k.le1@lancaster.ac.uk
2 Journal of Service Research 0(0)

(Huang and Rust, 2022; Le, Sajtos, and Fernandez, 2023; Consequently, practitioners can gain a deeper understanding
Noble et al., 2022; Wirtz et al., 2018). However, such of the benefits of integrating these collaborative cues into a
research has not yet examined this collaboration empirically new HE-DE joint service delivery mode.
by outlining its components and processes from the cus-
tomer’s viewpoint. In contrast to the dominant perception of
replacement or augmentation, which can only take place at
HE-DE Collaboration Cues and Mechanisms
either task or role level (De Keyser et al., 2019), the col- Research in marketing examining the role of robotic technol-
laboration view emphasizes how actors undertake their tasks ogies has underlined the debate over whether these technologies
at the process level (Le, Sajtos, and Fernandez, 2023). This will replace (De Keyser et al., 2019; McLeay et al., 2021) or
research has two objectives. The first objective is to con- augment humans (Grewal Dhruv et al., 2020; Noble et al.,
ceptualize and study the impacts of HE-DE collaborative 2022). This research stream considers robotic technologies as
cues on customers’ perception of a fluent service process and of tools that either take over or enhance the completion of indi-
HE-DE team cohesion, and ultimately of customer satisfaction vidual tasks (Huang and Rust, 2018). However, this task-
with such a service team. The second objective is to explore how a focused approach, utilizing the machine-as-a-tool principle,
transparent process, induced through HE-DE collaborative cues, overlooks the interaction, collaboration, and intricate interde-
could play a role in influencing the likelihood of adoption (i.e., use pendencies between the HE and DE within a workflow and
and recommendation) rather than the augmentation and substi- customer service process (Le, Sajtos, and Fernandez, 2023). On
tution types of service delivery mode. This research draws on the the other hand, previous research has also suggested that these
behavioral aspect of interdependence (Rusbult and Van Lange, non-human agents can be perceived as partners, especially when
2008) and proposes to examine how collaboration cues are people attribute human-like traits to them (Seeber et al., 2020;
communicated. Specifically, it focuses on conceptualizing the task Van Doorn et al., 2017). This research aims to explore the
and entity connection signals when the HE and DE work together collaboration between HEs and DEs from this partnership
in service teams1 to handle customer requests. An overview of key perspective, focusing on how they share the work process.
concepts, hypotheses, illustrative examples, and the results is Consequently, it underlines the need to develop a new theo-
presented in Figure 1. retical approach to conceptualize the foundational elements and
The current research makes three contributions to the mechanisms of HE-DE collaboration, and more importantly
research stream in human-robotic agent collaboration in service how such collaboration is perceived by customers during a
(De Keyser and Kunz, 2022; Huang and Rust, 2022; Noble service episode. Further, it is necessary to note that collabo-
et al., 2022; Van Doorn et al., 2017, 2023; Wirtz et al., 2018). ration and augmentation are related yet distinct concepts (Le,
First, this research complements empirical studies on DEs Sajtos, and Fernandez, 2023). This research proposes that these
(Garvey, Kim, and Duhachek, 2023; Huang and Dootson, 2022; two concepts differ in two crucial ways.
Jeon, 2022; Longoni, Bonezzi, and Morewedge, 2019; Longoni First, the underlying agency of a robotic agent in collabo-
and Cian, 2022; Yalcin et al., 2022) by investigating how the ration and augmentation is distinguishable. In the case of
collaboration between HEs and DEs can be depicted through collaboration, robotic agents have a significant level of agency
interdependent cues. It aims to demonstrate the benefits of to act independently (Seeber et al., 2020). For example, gen-
displaying such cues in the service process, thereby reinforcing erative AI-based applications such as ChatGPT, Microsoft Co-
positive service evaluations. Pilot, DALL-E, or Bing Chat can create answers to users’
Second, prior research has not yet examined empirically the questions on their own terms, and users have no control over
mechanisms that enable effortless customer experience, par- how the final answer looks. Second, collaboration is imbued in
ticularly in services provided jointly by human-robotic agent an actual working process that is shared between an HE and a
teams (Le, Sajtos, and Fernandez, 2023; Wirtz et al., 2018). This DE with a clear working structure (e.g., division of labor and
research identifies and tests mechanisms that explain how coordination) (Le, Sajtos, and Fernandez, 2023) whereas
communicating HE-DE collaboration cues impacts customer augmentation emphasizes task-based enhancement of human
outcomes, through perceived HE-DE team cohesion and service ability, requiring that the human and the augmenting technology
process fluency. This research finds that not only is the are fused together (e.g., exoskeleton and bionic eye) (Garry &
smoothness of the service process important to customer ex- Harwood, 2019; Grewal Dhruv et al., 2020). Making such a
perience but also the manner in which the HE-DE service team distinction between the two related concepts is necessary be-
manages this process in the customer’s presence. cause these theoretical perspectives serve different purposes.
Lastly, this research extends previous studies that have Collaboration requires a new conceptual approach that high-
examined the conditions under which customers prefer ser- lights its process-based nature and the machine-as-a-partner
vices provided by a team of humans and robotic agents principle (Le, Sajtos, and Fernandez, 2023). From the customer
(Longoni, Bonezzi, and Morewedge, 2019; Longoni and viewpoint, in augmentation, an actual working process between
Cian, 2022; Yalcin et al., 2022). It demonstrates that cus- the two agents might or might not be observable depending
tomers’ perceptions of a transparent service process, man- upon how such technologies are utilized by HEs, whereas in
aged collaboratively by an HE-DE team, can significantly collaboration, the process is clearly observable to customers
influence their response to the service provided. through HE-DE collaborative cues.
Le et al. 3

Figure 1. An overview of core concepts, hypotheses, and testing results.

Interdependence Theory interdependence (information exchange required to complete a

task). These are the aspects of structural interdependence
In developing our HE-DE collaboration framework, we draw on (Courtright et al., 2015), which help explain the emergence of
the interdependence theory, which originated in management, helping behaviors that influence team cooperativeness
psychology, and research on team management (Courtright (Courtright et al., 2015; Van Der Borgh, De Jong, and Nijssen,
et al., 2015; Rusbult and Van Lange, 2008; Wageman, 1995, 2019; Wageman, 1999); and these behaviors influence how
1999). Interdependence theory posits that an individual’s be- members act as a team. In behavioral interdependence, members
havior is determined by their own and their collaborator’s ac- engage in two discrete activities—taskwork and teamwork
tions (Rusbult and Van Lange, 2008; Wageman, 1999). The (Marks, Mathieu, and Zaccaro, 2001). This research deliber-
theory delineates the way that interdependence in collaboration ately draws on behavioral interdependence and posits that such
consists of goal interdependence (presence of a joint goal) behaviors can be served as behavioral cues that can be utilized in
(Courtright et al., 2015; Wageman, 1995) and task signaling the HE-DE collaboration to customers.
4 Journal of Service Research 0(0)

Further, the mechanisms explaining the impacts of be- This research proposes the concept of HE-DE frontline co-
haviorally interdependent cues can also be drawn from the presence, defining it as the visible sequencing of the co-presence
theory (Courtright et al., 2015). As individuals engage in of HE and DE in a customer-facing process. This component
collaboration, they commit their shared efforts toward the represents the workflow of the HE-DE team, which can be either
work process to achieve their joint objective (Courtright simultaneous (both agents serving the customer concurrently) or
et al., 2015; Johnson and Johnson, 2009; Rusbult and Van a sequential (one agent serving the customer at a time) (e.g.,
Lange, 2008; Wageman, 1995, 1999). Prior research suggests Shrestha, Ben-Menahem, and Von Krogh, 2019). For example,
that displaying such interdependent behavioral cues can Botnation.ai (2022) designs their DEs to assign conversations to
foster the customer’s recognition of (a) a positive alliance and an HE, enabling the HE to manage the conversation
(b) effortless HE-DE teamwork through their awareness of simultaneously.
the team’s joint efforts in handling service requests (Le, Further, this research defines a coordination cue as the visible
Sajtos, and Fernandez, 2023). This research refers to these communication regarding the orchestration of work handover
mechanisms as team cohesion and process fluency, respec- between DE and HE. A coordination cue specifically highlights
tively. Perceived HE-DE team cohesion is referred to as the to customers the act of being transferred between agents. For
customer perception of the HE-DE team’s “…bonding in example, Kasisto, a DE solution provider, enables its DEs to
which members share a strong commitment to one another notify customers when they are being transferred to an HE, who
and/or to the purpose of the team” (Mathieu et al., 2015, in turn acknowledges the information received from the DE. In a
p. 714). This is an entity connection mechanism, as it HE-HE collaboration context, Wang, Hoegg, and Dahl (2018,
highlights the emergent state of members’ relationship dynamics p. 195) conceptualized coordination as a subtle “referring ac-
(Marks, Mathieu, and Zaccaro, 2001). Process fluency is referred tion” to a partner, typically used when the first employee cannot
to as customer perception of “the subjective experience of the answer customer requests. In the HE-DE context, this is akin to
ease and speed with which an incoming stimulus is processed” human support stepping in where a robot cannot complete a task
(Orth and Wirtz, 2014, p. 296). This is a task connection (Choi, Mattila, and Bolton, 2021). Previous studies have as-
mechanism as it emphasizes the perception of service continuity. sociated coordination with helping behavior and considered it
These two mechanisms represent how the team and the process an optional action, not a formal part of the workflow. In contrast,
functioned and render the customer’s evaluation of the effec- this research emphasizes the proactive communication of co-
tiveness and efficiency of the team’s service quality (Le, Sajtos, ordination. It posits that service agents (HE-DE) should notify
and Fernandez, 2023), and they also act as facilitators of customer customers about their transfer as a formal requirement of the
outcomes (Cassab and MacLachlan, 2006; Marinova, Ye, and workflow, which makes the interconnectedness of the allocated
Singh, 2008; Shen et al., 2018). tasks more noticeable.
In sum, this research first utilizes the interdependence theory
and draws on the behavioral aspect (taskwork and teamwork
Entity Connection Cues
activities) to conceptualize the collaboration cues (task and
entity connection cues). Second, it proposes that these col- Teamwork activities consist of behavioral actions that em-
laboration cues will impact team mechanisms, which consist of phasize members’ joint efforts to maintain their relationship
task connection (process fluency) and entity connection (team (Mathieu et al., 2020). These activities include supervising their
cohesion) mechanisms. These, in turn, will drive customer partners’ work progress and setting shared goals (Mathieu et al.,
outcomes, such as satisfaction, in a service process actively 2020). From the customer’s perspective, these activities serve as
managed by a HE-DE team. cues for recognizing the nature of the HE-DE work
relationship. We refer to this second activity as the entity
connection cue. Within this context, supervisory and team goal
Task Connection Cues
cues inform customers about the hierarchical relationship
Taskwork activities consist of behavioral actions that emphasize (supervisor-subordinate) between HE and DE, as well as the
members’ joint efforts to complete tasks (Wageman 1995, HE’s motivation to work effectively with the DE.
1999). These activities include coordination efforts and the We define the supervisory cue as the visible communi-
scheduling of the work sequence (Marks, Mathieu, and Zaccaro cation that informs customers about which party, either the
2001). In our research, we examine how, from the customers’ DE or HE, is in charge of overseeing their partner’s work.
perspective, these activities provide cues that enable them to The supervisory cue represents the supervisor-subordinate
understand the progression through the tasks scheduled by the structure to customers (Shanks et al., 2021). For example,
HE-DE team, leading to the desired outcome. We identify this Hana, a DE on One New Zealand’s website, informs cus-
first activity as the task connection cue, as it signals to customers tomers at the start of a chat session that they can escalate
the orchestration of the HE-DE team’s taskwork. Within this their request to an HE at any time. This arrangement gives
cue, this study advances two components: HE-DE frontline co- the impression that the HE is monitoring the interactions
presence and coordination cues, which reflect the intercon- between the DE and customers. Recent research suggests
nectedness of the service workflow managed by the HE-DE that when the DE’s name is accompanied by the job title
service team. “manager,” participants perceive the DE as having a
Le et al. 5

position of authority (Jeon, 2022). Similarly, Shanks et al.’s Only a few empirical studies have investigated when and
(2021) experimental study found that participants attributed how the interplay between employees and robotic agents can
more power to a robot leading a nutrition counseling ses- improve customer experience. Choi, Mattila, and Bolton (2021)
sion. These findings indicate that a DE can also be perceived and Longoni, Bonezzi, and Morewedge (2019) found that a
as the supervisor of an HE. This supervisory arrangement sequential workflow where HE can intervene or make the final
can be recognized by customers through the supervisory decision appeal to customers. Several studies focused on the
cues displayed to them. supervisory cue and found that the HE as a leader generates a
Further, drawing on goal interdependence research more positive attitude in service provision (Huang and Dootson,
(Courtright et al., 2015; Wageman, 1995, 1999), we define team 2022; Longoni and Cian, 2022; Yalcin et al., 2022) than dis-
goal cue as the visible communication which informs the playing a robotic agent as the team leader (Jeon, 2022; Shanks
customer that HE and DE share a customer-related perfor- et al., 2021). Further, these empirical studies examined only the
mance objective. This cue signals to customers the shared aspects of such collaboration at the role level (see Table 1).
commitment of the HE-DE team in delivering services (Le, These studies also emphasize only the augmenting nature of
Sajtos, and Fernandez, 2023). For example, Virtual Vet Nurse, a robotic agents rather than actual teamwork between the two in a
DE solution provider, designs its agents to communicate the service process. This leads to an observation that prior research
team’s joint intention to provide the best possible veterinary has not examined other aspects, such as coordination or team
support. It is important to note while the DE is programmed to goal and how communicating such cues could enhance cus-
fulfill specific organizational objectives, the key factor is the tomer experience.
HE’s willingness to collaboration with the DE and in serving The current research extends this stream in two ways. First,
customers. When this shared motivation is communicated to this research will examine the collaboration as a process with an
customers, they can sense the united effort between the agents, emphasis on communicating HE-DE teamwork to customers
based on the HE’s willingness to work with the DE as a team to through interdependent behavioral cues to explain how cus-
achieve their shared goal. Figure 2 illustrates the conceptual tomers evaluate such teamwork in an actual work process that is
framework. managed by both agents. Second, this research proposes process
fluency and team cohesion as the mechanisms that explain how
the communication of such cues can transform the customer
An Overview of Empirical Research on Human-Robot experience.
Collaboration in Service
The proposed conceptualization raised a need to understand
how these components and mechanisms have been examined in Hypothesis Development
the literature. Table 1 below outlines recent empirical research
on working with robotic agents. These studies were selected
The Impact of Task Connection Cues
because they examined different aspects that related to the Figure 3 depicts our analytical framework. Within the context of
domain of collaborating with robotic agents. an HE-HE team, a process with a highly interdependent

Figure 2. Conceptual framework.

6 Journal of Service Research 0(0)

Table 1. A Review of Relevant Empirical Research on Human-Robotic Agent Interplay in the Services Marketing Discipline.

Collaboration Collaboration
Cues as a Process

Author(s) FC CC SC TC Y N Underlining Concept Mechanisms Main Findings

This research 3 3 3 3 3 Communication of Process fluency FC has no impact. CC, SC (partially

interdependent cues Team cohesion supported), and TC positively influence
process fluency and team cohesion,
which in turn, increase satisfaction
Choi, Mattila, and 3 3 Robot’s social Human Human intervention improved satisfaction
Bolton (2021) perception in service intervention when a non-humanoid robot caused the
failure failure, and a humanoid robot can
effectively recover by itself
Huang and 3 3 Disclosure timing of Coping Late disclosure of human availability had a
Dootson (2022) human agent’s mechanism positive impact on emotion-focused
availability coping mechanism, which increases
aggression
Jeon (2022) 3 3 Communication of Likability Displaying a transfer from an employee to
supervisory role AI manager results in a lower response
to manager’s likeability than a transfer
from an AI representative to an AI
manager.
Longoni, Bonezzi, 3 3 AI role allocation N/A Resistance toward the use of AI medical
and Morewedge (Study 9) service is not developed when AI service
(2019) provider supports human doctor
Longoni and Cian 3 3 AI role in utilitarian vs. N/A Even when a hedonic goal is activated,
(2022) hedonic context pairing with a human agent as a
supporter, the AI recommender did as
well as the human recommender
Shanks et al. (2021) 3 3 Communication of N/A A robot leader in a “robotic” team results
supervisory role in significantly less favorable behavioral
intentions compared to a human leader
Yalcin et al. (2022) 3 3 Type of decision maker Favorability of With favorable outcomes, the attitude
decision toward the company is less favorable if it
outcome is evaluated by the supervised algorithm
than by the human alone

FC, frontline co-presence; CC, coordination cue; SC, supervisory cue; TC, team goal cue.

Figure 3. The impacts of communicating collaborative cues on customer-perceived satisfaction in HE-DE service team context.
Le et al. 7

workflow requires more time and effort (Van de Ven, Delbecq, impression of a well-organized and efficient team (Wang,
and Koenig, 1976). In a sequential process, the downstream Hoegg, and Dahl, 2018). The emergence of intelligent sys-
member must wait for the upstream member’s output (Cannon- tems like DEs can also exhibit the capability to coordinate with
Bowers and Bowers, 2011), whereas in a simultaneous se- humans (Noble et al., 2022). We propose that the presence of a
quence, team members can complete their tasks concurrently coordination cue in HE-DE collaboration will have a positive
(Van de Ven, Delbecq, and Koenig, 1976). Although in human- impact on perceived fluency (Le, Sajtos, and Fernandez, 2023).
robot task sequencing, the sequential (vs. simultaneous) pres- Unlike the coordination in an HE-HE team context, which
ence increases the required time for task processing (Zhao, features flexibility where human collaborators can adapt to the
Henrichs, and Mutlu, 2020), and such research did not predict different situations in teamwork (Van de Ven, Delbecq, and
how human-robotic agent collaboration can influence the Koenig, 1976; Wang, Hoegg, and Dahl, 2018), the coordination
evaluation of work process fluency, especially from a third- in an HE-DE team context is rigid and rule-based, since it is
person view. strictly scheduled and planned (Zhao, Henrichs, and Mutlu,
Customers generally do not like to wait to be served 2020). Hence, in the HE-DE context, the presence of the co-
(Durrande-Moreau and Usunier, 1999). Consequently, we ordination cue can also serve as the signal of a timely and well-
propose that the sequential (vs. simultaneous) presence will managed process. When customers recognize such a cue, it
have a greater negative influence on perceived process fluency could create a sense that the HE understands how to work with
because sequential sequence dictates that the downstream agent the DE to create an efficient team, which may lead to an im-
relies on the output of their upstream partner, which results in a pression that their requests are being handled in a streamlined
delay in information exchange between agents. In contrast, a manner.
simultaneous process could result in a greater sense of fluency Further, the presence of the coordination effort can also
as HE and DE can work independently and concurrently. foster the emergence of the cooperative attitude among team
An interdependent workflow should create a greater sense of members (Courtright et al., 2015; Johnson and Johnson, 2009;
team cohesiveness in human work teams (Beal et al., 2003; Wang, Hoegg, and Dahl, 2018), which is an important factor in
Johnson and Johnson, 2009). A sequential presence denotes a creating a cohesive team (Mathieu et al., 2015). From the
mutual resource exchange and reliance among team members customer point of view, the communication of the coordination
which promotes unity (Wageman, 1995), whereas a simulta- cue could have a positive impact on perceived team cohesion
neous co-presence might create a sense of independence rather (Le, Sajtos, and Fernandez, 2023). Previous research has found
than interdependence (Cannon-Bowers and Bowers, 2011). that a coordination cue between employees when being prompted
Unlike in a HE-HE team setting, such an effect is unique to an (i.e., reactive coordination) enhances customer perception of such
HE-DE context due to the involvement of the non-human agent a team’s cohesion (Wang, Hoegg, and Dahl, 2018). In the context
that acts as a partner. Despite this, prior research has not ex- of HE-DE collaboration, the presence of the coordination cue
amined how an interconnected workflow can affect the eval- could create an impression that the HE is required to work with
uation of the partnership between the human and robotic agent. the DE rather than by themselves and has been trained to work
This research proposes that a sequential (vs. simultaneous) co- with the DE without any problem. Thus, the cue signals to
presence could foster the development of the customer’s im- customers an impression of the HE’s professionalism and will-
pression of a cohesive HE-DE team. Although simultaneous ingness to work cooperatively with the DE.
sequence emphasizes that both the DE and HE will appear
concurrently in front of the customers, they will likely be working Hypothesis 2a. The presence of a coordination cue in HE-DE
on different aspects of a process to optimize division of labor (Le, collaboration will have a positive impact on perceived
Sajtos, and Fernandez, 2023), which gives an impression of process fluency.
independent rather than interdependent. In contrast, a sequential Hypothesis 2b. The presence of a coordination cue in HE-DE
presence could give an impression of a cohesive HE-DE team due collaboration will have a positive impact on perceived HE-
to the required interaction nature of the sequential workflow DE team cohesion.
(Shrestha, Ben-Menahem, and Von Krogh, 2019).

Hypothesis 1a. A sequential HE-DE frontline co-presence

The Impact of Entity Connection Cues
(vs. simultaneous presence) will have a negative impact on In contrast to the power dynamic in an HE-HE team context,
process fluency. research on human-robotic agent collaboration suggests that
Hypothesis 1b. A sequential HE-DE frontline co-presence giving more power to the human agent in order for them to
(vs. simultaneous presence) will have a positive impact on intervene in the robotic agent’s operating cycle to correct its
perceived HE-DE team cohesion. actions, when necessary, could ensure a smooth operation
Further, in prior literature of collaboration between humans, (Johnson, Bradshaw, and Feltovich, 2018). Prior research has
coordination effort is necessary to maintain a smooth workflow documented that customers prefer the HE to be in the authority
in an HE-HE team (Johnson and Johnson, 2009; Marks, position (Longoni, Bonezzi, and Morewedge, 2019; Yalcin
Mathieu, and Zaccaro, 2001). Communicating such an effort et al., 2022) as the HE can intervene and redirect the DE if
through a coordination cue to customers could create an the agent is not performing as expected (Le, Sajtos, and
8 Journal of Service Research 0(0)

Fernandez, 2023). Human intervention can also enhance sat- goal between the DE and HE facilitates the belief that such a
isfaction when a robot fails to perform its task properly (Choi, team has a high sense of collective efficacy. A shared goal
Mattila, and Bolton, 2021), and customers generally prefer a motivates members to work smoothly and cooperatively
human adviser when the task is highly consequential (Castelo, (Johnson and Johnson, 2009). In human-robot interaction, the
Bos, and Lehmann, 2019). By having the HE as a supervisor, presence of a shared goal helps humans to orchestrate a smooth
customers can be reassured that an actual human being is coordination with their robotic partner (Johnson, Bradshaw, and
monitoring their interaction with the DE, ensuring the conti- Feltovich, 2018). When customers see that the two agents share
nuity of the engagement. In contrast, customers report less a common goal, they may take it as a sign of an efficient team as
positivity when the DE acts as the supervisor of the HE (Jeon, it indicates that the HE is committed in their effort to work with
2022; Shanks et al., 2021) as they may feel uncomfortable the DE to ensure a smooth encounter (Le, Sajtos, and
seeing the DE acting like a boss and find such a process dis- Fernandez, 2023). Further, displaying such a cue may also
quieting. Thus, signaling the HE (vs. DE) as the supervisor may reinforce the customer’s belief that the agents orchestrate their
enhance customer-perceived process fluency. effort effectively to serve them—the cue serves as a signal that
In the HE-HE team context, an imbalanced power distri- the team knows what to do and can avoid delays.
bution activates a high sense of responsibility for the power In addition, sharing a common goal can foster the emergence
holder (vs. the subordinate), which shapes the commitment of of cohesion perception through prosocial motives (Courtright
both entities toward a joint goal (Rusbult and Van Lange, 2008). et al., 2015; Johnson and Johnson, 2009). Hence, a team goal
In an HE-DE team context, such an imbalanced power distri- can be the driver of cohesion and the tendency to support one
bution not only shapes human responsibility in teaming with DE another (Courtright et al., 2015; Menguc, Auh, and Uslu, 2013).
(Le, Sajtos, and Fernandez, 2023), it also demonstrates the HE’s In contrast to the HE-HE context, goal sharing in the HE-DE
capability to manage intelligent systems in service provision team context also implies a sense of machine agency and au-
(Noble et al., 2022). From the customer’s perspective, by tonomy, which implies that the robotic agent has the intention to
communicating that the HE is in the position of authority, work cooperatively with the human partner (Le, Sajtos, and
customers may be reassured that the HE will be responsible for Fernandez, 2023). Communicating the cue that signals that the
the team’s actions. This human-based leadership eliminates the HE is sharing a service goal with DE may strengthen customer
perception of risk associated with incompetent teamwork since perception of a cohesive HE-DE team (Le, Sajtos, and
HE can intervene and help if the robot fails (see also Choi, Fernandez, 2023), because it implies mutual support—the
Mattila, and Bolton, 2021). Such behaviors can be considered as DE is an agent capable of cooperating with an HE (Walliser
a sign of a cooperative and cohesive team (Marks, Mathieu, and et al., 2019) while the HE is capable of managing teamwork
Zaccaro, 2001). In contrast, distributing more power to DE may with the machine agent to achieve their mutual service objective
not create an impression of a cohesive team because customers (Noble et al., 2022). As customers are the beneficiaries of the
generally perceive DEs as incompetent to manage human agents HE-DE teamwork, the commitment of HE to work with DE as a
(Jeon, 2022), and know that a DE is not capable of intervening team to serve them may be important for their experience.
to help the employee when they are stuck, especially where
tasks are intuitive and complex in nature. Hypothesis 4a. The presence of a team goal cue in HE-DE
collaboration will have a positive impact on process fluency.
Hypothesis 3a. The supervisory cue where the HE (vs. DE) Hypothesis 4b. The presence of a team goal cue in HE-DE
acts as a supervisor will have a positive impact on process collaboration will have a positive impact on perceived HE-
fluency. DE team cohesion.
Hypothesis 3b. The supervisory cue where the HE (vs. DE)
acts as a supervisor will have a positive impact on perceived
The Impacts of Process Fluency and Team Cohesion
HE-DE team cohesion.
In the context of our research, DEs by default carry the firm’s A seamless service experience will create a strong impression of
goal in their design for better serving customers. Current au- good service (Lemon and Verhoef, 2016). Customers are more
tomated systems might not be able to set goals that differ from easily satisfied when the service is undisrupted (Cassab and
the firm’s goal. However, this does not equate to a situation MacLachlan, 2006; Shen et al., 2018; Sun et al., 2020). The
where a DE does not have the capacity to share a performance perception of an effortless experience could make customers
objective with a HE. In fact, modern DEs are fully capable of appreciate the effort that the HE-DE team puts in. Hence,
sharing work goals with HEs (Kim et al., 2022; Noble et al., customers may be inclined to be more satisfied when they
2022; Walliser et al., 2019). For example, a teacher can set a perceive smooth teamwork. Further, research suggests that
desirable grade level with a DE as a joint objective in improving cooperativeness among service providers is an important factor
a student’s performance (Hew et al., 2022). Hence, our interest in enhancing service quality (Gracia, Cifre, and Grau, 2010). A
is customer perception when this team goal cue is activated, strong cohesive team affords a greater perception of coopera-
telling the customer that the HE is sharing a service goal with the tion, which directly and positively influences customers’ per-
DE, rather than investigating a situation where the DE could ception of service quality (Wang, Hoegg, and Dahl, 2018). In an
have different objectives from the HE. The presence of a shared HE-DE team, cohesiveness is a sign that HE understands how to
Le et al. 9

work effectively with DE. It reinforces the customer’s im- Study 1

pression that the HE-DE team is motivated to serve them as a
united team in resolving their requests. Design and Procedure
Participants recruited from MTurk (N = 117) were asked to read
Hypothesis 5. Customer perception of process fluency will about an education program’s advisory service. The participants
have a positive impact on their perceived satisfaction. were asked to play the role of a prospective student looking for a
Hypothesis 6. Customer perception of HE-DE team cohesion data science program to enhance their career prospects and to
will have a positive impact on their perceived satisfaction. imagine interacting with the advisory team, which consisted of a
human advisor (HE) and a chatbot advisor (DE). Frontline co-
presence was manipulated by altering whether HE and DE
An Overview of Empirical Studies appeared simultaneously or sequentially in serving the partic-
ipants. The coordination cue was manipulated by communi-
A detailed overview of the design, measurement, and justification cating the transfer of the task between them.
of the empirical studies can be found in the Web Appendix. The
structure of the empirical investigations is as follows. First, we
investigate the hypothesized effects in Figure 3. This is the main Manipulation Check
investigation which consists of four main studies. Study 1 focuses
The manipulation check for frontline co-presence was under-
on the impact of task-based connection cues (frontline co-presence
taken by asking participants to indicate a simultaneous or se-
and coordination). Study 2 adds an additional layer of entity-based
quential presence (1 = “I felt like I was being accompanied by
connection cues (supervisory). Studies 3A and 3B focus on both
EduBot and Sam at the same time” and 7 = “I felt like I was
task-based and entity-based connection cues (coordination, su-
being accompanied by EduBot and Sam one after the other”).
pervision, and team goal). Study 4 aims to replicate the impacts of
An independent-samples t-test confirmed the manipulation
coordination and team goal cues in a different setting (main
(MSimultaneous = 2.36, MSequential = 6.33), t (115) = 15.48, p <
analysis). Complementing this main analysis, Study 4 also ex-
.001. The coordination cue manipulation check was tested by
plores the potential influence of type of collaboration (HE-HE vs.
asking participants whether they observed EduBot and Sam
HE-DE). Further, beyond measuring the impacts on satisfaction,
working together (1 = strongly disagree; 7 = strongly agree).
Study 4 also explores the potential influences of these cues on
The result confirmed the manipulation (MAbsence = 2.60,
behavioral outcomes. In particular, this research measures two
MPresence = 5.96, t (113.43) = 13.91, p < .001).
additional behavioral outcomes—acceptance of the HE-DE team’s
recommendations and email subscriptions. These behavioral
outcomes were selected because they are among many behavioral Results
manifestations of engagement during the service provision (re-
sponse to the HE-DE team) and post-engagement (email sub- We used PROCESS macro (Hayes, 2017) to analyze the impact
scription) (e.g., Van Doorn et al., 2010). To streamline the of co-presence (0 = simultaneous; 1 = sequential) and coor-
investigation, the detailed results on these behavioral outcomes are dination cues (see part A in Table 2).
documented in Web Appendix F. Second, Study 5 aims to extend This study found a negative but insignificant effect of se-
the scope of this research and explores the additional impacts of quential presence on process fluency (b (SE) = 0.28 (0.18), p >
collaboration (as manifested through coordination and team goal .10). We did not find evidence to support hypothesis 1b, which
cues) compared to augmentation and substitution as a type of proposed that under sequential (vs. simultaneous) presence,
service delivery mode. The main investigation of this extension is team cohesion is higher (b (SE) = 0.04 (0.11), p > .10). Thus, we
the role of perceived process transparency—the extent to which a found no support evidence for both H1a and H1b. However, we
service provider makes the details of their service process visible to found a significant positive effect of coordination cue on
their customers (see also Le, Sajtos, and Fernandez, 2023). This is process fluency (b (SE) = 0.36 (0.09), p < .01) and on perceived
the concept of interest of the extended study because the con- HE-DE team cohesion (b (SE) = 0.82 (0.05), p < .01), which
ceptualization of collaborative cues in this research focuses on supports H2a and H2b. Both process fluency (b (SE) = 0.23
revealing to the customers how HE and DE work together. Thus, (0.08), p < .01) and team cohesion (b (SE) = 0.35 (0.14), p < .05)
this process transparency mechanism can complement the current positively influenced satisfaction, which supports H5 and H6.
investigations by extending the scope of the impact of revealing
those collaborative cues. Study 5 also diverges from prior studies
Discussion
and focuses on exploring the impacts of different service delivery
modes on usage and recommendation likelihoods, which helps to Study 1 aimed to establish the impact of task connection cues.
capture the overall adoption tendency. This could be a useful We find that the coordination cue improves both perceived
insight when HE-DE collaboration becomes popular in the future. fluency and team cohesion, whereas frontline co-presence as
Taken together, the series of empirical studies in this research aims sequential (vs. simultaneous) has no impact on perception of
to capture the complexity and nuances of communicating HE-DE process fluency or team cohesion. Previous research has found
collaborative cues. some beneficial effects of reactive coordination (i.e., one service
10 Journal of Service Research 0(0)

Table 2. Regression Analysis Results Across Studies 1–4.

Service Satisfaction (Input as

Process Fluency Team Cohesion Y)

Part A—Study 1 (N = 117, standardized score, model 4)

Constant .16 (0.13)ns 0.02 (0.08)ns 0.09 (0.12)ns
ns
Frontline co-presence (input as covariate; simultaneous 0.28 (0.18) .04 (0.11)ns (H1b) .15 (0.15)ns
presence = 0) (H1a)
Coordination cue (input as X) .36 (0.09)*** (H2a) .82 (0.05)*** (H2b) .13 (0.13)ns
Process fluency (input as M1) .23 (0.08)*** (H5)
Team cohesion (input as M2) .35 (0.14)** (H6)
Model summary
R2 .14 .68 .37
MSE .88 .32 .66
F 9.11*** 123.03*** 16.27***
Df 2, 114 2, 114 4, 112
Part B—Study 2 (N = 315, standardized score, model 4)
Constant .05 (0.10)ns 0.11 (0.09)ns 0.06 (0.08)ns
ns
Frontline co-presence (input as covariate; simultaneous 0.15 (0.11) 0.02 (0.09)ns .00 (0.08)ns
presence = 0) (H1a) (H1b)
Coordination cue (input as X) .30 (0.05)*** (H2a) .62 (0.04)*** (H2b) 0.04 (0.05)ns
Supervisory cue (input as covariate) .06 (0.11)ns (H3a) .20 (0.09)** (H3b) .11 (0.08)ns
Process fluency (input as M1) .27 (0.05)*** (H5)
Team cohesion (input as M2) .54 (0.06)*** (H6)
Model summary
R2 .10 .40 .51
MSE .91 .61 .50
F 11.24*** 67.98*** 64.91***
Df 3, 311 3, 311 5, 309
Part C—Study 3A (N = 294, standardized score, model 4)
Constant 0.03 (0.09)ns .07 (0.07)ns 0.05 (0.07)ns
Coordination cue (input as X) .18 (0.07)*** (H2a) .44 (0.05)*** (H2b) .04 (0.06)ns
Supervisory cue (input as covariate) .05 (0.12)ns (H3a) 0.12 (0.09)ns .09 (0.09)ns
(H3b)
Team goal cue (input as covariate) .12 (0.07)* (H4a) .30 (0.05)*** (H4b) .01 (0.06)ns
Process fluency (input as M1) .25 (0.05)*** (H5)
Team cohesion (input as M2) .46 (0.07)*** (H6)
Model summary
R2 .07 .43 .43
MSE .94 .58 .58
F 7.42*** 71.76*** 42.99***
Df 3, 290 3, 290 5, 288
Part D—Study 3B (N = 262, standardized score, model 4)
Constant 0.19 (0.11)* .02 (0.08)ns .06 (0.09)ns
Coordination cue (input as X) .16 (0.07)** (H2a) .53 (0.05)*** (H2b) 0.03 (0.07)ns
Supervisory cue (input as covariate) .26 (0.13)** (H3a) 0.03 (0.09)ns 0.08 (0.11)ns
(H3b)
Team goal cue (input as covariate) .21 (0.07)*** (H4a) .31 (0.05)*** (H4b) .04 (0.06)ns
Process fluency (input as M1) .34 (0.05)*** (H5)
Team cohesion (input as M2) .43 (0.08)*** (H6)
Model summary
R2 .14 .55 .46
MSE .87 .45 .56
F 14.05*** 105.55*** 42.76***
Df 3, 258 3, 258 5, 256
Le et al. 11

Base Model Full Model

Part E—Study 4 (N = 254, Standardized Process Team Service Service

Score, Model 4) Fluency Cohesion Satisfaction Process Fluency Team Cohesion Satisfaction

Constant 0.02 (0.08) 0.04 (0.06) 0.04 (0.07)ns 0.02 (0.08)ns 0.04 (0.06)ns 0.05 (0.06)ns
ns ns

Coordination cue (input as X) .28 (0.07)*** .65 (0.05)*** .01 (0.07)ns .25 (0.07)*** .64 (0.05)*** .01 (0.07)ns
(H2a) (H2b)
Team goal cue (input as covariate) .07 (0.07)ns .21 (0.05)*** .07 (0.06)ns .02 (0.07)ns .19 (0.05)*** .06 (0.06)ns
(H4a) (H4b)
Type of collaboration (input as covariate) .04 (0.12)ns .08 (0.08)ns .07 (0.09)ns .04 (0.12)ns .08 (0.08)ns .10 (0.09)ns
(HE-HE = 0)
Process fluency (input as M1) .40 (0.05)*** .36 (0.05)***
(H5)
Team cohesion (input as M2) .36 (0.08)*** .35 (0.08)***
(H6)
Control variables
Perceived job complexity .01 (0.06)ns 0.003 (0.04)ns .09 (0.05)ns
Perceived job significance .30 (0.06)*** .15 (0.04)*** .08 (0.05)ns
Technology affinity .04 (0.07)ns 0.03 (0.04)ns .10 (0.05)ns
Prior experience with DE .04 (0.08)ns .03 (0.05)ns .03 (0.06)ns
Knowledge about DE 0.10 (0.08)ns 0.04 (0.05)ns 0.02 (0.06)ns
Model summary
R2 .10 .61 .47 .20 .63 .50
MSE .91 .40 .54 .83 .38 .52
F 9.73*** 129.56*** 44.30*** 7.59*** 52.09*** 24.58***
Df 3, 250 3, 250 5, 248 8, 245 8, 245 10, 243

***p < .01; **p < .05; *p < .10.

agent who, being unable to help, refers the customer to another asking the participants to indicate whether EduBot or Sam had
agent) in increasing customer perception of service quality greater control over the process on a 7-point scale ( 3 to 3 with
(Wang, Hoegg, and Dahl, 2018). This research provides evi- 0 as the midpoint, implying a balanced power distribution). The
dence of the positive effects of proactive communication about manipulations for coordination cue (MAbsence = 3.16, MPresence =
HE-DE coordination on the positive evaluation of the fluency of 5.51, t (307.52) = 15.24, p < .001), frontline co-presence
the process as well as of the cohesion of the HE-DE team. (Msimultaneous = 2.36, Msequential = 5.98, t (249.38) = 19.54,
Further, our study extends current research on human-robot p < .001), and supervisory cue (MEdubot (DE) = 0.95 MSam
interaction in the service context (Choi, Mattila, and Bolton, (HE) = 0.93, t (313) = 13.32, p < .001) worked as intended.
2021; Huang and Dootson, 2022; Yalcin et al., 2022) by
identifying two critical factors, namely, process fluency and
team cohesion, which exert a positive effect on service Results
satisfaction. Study 2 did not find evidence to support H1a (b (SE) = 0.15
(0.11), p > .10) or H1b (b (SE) = 0.02 (0.09), p > .10), but
confirmed H2a (b (SE) = 0.30 (0.05), p < .01), H2b (b (SE) =
Study 2 0.62 (0.04), p < .01), H5 (b (SE) = 0.27 (0.05), p < .01), and H6
(b (SE) = 0.54 (0.06), p < .01) (for detailed results see part B in
Design and Procedure
Table 2). When HE (vs. DE) acted as a supervisor, our findings
Participants were recruited from MTurk (N = 315). We used the showed a significant positive effect on perception of team
same scenario and procedure as for Study 1. We manipulated cohesion (b (SE) = 0.20 (0.09), p < .05) but not on process
the supervisory cue by placing the word “supervisor” next to the fluency (b (SE) = 0.06 (0.11), p > .10), which support H3b, but
icon for HE or DE (See Web Appendix for details). not H3a.

Manipulation Check Discussion

Study 2 used the same manipulation and measurement for This study successfully replicated the results of Study 1. Prior
coordination cue and frontline co-presence as in the previous research has examined the role of HE (vs. DE) being in the
study. The manipulation of the supervisory cue was checked by position of authority and found that it is likely to lead to a higher
12 Journal of Service Research 0(0)

preference for and reduced resistance towards accepting the evaluation of the service due to increased perception of co-
service (Longoni, Bonezzi, and Morewedge, 2019; Longoni and hesion. The insignificant effect of HE as a supervisor on process
Cian, 2022; Shanks et al., 2021). Aligning with previous fluency and team cohesion, compared to its significant effect in
research, this study also found the positive effect of HE as a Study 2, could be due to Study 3A examining this effect under a
supervisor, but our study does this in a collaborative as opposed simultaneous presence condition. That is, the DE and HE are
to a comparative (HE vs. DE) environment. Additionally, our present at the same time throughout the process, and the cus-
study also suggests that HE’s effectiveness as a supervisor lies tomer may automatically perceive that the human co-worker is
in increasing the customer’s perception of a cohesive team. in charge during the engagement. Thus, such a cue might be
redundant.
Study 3A
Study 3B
Design and Procedure
Design and Procedure
Study 3A aimed to test the joint effect of task and entity
connection cues. Participants were recruited from MTurk (N = We recruited participants from MTurk (N = 262). We used the
294). The manipulations for coordination and supervisory cues same procedure and context as for Study 3A (see Web Appendix
were the same as in our previous studies. The team goal cue was B) with the exception that sequential presence was held constant
manipulated by displaying the following message to the par- across all conditions. The difference between Studies 3A and 3B
ticipants: “Sam and I work together as a team, and our team’s is that we held frontline co-presence constant as a simultaneous
goal is to find you the program and course options that best sequence in Study 3A, and as a sequential sequence in
match your profile,” whereas in the absence condition, this Study 3B.
sentence was not displayed.
Manipulation Check
Manipulation Check
We used the same manipulation checks for coordination, su-
The manipulation checks for coordination cue (MAbsence = 3.23, pervisory and team goal cues as in Study 3A. An independent
MPresence = 5.68, t (227.96) = 14.04, p < .001) and supervisory t-test confirms that our manipulations were effective for co-
cue visibility (MHE as supervisor = 1.03, MDE as supervisor = 1.09, ordination cue (MAbsence = 2.93, MPresence = 5.58, t (156.06) =
t (242.50) = 13.58, p < .001) worked as intended. The team goal 14.24, p < .001); supervisory cue (MHE as supervisor = 1.12,
cue was checked by asking participants to indicate their MDE as supervisor = 1.09, t (260) = 13.03, p < .001), and team
agreement with whether Edubot and Sam had a team goal (1 = goal cue (MAbsence = 3.14, MPresence = 5.66, t (222.11) = 15.77,
strongly disagree and 7 = strongly agree). The manipulation p < .001).
worked as intended (MPresence = 5.72, MAbsence = 3.21, t
(233.05) = 15.14, p < .001).
Results
The results are in Part D of Table 2. The coordination cue had a
Results positive impact on process fluency (b (SE) = 0.16 (0.07), p <
The results are in part C in Table 2. With the exception of H3a (b .05) and team cohesion (b (SE) = 0.53 (0.05), p < .001), which
(SE) = 05 (0.12), p > .10) and H3b (b (SE) = 0.12 (0.09), p > reinforces H2a and H2b. We found evidence that HE as su-
.10), this study replicates and confirms previous hypotheses pervisor enhances process fluency in this study (b (SE) = 0.26
including H2a (b (SE) = 0.18 (0.07), p < .01), H2b (b (SE) = (0.13), p < .05) but not team cohesion (b (SE) = 0.03 (0.09),
0.44 (0.05), p < .01), H5 (b (SE) = 0.25 (0.05), p < .01), and H6 p > .10), which provides support for H3a, but not for H3b. The
(b (SE) = 0.46 (0.07), p < .01). This study tested two new team goal cue enhanced process fluency (b (SE) = 0.21 (0.07),
hypotheses and found a weak significant positive effect of a p < .001) and cohesion (b (SE) = 0.31 (0.05), p < .001), which
team goal cue on process fluency (b (SE) = 0.12 (0.07), p < .10) supports H4a and H4b. We also found a significant positive
and a strong positive effect of the team goal on team cohesion effect of process fluency (b (SE) = 0.34 (0.05), p < .001) and
(b (SE) = 0.30 (0.05), p < .01), which confirms H4a and H4b. team cohesion (b (SE) = 0.43 (0.08), p < .001) on service
satisfaction, which support H5 and H6.
Discussion
Discussion
In addition to confirming the findings of our previous studies,
this study also shows that the presence of a team goal enhances Study 3B successfully replicated the results of Study 3A under
perception of team cohesion. Building on previous research that sequential presence. Regarding the inconsistent impact of the
underlined that DE is considered to be a partner to HE (Noble HE as the supervisor, this might be due to the sequential setting
et al., 2022), this study provides evidence that creating a visible implying that the HE is handling more critical tasks and is
partnership between HE and DE can lead to enhanced customer responsible for monitoring the interaction (see also Longoni,
Le et al. 13

Bonezzi, and Morewedge, 2019). However, the impact of the impact of team goal cue on team cohesion (b (SE) = 0.19 (0.07),
HE as the supervisor is not strong enough to influence perceived p < .01), but not on process fluency (b (SE) = 0.02 (0.07), p >
team cohesion as this cue alone might not be a strong signal .10). The results also demonstrate the robust effect of team
compared to the coordination cue. Although the sequencing of cohesion (b (SE) = 0.35 (0.08), p < .01) and process fluency (b
how the HE and DE process their tasks (simultaneous or se- (SE) = 0.36 (0.05), p < .01) on satisfaction. Types of collab-
quential) has been proposed as an important aspect of HE-DE oration (HE-HE = 0; HE-DE = 1) have no major direct impact
team design (Le, Sajtos, and Fernandez, 2023) in a customer- (see Table 2, part E).
facing setting, the role that the supervisory cue plays may
depend upon the way in which the work processing plan is
designed, as to whether it influences either customer perception Discussion
of team cohesion or process fluency. Study 4 replicates all previous results in a different context,
other than the influence of team goal on process fluency. This
might be due to the nature of the service offering—that is the
Study 4 financial advising context is information-oriented whereas
Design and Procedure educational advising is people-oriented. In people-oriented
services, human relationship is the focus, and such a cue
Our previous studies have shown that both coordination and might signify a joint motivation from the service team, which
team goal cues have a consistent impact on perception of might affect a customer’s confidence in the team’s capability to
process fluency and team cohesion. We decided to examine the provide consistent service and minimize potential mistakes
generalizability of these effects in a different context (a finance (Wang, Hoegg, and Dahl, 2018). In information-oriented ser-
consulting setting). Study 4 also extends our previous studies by vices, efficiency is the focus. Customers may be less sensitive to
examining the impact of these cues under sequential co- the team goal cue in evaluating fluency—in this context, they
presence (similarly to Study 3B) and by comparing HE-DE may be more concerned about the accuracy of the work. Thus, a
and HE-HE collaboration. Further, we employed two binary subjective stimulus like a team goal cue might not be necessary
outcomes reflecting the customer’s behavioral intention, by to activate the perception of fluency.
asking the participants whether or not they would be willing to Previous research has underlined the importance of dis-
accept or reject the team’s recommendation (Longoni, Bonezzi, closing whether the provider is a human or non-human agent
and Morewedge, 2019), and to subscribe to the company’s (Huang and Dootson, 2022; Luo et al., 2019), and our study
mailing list (i.e., post-engagement behavior). We employed a 2 complements this stream by highlighting that disclosing the
(Coordination: Presence vs. Absence) x 2 (Team goal: Presence types of team composition can cause customers to behave
vs. Absence) x 2 (Type of collaboration: HE-HE vs. HE-DE) differently in a HE-DE collaborative service setting. In the
between-subject design. The design of this study is in the Web presence of a team goal cue, the HE-DE team (vs. HE-HE team)
Appendix. We recruited participants from Prolific. In the ex- generates the highest likelihood of accepting the team’s rec-
periment, we also controlled additional job characteristics ommendation. As there is likely to be a certain degree of bias in
(complexity and significance—e.g., Castelo, Bos, and human team decision-making, communicating such a shared
Lehmann, 2019), and the customers’ technology affinity goal in an HE-DE team setting might lead to a better impression
(Belanche, Casaló, and Flavián, 2021) and prior experience with of the HE-DE’s team ability to minimize such human bias, thus
and knowledge about DEs (Luo et al., 2019). The results of improving customer confidence in the HE-DE team’s recom-
measurement validity of the additional control variables are in mendation. Additionally, we observed that the overall likeli-
Web Appendix D. hood of subscribing to the provider’s e-mailing list is
significantly smaller compared to the likelihood of accepting
Manipulation Check service team’s recommendation. However, the presence of a
coordination cue in HE-DE team makes customers more likely
We checked the manipulation of coordination and team goal to subscribe than when in the presence of such a cue in HE-
cues by using the same approach as in the previous studies. HE team.
Results confirmed that the manipulations (coordination cue –
MAbsence = 3.41, MPresence = 5.58, t (252) = 12.37, p < .001; team
goal cue – MAbsence = 3.10, MPresence = 5.43, t (252) = 13.68, p < Study 5
.001) had been successful.
Design and Procedure
Earlier we underlined the differences between substitution,
Results augmentation, and collaboration approaches. In this study, we
The results are in Table 2, part E. Results confirm the replication aim to compare the impact of HE-DE collaboration (through
of the effects of coordination cue on process fluency (b (SE) = coordination and team goal cues) to that of substitution and
0.25 (0.07), p < .01) and on team cohesion (b (SE) = 0.64 (0.05), augmentation. We utilized a similar experimental setting as in
p < .01) in a different setting.2 The results also confirm the Longoni, Bonezzi, and Morewedge (2019). Participants were
14 Journal of Service Research 0(0)

recruited from Prolific (N = 161) to read an advertisement about recommendation likelihood (b (SE) = 0.52 (0.07), p < .001). A
a hypothetical new medical screening service for lung cancer, mediation analysis was conducted and revealed that the impact
offered by the National Health Service (see Web Appendix B for of collaboration is significantly mediated through process
the scenario). Our outcome variables were the likelihood of transparency in influencing usage likelihood (effect (SE) = 0.25
using and recommending the medical service. Since this study (0.10), CI95% = [0.08 to 0.45]) and recommendation likelihood
underlines the importance of communicating the HE-DE col- (effect (SE) = 0.28 (0.11), CI95% = [0.08 to 0.50]) and this
laboration cues to customers, we measured perceived service mediation effect is greater than that of an augmentation type of
process transparency (see also Le, Sajtos, and Fernandez, 2023). service provision.
We measured this concept on a single item, where participants
indicated how would they rate the level of transparency of the
process (1 = extremely not transparent and 7 = extremely Discussion
transparent). We anticipated that the communication of coor- In the literature, Longoni, Bonezzi, and Morewedge (2019)
dination and team goal cues in an HE-DE collaboration would found that customers prefer a service provision where a human
lead to a greater sense of transparency, as customers can observe employee plays a major role in making decisions and DE plays
the steps and actions that the service agent executes, which will the support role. Our first analysis corroborates the result: we
eventually lead to better service outcomes. found that customers are more inclined to use and recommend
the service under the augmentation setting. Further, although
prior research has discussed the importance of a transparent
Results service (Grewal Dhruv et al., 2020; Le et al., 2023), no research
We analyzed process transparency as the mediating mechanism has provided empirical evidence as to the role of transparency in
and examined the impact of the type of service provision on explaining the impact of collaboration between HE and DE. We
likelihood of using and recommending through this mecha- have shown that communicating HE-DE collaboration (man-
nism.3 The results are in Table 3. Controlling for customer prior ifested through coordination and team goal cues) creates the
experience, we found that an HE-DE collaboration type of most transparent service as perceived by customers. We also
service provision generated a greater positive impact on process revealed that the effects of HE-DE collaboration on usage and
transparency (b (SE) = 0.54 (0.19), p < .05) when compared to recommendation likelihoods are due to process transparency.
substitution. Similarly, augmentation also generated a greater
level of process transparency (vs. substitution), but this effect is
less than that of HE-DE collaboration (b (SE) = 0.38 (0.19), p =
General Discussion
.051). In turn, process transparency has positive impacts on As service provision is increasingly being managed by both the
usage likelihood (b (SE) = 0.47 (0.07), p < .001) and HE and DE (Huang and Rust, 2022; Noble et al., 2022), this

Table 3. Regression Results of Process Transparency as the Mediation Mechanism.

PROCESS Model 4 (Separated Analyses

on DVs)—Standardized Score Process Transparency (3) Usage Likelihood (4) Recommendation Likelihood (5)

Constant 0.37 (0.30)ns 0.19 (0.26)ns 0.22 (0.26)ns

Substitution (base line—coded 0) — — —
Augmentation (1) .38 (0.19)* (p = 0.051) .35 (0.17)** .28 (0.17)ns
Collaboration (2) .54 (0.19)*** 0.01 (0.17)ns .04 (0.17)ns
Process transparency (3) .47 (0.07)*** .52 (0.07)***
Control variables
Prior experience with DE .03 (0.10)ns .03 (0.09)ns .05 (0.09)ns
Model summary
R2 .05 .26 .29
MSE .97 .76 .73
F 2.77** 13.70*** 16.10***
Df 3, 157 4, 156 4, 156

Mediation Analysis Effect (SE) CI95%

(1) → (3) → (4) .18 (0.10) [-0.004 to 0.40]

(2) → (3) → (4) .25 (0.10) [0.08 to 0.45]
(1) → (3) → (5) .19 (0.11) [-0.01 to 0.45]
(2) → (3) → (5) .28 (0.11) [0.08 to 0.50]
***p < 0.01; **p < 0.05; *p < 0.10.
Le et al. 15

research aimed to understand the effectiveness of HE-DE strategic preparation for a future workforce landscape that in-
collaborative cues in service provision. Prior research has creasingly relies on human and robotic technology collabora-
highlighted how DEs can, in some circumstances, replace HEs tion, which is essential for optimizing service delivery. Hence,
entirely in service provision (Xiao and Kumar, 2021) or how this research serves as a starting guide for future initiatives
they can augment HEs in the decision-making process aiming to harness the full potential of human-robot collabo-
(Longoni, Bonezzi, and Morewedge, 2019). However, such ration in the service industry.
research has not fully explored the actual collaboration process
between the two agents and how customers perceive it.
Theoretical Implications
The experimental studies illuminate insights into the dy-
namics of HE-DE collaborative cues as perceived by cus- This research draws on interdependence theory (Courtright
tomers. The frontline co-presence cue was tested in Study et al., 2015; Rusbult and Van Lange, 2008) to conceptualize
1 and Study 2, and it has an insignificant impact on either HE-DE collaborative cues through the notion of task- and
process fluency or team cohesion. The HE-DE coordination entity-based interdependent behavioral cues. The insight from
cue was tested in Study 1 to 4, and we observed a consistent this research is novel in several ways. First, we develop the
and positive impact of this cue on both mechanisms. The concept of HE-DE collaboration using two types of behavioral
robustness of the HE-DE coordination cue underscores its cues, namely task and entity connection cues. Unlike previous
potential as a cornerstone in showcasing the effectiveness of a research (Choi, Mattila, and Bolton, 2021; Longoni, Bonezzi,
hybrid HE-DE service team to customers in the future. The and Morewedge, 2019), our research takes a process view of
insignificant impacts of frontline co-presence cue should not collaboration and distinguishes it from augmentation and
lead to a dismissal of its value. Instead, it should motivate substitution. Our research reveals that HE-DE service team
future work to examine the conditions under which frontline collaborative cues influence customers’ likelihood of using and
co-presence might be more influential. recommending due to their perception of a transparent process.
The effect of supervisory cue was tested in Study 2, 3A, Second, this research also responds to previous calls (Huang
and 3B and it was not consistent. In Study 2, it has a positive and Rust, 2022; Wirtz et al., 2018; Xiao and Kumar, 2021) by
impact on team cohesion, but not process fluency. In Study identifying the mechanisms that are effective in enabling a
3A, holding the workflow sequence as simultaneous presence smooth integration of DEs into human work structure, namely,
in the experiment, the supervisory cue has no significant process fluency and team cohesion. Our research finds that task
impact on either process fluency or team cohesion. In Study connection (manifested through coordination cue) has the
3B with a sequential presence, the supervisory cue has a greatest impact on customers’ evaluation of the service through
positive impact on process fluency, but not on team cohesion. a perception of HE-DE team cohesion over process fluency
The team goal cue was tested in Study 3A, 3B, and 4. This cue (Web Appendix H). To this end, this is the first study that
has a positive impact on both process fluency and team explores the HE-DE team cohesiveness as perceived by the
cohesion (Study 3A and Study 3B), but only the positive customers, and not only highlights its importance in driving
impact on team cohesion was replicated in Study 4. The satisfaction, but also reveals how this cohesion is created. While
inconsistent impact of the supervisory cue highlights the prior research highlights the importance of enhancing the ex-
possible influence of sequencing the presence of HE and DE, perience of customers by providing them with a smooth and
which can affect how customers perceive the power dynamic effortless process (Lemon and Verhoef, 2016), our research
of the two entities. The team goal cue demonstrates to the suggests that the impression of a united service team
customers a commitment of HE-DE service team to meeting (i.e., cohesion) is far more important.
customer needs and it reassures customers that the DE is Third, considering only task-connection cues, our research
being used to enhance their experience, not replace the hu- finds that the coordination cue has a significant, large, and
man touch in service delivery. Additionally, in Study 5, the consistent effect in creating an impression of a smooth process
HE-DE collaboration as a service delivery mode has a and a cohesive HE-DE team. Prior research has demonstrated
positive impact on process transparency, which in turn, drives that revealing the coordination when the initial service agent is
usage and recommendation likelihoods. This result shows unable to address a customer’s request can yield favorable
that HE-DE collaborative cues can also help build confidence results for customer-related outcomes within a human team
in the HE-DE team through the notion of process transpar- environment (Wang, Hoegg, and Dahl, 2018). However, this
ency, which allows customers to understand how decisions result primarily underscores the reactive aspect of coordination,
within the HE-DE service team are made. These findings as it is exposed to the customer only when there is a service
suggest that when consumers are effectively informed about issue. Our research conceptualizes coordination cues as a
the collaborative nature of a HE-DE team, it could enhance proactive communication of taskwork orchestration. Through
their outcomes. this communication, including the transparent tasks assignment
Taken altogether the empirical studies, this research high- and clear task handovers between the HE and DE, the cus-
lights the innovative nature and efficacy of conveying collab- tomer’s perception of a seamless process and a cohesive HE-DE
orative cues to customers in the context of the HE-DE service team can be improved significantly, which result in heightened
team. This research underscores the growing importance of satisfaction.
16 Journal of Service Research 0(0)

In terms of the frontline co-presence cue, while previous improve customer experience with such a service team in the
research has highlighted the importance of frontline co-presence future. Both task- and entity-based connection cues represent
in influencing customer attitude (Choi, Mattila, and Bolton, different ways that firms can showcase the HE-DE collab-
2021; Peng et al., 2022), this research finds no support for the oration to customers. Firms can choose to infuse this col-
impact of a frontline co-presence cue. This may be because prior laboration through the service process with task-based
research focused only on social co-presence in either an aug- connection cues or through the HE-DE service team rela-
mentation or a substitution setting (e.g., Longoni, Bonezzi, and tionship with entity-based connection cues. Either approach
Morewedge, 2019), where the DE could either replace human could lead to a better experience for customers.
presence completely or play only a supportive role (e.g., Peng In terms of task-based connection cues, only coordination
et al., 2022). These approaches are likely to magnify the im- has a consistently positive impact on both process fluency and
portance of the frontline co-presence cue of the DE. In contrast, team cohesion. Hence, firms should develop a clear commu-
the collaboration approach is likely to diminish the role of the nication strategy for the HE-DE team as they are interacting
individual and shifts the customer’s attention to the HE-DE with the client. In particular, this communication strategy
service team as a whole and to the team collaborative cues. should include the task allocated to each service agent, and more
However, the cross-studies meta-analysis of the overall effect importantly, the information that each entity captures, handles,
size of frontline co-presence illustrates a negative impact on transfers, and receives. This could create a better impression of a
process fluency in the sequential setting (see Web Appendix I). fluent process and a cohesive team as we hypothesized and
Thus, the potential of frontline co-presence cue in HE-DE tested. For instance, AiChat, a DE solution provider from
collaboration requires further examination. Singapore, creates DEs that are capable of notifying customers
Fourth, regarding entity connection cues, previous research when their requests are being transferred to a human employee
has highlighted the importance of having the HE in the position through a ticketing system. This notification is made visible to
of authority (Longoni, Bonezzi, and Morewedge, 2019; Peng the customer through a live chat console. Information such as
et al., 2022; Yalcin et al., 2022). Our research adds to the lit- attached documents and photos that the customer provides to the
erature by suggesting that the benefits of an HE as the supervisor DE is stored and associated with their ticket, and when the DE is
may depend on the sequence of co-presence. In a simultaneous ready to be handed over, it notifies the customer that they are
co-presence setting, the supervisory cue might be redundant. being transferred to an HE. All the associated files are also
However, in a sequential setting where the HE acts as the handed over to the HE so that they can pick up the case in the
downstream agent who receives information from the DE, this same chat console and work on the customers’ request im-
supervisory cue could have an impact on process fluency (Study mediately without the need to ask customers to repeat that
3B). The presence of a human supervisor as a communication information or upload the files again.
signal seems relevant only when the human is not ‘in sight’ and Firms should also communicate to customers a message
is involved only in the latter part of an interaction. In com- about a joint service goal, which represents the HE-DE team’s
plementing previous research that highlighted the importance of commitment to provide high service standards and it is also an
human supervision (Choi, Mattila, and Bolton, 2021), this indication of a well-trained team that is capable of handling
research underlines that the presence of a human supervisor customer requests. For instance, Virtual Vet Nurse partners with
could lead to the team being perceived as more cohesive (Study Dandenong Ranges Veterinary Centre to deploy a DE—
2) and the process being perceived as more fluent under a Sophie—that works with onsite veterinarians. The DE com-
sequential presence (Study 3B). However, this inconsistent municates to the customers that the vet team has a joint intention
result needs to be interpreted with care and takes into account to provide accessible and the best possible veterinary support
the role of workflow sequencing as we can only partially verify for pets. Our research suggests that firms should frame such a
these hypotheses in individual studies. message at a team level to include the DE in the communication
In terms of team goal cues, showcasing HE-DE team more explicitly. Such a signal in service provision will reinforce
commitment to serve the customers can enhance their per- customer confidence in the team’s capability and motivation,
ception of a fluent process and a united team. Research has which would ultimately enhance the impression they give of
highlighted the importance of a shared goal in human-robotic team cohesion and process fluency.
agent teams (Noble et al., 2022). Our findings complement this The explicit communication of both the coordination and
literature stream demonstrating that communicating the shared goals underlines the importance of process transparency. Our
goal of the HE-DE service team to customers can enhance their Study 5 also directly supports the importance of process
perception of both a fluent process and a cohesive HE-DE team. transparency. In addition to the collaborative cues that we
examined in this research, firms could also enhance process
transparency by providing customers with a summary of all of
Managerial Implications their interactions (e.g., between the customer and the service
This research investigates the potency of communicating HE- team or between HE and DE) at the end of their chat session so
DE collaborative cues in service provision. We offer several that they can review the conversations for any information
suggestions that may assist managers and practitioners to provided by the service providers. This will also reduce cus-
embrace HE-DE collaboration, and to a larger extent to tomer effort, as all activities in the conversation are tracked and
Le et al. 17

documented, so customers do not have to memorize information We acknowledge that these cues could also be applicable to
provided by the service agents after the chat. human-based collaborations (e.g., HE-HE team context). In this
Further, we have also demonstrated that both mechanisms current research, the additional analysis of Study 4 (see Web
(process fluency and team cohesion) play important roles in Appendix F) did not indicate support evidence for the effect of
explaining the impact of communicating HE-DE collaborative team setting (HE-HE vs. HE-DE) in moderating the impacts of
cues on customer satisfaction. We showed through a compar- the collaboration cues on process fluency and team cohesion
ison that the mediation effect of an HE-DE team cohesion is and the subsequent impacts of the mediators on customer
consistently larger than that of process fluency in terms of satisfaction. This could be due to our Study 4 examining col-
explaining the impact of the collaborative cue on satisfaction laborative cues from the customer perspective. This means that
(see Web Appendix H). Accordingly, we suggest that firms participants did not engage in the collaboration and hence, they
should prioritize improvements in collaborative cues that en- might not have been sensitive to the difference of team settings.
hance team cohesion. Future research could focus explicitly on this research avenue,
comparing the effectiveness of these collaboration cues on
process fluency and team cohesion in two different team set-
Limitations and Future Research tings, especially from the employee perspective. Further, we
This research also has several limitations that need to be ac- encourage such investigations to be conducted in conjunction
knowledged when interpreting the results. First, this research with other potential variables or in different situations. Given
uses scenario-based experiments, which could have conse- the recent emergence of human-robotic agent collaboration in
quences on the findings. Participants might find it difficult to the literature, this research only scratches the surface of this
imagine the simultaneous sequence because of its unpopularity phenomenon. Thus, it is possible that there are other situations
in practice. Despite having satisfactory scores on the realism where the impacts of these cues might be different between the
and believability checks in all studies (see Web Appendix), the HE-DE and HE-HE team setting (e.g., handling complaints),
results of this research should be interpreted with care. Previous which could be fruitful for future research.
research has provided evidence that the physical presence of
robots creates anxiety, resulting in compensatory behavior (e.g., Declaration of Conflicting Interests
Mende et al., 2019). The lack of significant effects from the The author(s) declared no potential conflicts of interest with respect to
presence of DEs in our research could be attributed to DEs being the research, authorship, and/or publication of this article.
digital. Physical robots are more noticeable when in close
proximity to humans and might elicit a stronger response to their
Funding
frontline co-presence.
Second, although the manipulations across all of our studies The author(s) received no financial support for the research, authorship,
worked as intended, this research acknowledges that the results and/or publication of this article.
should be interpreted with care. We demonstrated in the Web
Appendix that the manipulations of team goal and coordination ORCID iDs
cues were not independent. Further, the utilization of single- Khanh B. Q. Le  https://orcid.org/0000-0002-9271-0098
item measurements such as satisfaction is another limitation of Werner H. Kunz  https://orcid.org/0000-0001-6264-183X
the current research. Thus, scholars and practitioners need to
take these limitations into account when interpreting the results. Supplemental Material
Third, the effect of a supervisory cue on process fluency or
Supplemental material for this article is available online
team cohesion can only be partially verified in individual
studies. Ultimately, we reject the hypothesized effects based on
the synthesized effect size (see Figure 1). When tasks are Notes
perceived as highly consequential, customers do not trust al- 1. A team typically consists of two or more entities that exhibit in-
gorithms to make the decision (Castelo, Bos, and Lehmann, terdependencies (Kozlowski and Ilgen, 2006). In our research, we
2019). Further, evidence also suggests that customers prefer a consider an HE and a DE to be part of a frontline customer service
human to be in the position of authority (Longoni, Bonezzi, and team and this “HE-DE” team is a sub-set of teams.
Morewedge, 2019). Future research should explore the cir- 2. The results on behavioral outcome are in Web Appendix F. We
cumstances under which the impact of supervisory cues is found a negative effect of a team goal cue on customer’s decision to
important to customer evaluations of cohesiveness and fluency. accept the service team’s recommendation (b (SE) = 0.43 (0.18),
Future research could also examine the conditions under which p < .05). Further, this effect is moderated by the types of collab-
the effects of behavioral interdependence cues on either process oration (b (SE) = 0.54 (0.25), p < .05). Activating a team goal cue in
fluency or team cohesion are amplified or mitigated. For ex- HE-DE collaboration can result in a higher likelihood of accepting
ample, future research could examine how customers perceive the team’s recommendation (71.78%) compared to HE-HE col-
who (HE or DE) defines the goals (Noble et al., 2022). laboration (54.70%). However, in the absence of a team goal cue,
Fourth, the current research aims to understand the impacts the effect is reversed. For subscribing behavior, we found no direct
of these collaboration cues mainly in the HE-DE team context. effect from coordination, team goal cue or the types of
18 Journal of Service Research 0(0)

collaboration. However, we found a marginal significant interaction Infusion: Conceptual Archetypes and Future Research Direc-
effect between coordination cue and types of collaboration (b (SE) = tions,” Journal of Service Management, 30 (1), 156-183.
0.53 (0.27), p = 0.051). Activating the coordination cue in HE-DE Durrande-Moreau, Agnes and Jean-Claude Usunier (1999), “Time
collaboration generates a greater propensity to subscribe to email Styles and the Waiting Experience: An Exploratory Study,”
communication (33.02%) compared to HE-HE collaboration Journal of Service Research, 2 (2), 173-186.
(11.96%). When this cue is deactivated, the effect is reversed, but Fuscaldo, Donna (2019), “Bank of America’s Virtual Assistant Now
the difference is minimal. Has More Than 10 Million Users,” Forbes, available at: https://
3. We analyzed the impacts of the type of service delivery mode on www.forbes.com/sites/donnafuscaldo/2019/12/11/bank-of-
perceived transparency. The results are in Web Appendix G. In line americas-virtual-assistant-now-has-more-than-10-million-users/?
with our prediction, HE-DE collaboration induces a greater level of sh=496db2e9f69b (accessed January 12, 2023).
process transparency when compared to both augmentation and Garry, Tony and Tracy Harwood (2019), “Cyborgs as Frontline Service
substitution. However, a one-way ANOVA also reveals significant Employees: A Research Agenda,” Journal of Service Theory and
differences in the likelihood of using and recommending service Practice, 29 (4), 415-437.
across the types of service provision in which augmentation is the Garvey, Aaron M., TaeWoo Kim, and Duhachek Adam (2023), “Bad
preferred type of service delivery mode. News? Send an AI. Good News? Send a Human,” Journal of
Marketing, 87 (1), 10-25.
Gracia, Esther, Eva Cifre, and Rosa Grau (2010), “Service Quality: The
References
Key Role of Service Climate and Service Behavior of Boundary
Beal, Daniel J., Robin R. Cohen, Michael J. Burke, and Christy L. Employee Units,” Group & Organization Management, 35 (3),
McLendon (2003), “Cohesion and Performance in Groups: A 276-298.
Meta-Analytic Clarification of Construct Relations,” Journal of Grewal Dhruv, Kroschke Mirja, Mende Martin, L. Roggeveen Anne,
Applied Psychology, 88 (6), 989-1004. and L. Scott Maura (2020). Frontline Cyborgs at Your Service:
Belanche, Daniel, Luis V. Casaló, and Carlos Flavián (2021), How Human Enhancement Technologies Affect Customer Ex-
“Frontline Robots in Tourism and Hospitality: Service En- periences in Retail, Sales, and Service Settings. Journal of In-
hancement or Cost Reduction?” Electronic Markets, 31 (3), teractive Marketing, 51(1), 9–25.
477-492. Hayes, Andrew F. (2017). Introduction to Mediation, Moderation, and
Botnation.ai (2022), “The Advanced Mode of Escalation to Human,” Conditional Process Analysis: A Regression-Based Approach,
Botnation.ai, available at: https://botnation.ai/site/en/climbing-to- 2nd ed. New York: The Guilford Press.
human/#The_conversation_on_the_user_side (accessed January Hew, Khe Foon, Weijiao Huang, Jiahui Du, and Chengyuan Jia (2022),
18, 2023). “Using Chatbots to Support Student Goal Setting and Social
Cannon-Bowers, Janis A. and Clint Bowers (2011), “Team Devel- Presence in Fully Online Activities: Learner Engagement and
opment and Functioning,” in APA Handbook of Industrial and Perceptions,” Journal of Computing in Higher Education, 35 (1),
Organizational Psychology, Vol. 1: Building and Developing the 40-68.
Organization, S. Zedeck, ed. Washington, DC: American Psy- Huang, Ming-Hui and Roland T. Rust (2018), “Artificial Intelligence in
chological Association, 597-650. Service,” Journal of Service Research, 21 (2), 155-172.
Cassab, Harold and Douglas L. MacLachlan (2006), “Interaction Huang, Ming-Hui and Roland T. Rust (2022), “A Framework for
Fluency: A Customer Performance Measure of Multichannel Collaborative Artificial Intelligence in Marketing,” Journal of
Service,” International Journal of Productivity and Performance Retailing, 98 (2), 209-223.
Management, 55 (7), 555-568. Huang, Yu-Shan Sandy and Paula Dootson (2022), “Chatbots and
Castelo, Noah, Maarten W. Bos, and Donald R. Lehmann (2019), Service Failure: When Does it Lead to Customer Aggression,”
“Task-Dependent Algorithm Aversion,” Journal of Marketing Journal of Retailing and Consumer Services, 68, 1-11.
Research, 56 (5), 809-825. Jeon, Yongwoog Andrew (2022), “Let Me Transfer You to Our AI-
Choi, Sungwoo, Anna S. Mattila, and Lisa E. Bolton (2021), “To Err Based Manager: Impact of Manager-Level Job Titles Assigned to
is Human (-oid): How do Consumers React to Robot Service AI-Based Agents on Marketing Outcomes,” Journal of Business
Failure and Recovery?” Journal of Service Research, 24 (3), Research, 145 (5), 892-904.
354-371. Johnson, David W. and Roger T. Johnson (2009), “An Educational
Courtright, Stephen H., Gary R. Thurgood, Greg L. Stewart, and Psychology Success Story: Social Interdependence Theory and
Abigail J. Pierotti (2015), “Structural Interdependence in Teams: Cooperative Learning,” Educational Researcher, 38 (5), 365-379.
An Integrative Framework and Meta-Analysis,” Journal of Ap- Johnson, Matthew, Jeffrey M. Bradshaw, and Paul J. Feltovich (2018),
plied Psychology, 100 (6), 1825-1845. “Tomorrow’s Human–Machine Design Tools: From Levels of
De Keyser, Arne and Werner H. Kunz (2022), “Living and Working Automation to Interdependencies,” Journal of Cognitive Engi-
with Service Robots: A TCCM Analysis and Considerations for neering and Decision Making, 12 (1), 77-82.
Future Research,” Journal of Service Management, 33 (2), Kim, Jun Hyung, Minki Kim, Do Won Kwak, and Sol Lee (2022),
165-196. “Home-Tutoring Services Assisted with Technology: Investi-
De Keyser, Arne, Sarah Köcher, Linda Alkire, Cédric Verbeeck, and gating the Role of Artificial Intelligence Using a Randomized
Jay Kandampully (2019), “Frontline Service Technology Field Experiment,” Journal of Marketing Research, 59 (1), 79-96.
Le et al. 19

Kozlowski, Steve W. J. and Daniel R. Ilgen (2006), “Enhancing the Orth, Ulrich R. and Jochen Wirtz (2014), “Consumer Processing of
Effectiveness of Work Groups and Teams,” Psychological Science Interior Service Environments: The Interplay Among Visual
in the Public Interest, 7 (3), 77-124. Complexity, Processing Fluency, and Attractiveness,” Journal of
Lake, Katrina (2018), “Stitch Fix’s CEO on Selling Personal Style to Service Research, 17 (3), 296-309.
the Mass Market,” Harvard Business Review, available at: https:// Peng, Chenming, Jenny van Doorn, Felix Eggers, and Jaap E. Wieringa
hbr.org/2018/05/stitch-fixs-ceo-on-selling-personal-style-to-the- (2022), “The Effect of Required Warmth on Consumer Accep-
mass-market (accessed December 21, 2021). tance of Artificial Intelligence in Service: The Moderating Role of
Le, Khanh B. Q., Laszlo Sajtos, and Karen V. Fernandez (2023), AI-Human Collaboration,” International Journal of Information
“Employee-(ro)bot Collaboration in Service: An Interdependence Management, 66 (5), 1-11.
Perspective,” Journal of Service Management, 34 (2), 176-207. Rusbult, Caryl E. and Paul A. M. Van Lange (2008), “Why We Need
Lemon, Katherine N. and Peter C. Verhoef (2016), “Understanding Interdependence Theory,” Social and Personality Psychology
Customer Experience Throughout the Customer Journey,” Compass, 2 (5), 2049-2070.
Journal of Marketing, 80 (6), 69-96. Seeber, Isabella, Eva Bittner, Robert O. Briggs, Triparna De Vreede,
Longoni, Chiara and Luca Cian (2022), “Artificial Intelligence in Gert-Jan De Vreede, Aaron Elkins, Ronald Maier, Alexander B.
Utilitarian vs. Hedonic Contexts: The “Word-of-Machine” Ef- Merz, Sarah Oeste-Reiß, Nils Randrup, Gerhard Schwabe, and
fect,” Journal of Marketing, 86 (1), 91-108. Matthias Söllner (2020), “Machines as Teammates: A Research
Longoni, Chiara, Andrea Bonezzi, and Carey K. Morewedge (2019), Agenda on AI in Team Collaboration,” Information & Man-
“Resistance to Medical Artificial Intelligence,” Journal of Con- agement, 57 (2), 1-22.
sumer Research, 46 (4), 629-650. Shanks, Ilana, Maura Scott, Martin Mende, Jenny van Doorn, and
Luo, Xueming, Siliang Tong, Zheng Fang, and Zhe Qu (2019), Dhruv Grewal (2021). Power to the Robots!? How Consumers
“Frontiers: Machines vs. Humans: The Impact of Artificial In- Respond to Robotic Leaders in Cobotic Service Teams. Marketing
telligence Chatbot Disclosure on Customer Purchases,” Market- Science Institute Working Paper Series, 21-128.
ing Science, 38 (6), 937-947. Shen, Xiao-Liang, Yang-Jun Li, Yongqiang Sun, and Nan Wang (2018),
Marinova, Detelina, Jun Ye, and Jagdip Singh (2008), “Do Frontline “Channel Integration Quality, Perceived Fluency and Omnichannel
Mechanisms Matter? Impact of Quality and Productivity Orien- Service Usage: The Moderating Roles of Internal and External Usage
tations on Unit Revenue, Efficiency, and Customer Satisfaction,” Experience,” Decision Support Systems, 109, 61-73.
Journal of Marketing, 72 (2), 28-45. Shrestha, Yash Raj, Shiko M. Ben-Menahem, and Georg Von Krogh
Marks, Michelle A., John E. Mathieu, and Stephen J. Zaccaro (2001), (2019), “Organizational Decision-Making Structures in the Age of
“A Temporally Based Framework and Taxonomy of Team Pro- Artificial Intelligence,” California Management Review, 61 (4), 66-83.
cesses,” Academy of Management Review, 26 (3), 356-376. Spataro, Jared (2023), “Introducing Microsoft 365 Copilot – Your
Mathieu, John E., Margaret M. Luciano, Lauren D’Innocenzo, Copilot for Work,” Microsoft, available at: https://blogs.microsoft.
Elizabeth A. Klock, and Jeffery A. LePine (2020), “The Devel- com/blog/2023/03/16/introducing-microsoft-365-copilot-your-
opment and Construct Validity of a Team Processes Survey copilot-for-work/(accessed August 1, 2023).
Measure,” Organizational Research Methods, 23 (3), 399-431. Sun, Yongqiang, Chaofan Yang, Xiao-Liang Shen, and Nan Wang (2020),
Mathieu, John E., Michael R. Kukenberger, Lauren D’innocenzo, and “When Digitalized Customers Meet Digitalized Services: A Digi-
Greg Reilly (2015), “Modeling Reciprocal Team Cohesion– talized Social Cognitive Perspective of Omnichannel Service Usage,”
Performance Relationships, as Impacted by Shared Leadership International Journal of Information Management, 54, 1-13.
and Members’ Competence,” Journal of Applied Psychology, 100 Van de Ven, Andrew H., Andre L. Delbecq, and Richard Koenig Jr
(3), 713-734. (1976), “Determinants of Coordination Modes Within Organi-
McLeay, Fraser, Victoria Sophie Osburg, Vignesh Yoganathan, and zations,” American Sociological Review, 41 (2), 322-338.
Anthony Patterson (2021), “Replaced by a Robot: Service Im- Van Der Borgh, Michel, Ad De Jong, and Edwin J. Nijssen (2019),
plications in the Age of the Machine,” Journal of Service “Balancing Frontliners’ Customer-and Coworker-Directed Be-
Research, 24 (1), 104-121. haviors When Serving Business Customers,” Journal of Service
Mende, Martin, Maura L. Scott, Jenny van Doorn, Dhruv Grewal, and Research, 22 (3), 323-344.
Ilana Shanks (2019), “Service Robots Rising: How Humanoid Van Doorn, Jenny, Martin Mende, Stephanie M. Noble, John Hulland,
Robots Influence Service Experiences and Elicit Compensatory Amy L. Ostrom, Dhruv Grewal, and J. Andrew Petersen (2017),
Consumer Responses,” Journal of Marketing Research, 56 (4), “Domo Arigato Mr. Roboto: Emergence of Automated Social
535-556. Presence in Organizational Frontlines and Customers’ Service
Menguc, Bulent, Seigyoung Auh, and Aypar Uslu (2013), “Customer Experiences,” Journal of Service Research, 20 (1), 43-58.
Knowledge Creation Capability and Performance in Sales Teams,” Van Doorn, Jenny, Katherine N. Lemon, Vikas Mittal, Stephan Nass,
Journal of the Academy of Marketing Science, 41 (1), 19-39. Doreén Pick, Peter Pirner, and Peter C. Verhoef (2010), “Cus-
Noble, Stephanie M., Martin Mende, Dhruv Grewal, and A. tomer Engagement Behavior: Theoretical Foundations and
Parasuraman (2022), “The Fifth Industrial Revolution: How Research Directions,” Journal of Service Research, 13 (3),
Harmonious Human–Machine Collaboration is Triggering a 253-266.
Retail and Service [R] evolution,” Journal of Retailing, 98 (2), Van Doorn, Jenny, Edin Smailhodzic, Stefano Puntoni, Jia Li, Jan
199-208. Hendrik, Schumann, and Jana Holthöwer (2023), “Organizational
20 Journal of Service Research 0(0)

Frontlines in the Digital Age: The Consumer–Autonomous Communication, Naples, Italy (31 August–04 September 2020),
Technology–Worker (CAW) Framework,” Journal of Business 1143-1149.
Research, 164, 1-13.
Wageman, Ruth (1995), “Interdependence and Group Effectiveness,” Author Biographies
Administrative Science Quarterly, 40 (1), 145-180.
Khanh B. Q. Le (PhD) is a Lecturer in marketing at Lan-
Wageman, Ruth (1999), “The Meaning of Interdependence,” in Groups
caster University, United Kingdom. His primary research
at Work: Advances in Theory and Research, M Turner, ed.
interests include service management and exploring the
Hillsdale, NJ: Erlbaum, 197-218.
impacts of robotic technologies in service frontline. He has
Walliser, James C., Ewart J. de Visser, Eva Wiese, and Tyler H. Shaw
published his research in the Journal of Service
(2019), “Team Structure and Team Building Improve Human–
Management.
Machine Teaming with Autonomous Agents,” Journal of Cog-
nitive Engineering and Decision Making, 13 (4), 258-278. Laszlo Sajtos (PhD), Associate Professor at the University of
Wang, Chen, JoAndrea Hoegg, and Darren W. Dahl (2018), “The Auckland Business School, researches the impact of emergent
Impact of a Sales Team’s Perceived Entitativity on Customer technologies on service management. His work is published in
Satisfaction,” Journal of the Academy of Marketing Science, 46 top journals including the International Journal of Research in
(2), 190-211. Marketing, Journal of Service Research, and Journal of In-
Wirtz, Jochen, Paul G. Patterson, Werner H. Kunz, Thorsten Gruber, teractive Marketing, among others.
Vinh Nhat Lu, Stefanie Paluch, and Antje Martins (2018), “Brave
Werner H. Kunz (PhD) is Professor of Marketing and Director
New World: Service Robots in the Frontline,” Journal of Service
of the digital media lab at the University of Massachusetts
Management, 29 (5), 907-931.
Boston. His research interests are in service technology, incl. AI,
Xiao, Li and Vikas Kumar (2021), “Robotics for Customer Service: A
service robots, and innovation. Among many honors, he is the
Useful Complement or an Ultimate Substitute?” Journal of
inaugural recipient of the Bo Edvardsson “Industry Impact in
Service Research, 24 (1), 9-29.
Services” Award.
Yalcin, Gizem, Sarah Lim, Stefano Puntoni, and Stijn M. J. van
Osselaer (2022), “Thumbs up or Down: Consumer Reactions Karen V. Fernandez (PhD) is Associate Professor of Mar-
to Decisions by Algorithms Versus Humans,” Journal of Mar- keting at the University of Auckland, New Zealand. Her
keting Research, 59 (4), 696-717. research interests include consumer culture, consumer identity,
Zhao, Fangyun, Curt Henrichs, and Bilge Mutlu (2020), “Task in- technology, and attachment. Her work has appeared in the
terdependence in Human-Robot Teaming,” in 29th IEEE Inter- Journal of Consumer Research, the European Journal of
national Conference on Robot and Human Interactive Marketing, and Internet Research.