Sluggish, but innovative? Orchestrating collaboration in multi-stakeholder networks despite low commitment

ABSTRACT Network orchestrators try to mobilise value from multi-stakeholder settings for innovation. While most of these settings for innovation are dynamic and fast-paced, we explore an empirical case in which members are not deeply engaged initially, whereas the orchestrators themselves are highly active and committed. Building on a qualitative study of a multi-stakeholder innovation network in the context of autonomous driving, we carve out the challenges orchestrators face in these ‘sluggish’ network settings. In addition, our findings unfold three practices orchestrators use to cultivate members’ commitment. Based on these findings, we expand the theoretical understanding of network orchestration by showing practices of orchestration that specifically focus on the creation of commitment among network members that are generally interested but remain passive at first. Furthermore, by assessing how orchestrators try to align members’ time horizons with their own, we unpack the idea of orchestrators as temporal brokers that sequence activities in multi-temporal innovation environments. Overall, we show how orchestrators can foster members' commitment levels to eventually collaborate on innovative projects.


Introduction
In multi-stakeholder networks for innovation, actors from different sectors collaborate to develop innovative solutions to complex problems (Reypens et al., 2021). Such networks are seen as promising to do so as they allow for the pooling of different resources, perspectives, and expertise (Ferraro, Etzion, & Gehman, 2015;Levén et al., 2014). Oftentimes, multi-stakeholder networks for innovation are coordinated by a network orchestrator, a central actor who tries to create value with and mobilise value from multi-stakeholder settings (Dhanaraj & Parkhe, 2006). Network orchestrators try to do so by developing a strategic vision for the network (Aarikka-Stenroos et al., 2017), by providing actors with the necessary resources to collaborate (Ritala et al., 2009) and by creating legitimacy for the network's activities (Paquin & Howard-Grenville, 2013).
One of the tasks for orchestrators in the early stages of network collaboration to create innovative outcomes pertains to securing the commitment of network members (Paquin & Howard-Grenville, 2013). For most orchestrators, fostering commitment might not be CONTACT Leona A. Henry leona.henry@uni-wh.de as problematic either because the innovation setting is dynamic and competitive or because members' commitment is contractually enforced (e.g., Nambisan & Sawhney, 2011). Yet, sometimes orchestrators find themselves in settings in which members collaborate on a voluntary basis and are not overly engaged, while the orchestrators themselves are keen to generate innovate outcomes. In such settings, which we label as 'sluggish', orchestrators need to put substantial time and effort in securing their members' emotional and social commitment before they can engage in collaborative activities with them. In this paper, we seek to understand how orchestrators accomplish this challenging task. We believe this is a highly relevant question because these sluggish networks can hold great potential for innovation which may be wasted unnecessarily without effective facilitation and generation of commitment on the side of the orchestrators. We assess this question in a network we called AutoNet (pseudonym), which is a multi-stakeholder network that searches for technological innovations related to the field of autonomous driving and it is orchestrated by an external management agency responsible for the overall coordination and 'dramaturgy' of the network's activities. AutoNet's members only meet occasionally, they see the network as a 'part-time' activity and most of them have only vague initial expectations as to the outcomes of the network. The orchestrators, on the other hand, aim to develop one or more innovation proposals with this network within a one-year timeframe, which, upon positive evaluation will result in government funding for them. While the orchestrators are thus highly committed, most of the members find the network interesting but remain passive at first. Our findings unfold in detail the collaborative nature of this network setting, the challenges orchestrators face in fostering members' commitment as well as three orchestrating practices the orchestrators engaged in to do so (1) fictional roadmapping, (2) triggering exchanges and (3) tailoring activities and timelines. Although in different manners, these practices allowed the orchestrators to move the network from a largely passive and lowcommitment state to one in which members became sufficiently intrigued for collaboration to take place.
Our study has three important implications for the literature on network orchestration in the context of multi-stakeholder innovation (Levén et al., 2014;Reypens et al., 2021). First, our study highlights specific practices that orchestrators engage in to foster network members' commitment in innovation settings where this is not a given. While studies have emphasised the need for orchestrators to convince members of a network's value in order for them to collaborate (e.g., Paquin & Howard-Grenville, 2013), few studies have closely assessed how this can be done. By specifically focusing on the 'precollaborative' phase in which orchestrators must secure members' motivation to continue, we closely scrutinise this important orchestrating task. Secondly, our findings contribute to the literature on network orchestration by highlighting the relevance of temporalities and 'temporal brokerage' (Reinecke & Ansari, 2015) in the context of multi-stakeholder innovation. While other studies touch upon the salience of temporal dynamics for orchestrators (e.g., Giudici et al., 2018), we still lack an understanding of how orchestrators cope with them to foster commitment and eventually collaboration in these settings. Finally, by showing how orchestrators in sluggish networks act as directors that mobilise network members' engagement through predefined defined scripts, staged interactions and improvisation, our paper unfolds the dramaturgy behind network orchestration (e.g., Khoury et al., 2021). Besides theoretical contributions, our paper also contributes to developing practical insights for orchestrators who find themselves in lowcommitment settings that may be promising for innovation yet slow and challenging to coordinate.

Creating commitment in multi-stakeholder networks for innovation
In multi-stakeholder networks for innovation multiple different actors such as universities, governmental bodies and firms interact over time to develop innovative solutions (Levén et al., 2014;Reypens et al., 2021). Because of their unique capability to bundle perspectives and expertise, such networks are set to achieve 'collaborative advantage', i.e., produce outcomes that no single sector could achieve individually (Huxham & Vangen, 2013;Kogut, 2000). Multi-stakeholder networks are particularly conducive to generating innovative outcomes as they leverage capabilities distributed across a range of organisations (Boland et al., 2007;Coombs & Metcalfe, 2002), allow for sharing knowledge and resources (Westergren & Holmström, 2012) and foster creativity among network members (Aarikka-Stenroos et al., 2017).
Oftentimes, multi-stakeholder networks for innovation are coordinated by a central actor, also referred to as 'network orchestrator', who engages in deliberate and purposeful actions to create and mobilise innovative value from networks (Dhanaraj & Parkhe, 2006;Paquin & Howard-Grenville, 2013;Zaoual & Lecocq, 2018). Orchestrators try to do so by selecting and connecting complementary partners (Zaoual & Lecocq, 2018), by providing spaces for interaction (Paquin & Howard-Grenville, 2013) and by creating trust among network participants (Zucchella & Previtali, 2019). In addition to building the relational infrastructure necessary for producing innovative outcomes, orchestrators also try to mobilise value by facilitating knowledge and resource exchanges (Dhanaraj & Parkhe, 2006;Ritala et al., 2009) and by defining and developing a strategic vision for the network (Aarikka-Stenroos et al., 2017). Finally, in innovation contexts that are tied to local industries, orchestrators play an important role in connecting network members to local public and private actors (Levén et al., 2014). Although some orchestrators also support networks' independent pursuit of new business opportunities (Giudici et al., 2018), usually it is the collective contribution of network members to a common innovation effort that they coordinate.
One of the essential ingredients for generating innovative outcomes with multiple stakeholders is ensuring these stakeholders' commitment to the collaborative process (Hagedoorn et al., 2000;Nummela, 2003). In collaborative contexts, commitment can be seen as an actors' intention to stay in the partnership and actively contribute to make the collaboration a success (Wong et al., 2005). An active contribution here goes beyond mere contractual commitment in the sense that members contribute to the collaborative process 'both in thought and in action' (Sol et al., 2013, p. 42), i.e., they become emotionally and socially attached to the network's activities. Actors' willingness to make such an active contribution has shown to depend on multiple factors, including a belief that benefits of the collaboration will accrue to all parties (Clarke, 2006), abundant similarity between their values and the network's values (Ring & Van de Ven, 1992) and compatibility in actors' goals (Nummela, 2003).
To foster members' commitment, the literature on network orchestration highlights the importance of creating an environment in which procedural justice is ensured, for example, by establishing consistent and inclusive decision-making processes (Dhanaraj & Parkhe, 2006). According to these authors, it is mostly members' fear of free riding and non-equitable distribution of value that may hamper actors' commitment to the collaborative innovation process. In a similar vein, studies have emphasised the value of creating contractual mechanisms to secure members' commitment, such as intellectual property agreements that alleviate actors' concerns about knowledge creation (Leten et al., 2013;Nambisan & Sawhney, 2011). Such mechanisms are especially fruitful in settings that are highly dynamic and competitive of nature, which is the case for many settings in which orchestration has been studied. For instance, Aarikka-Stenroos et al. (2017) emphasise how orchestrators must manage the 'dynamic nature of the range of relevant actors' (p. 102) in the orchestrating process. Going even further, studies have shown how orchestrators in R&D consortia are oftentimes confronted by 'learning races' amongst network members (Hurmelinna-Laukkanen et al., 2012).
While these are fruitful insights, they leave largely unaddressed how orchestrators can foster members' commitment in multi-stakeholder settings in which this cannot be contractually enforced, and which are not as dynamic in nature. In such settings, which we label as 'sluggish', orchestrators need to put substantial time and effort in securing members' commitment before they can engage in collaborative activities with their network members. In the following, we describe the nature of these networks and the commitment challenge this causes for orchestrators in more detail.

Sluggish network settings
Whereas many innovation networks are highly dynamic settings, in some innovation networks, members participate on a voluntary and loosely coupled basis, have infrequent meetings and participation is not contractually enforced. In such networks, members are generally interested and see potential, but oftentimes the network's activities are not directly related to their daily business and there is little sense of urgency. The orchestrators of such networks on the other hand, can be extremely committed and engaged for reasons that current literature has not emphasised so far. For example, they might operate in a sponsored programme which offers funding for orchestration activities in innovation networks. In order to secure this funding and any follow-up funding, orchestrators are under pressure to generate tangible outcomes from the network members' collaboration. Secondly, the orchestrators may be professionally committed and working for external management agencies or consultancies who offer the service of coordinating innovation networks. Thus, while the network may constitute a part-time, optional activity for network members, it may be the full-time, core business for network orchestrators. All in all, many innovation networks may be characterised by an asymmetric relationship between the orchestrators' commitment on the one hand and network members' commitment on the other. We label this kind of a collaborative setting as 'sluggish' and we claim that this sluggishness and the risk of wasted innovation potential could pose an important problem for business and society.
We do not claim that sluggish networks are solely found in the context of multistakeholder innovation networks. For example, they bear resemblance to 'interstitial spaces' (Furnari, 2014) in which actors from different fields interact occasionally and informally around common activities to which they devote limited time, such as hobbyist clubs, hangouts, workshops, meet-ups. In addition, due to their part-time nature sluggish networks are also similar to cross-sector partnerships in which actors temporarily collaborate around sustainability issues (e.g., Henry et al., 2022). Even though these settings and their 'sluggishness' might not seem promising in the context of innovation, they have been argued to be highly conducive to doing so. First of all, informal collaborative settings are said to foster innovative efforts as they temporarily free up actors' minds and help them let go of their dominant field structures (Villani & Phillips, 2021). In addition, voluntary forms of organising are increasingly common in the context of grand challenges to foster the participation of multiple and dispersed stakeholders that might otherwise not engage (Delmas & Terlaak, 2001).
Yet, opposed to multi-stakeholder networks for innovation, the above-mentioned settings are usually not centrally coordinated by an orchestrator who, for varying reasons, is highly committed to generate innovative outcomes with the network. While they might thus be conducive to generating innovative outcomes eventually, orchestrators in these sluggish networks are likely to spend substantial time and energy to motivate and incentivise members before they engage in actual collaboration. However, given that our knowledge on orchestration and the creation of commitment mostly comes from more dynamic and contractually enforced innovation contexts, we lack an understanding of how orchestrators foster members' commitment when this is not a given. As we believe that sluggish networks can nevertheless constitute promising environments for innovation, we ask: How do network orchestrators foster commitment in multi-stakeholder networks for innovation that are sluggish in nature?

Case setting
Our study is set up as a single case study, suited to investigate 'how' and 'why' questions in relation to empirical phenomena that have not been researched extensively before (Yin, 2018). The network 'AutoNet' (pseudonym) has the overall mission to develop innovative applications for autonomous vehicles. AutoNet operates in a funding programme initiated by the German Ministry of Economic Affairs with the overall aim of fostering innovative technology projects between SMEs, larger firms and research institutes around complex societal issues. As one of the goals of this programme is to foster SME development, the programme requires that these networks consist of a minimum of at least six SMEs to be eligible for funding. Besides these SMEs, the network may include an undefined number of other partners, such as research agencies, large firms (more than 500 employees), associations or non-profit firms. In terms of time frame, the network operates in a one-year 'exploration phase' (year 1) after which they have to hand in their innovation proposal(s). If those final innovation proposals are evaluated positively, the orchestrating agency receives governmental funding that they can use to coordinate the network for another two years and turn their proposals into action. This is important, as only through this funding, the network will be able to turn their proposals into concrete innovation outcomes. AutoNet is a fruitful case to study efforts of creating commitment as our research was conducted in the first year of the network, in which the orchestrators needed to formulate one or more innovation proposals with the network members. After the first year, the network entered the second phase in which it still found itself at the end of our observation period. Being able to follow this network from the first day allowed us to see how the process of fostering commitment evolved from the network's nascent stages.
During the period of observation, AutoNet consisted of six SME's, four research institutes and seven larger firms that operate in the automotive industry or have a technical expertise that is deemed promising to develop innovations for this sector. Among the research institutes was a mixture of universities, applied research centres and smaller research labs that work on technologies close to, but not necessarily directly in the automotive industry. Among the SMEs were predominantly suppliers who work on technologies such as 3D printing or relevant new materials. Finally, the larger members mainly included automotive manufacturing firms. Hence, in terms of sector composition, AutoNet found itself at the interface of the automotive and other sectors, such as 3D printing, software or plasma technology. Except for some of the SMEs, most of the members did not know each other prior to operating in AutoNet. At the end of our observation period, the network started expanding and the number of network members was fluctuating.
The network orchestrating agency in our case is an external network management agency that purposefully assembles networks around technological fields. This agency itself is not a member of the network, but a third-party mediator that is responsible for bringing actors together, organising and hosting network meetings, and supporting the network members in writing their innovation proposals. Importantly, the orchestrators themselves are no technological experts in the automotive field, i.e., they are purely responsible for the network's overall 'dramaturgy', not the actual content of the members' exchanges. The agency is composed of five actors who work as network orchestrators, of whom three have been responsible for orchestrating AutoNet. However, as all five of them are experienced in orchestrating networks generally, we also included the other two orchestrators in our research procedure to increase our understanding of their activities.
AutoNet's orchestrators held four network meetings in the first year which would take half a day on average. An exception was the third meeting, which took two days in total. While most of AutoNet's members joined these meetings, they devoted little time to any additional activities, as to all of them, participating in the network was a part-time activity that was not necessarily related to their core business. Hence, the amount of resources devoted to any follow-up activities that would help the orchestrators gather more information about members' interests was limited. This also resulted in the fact that between the network meetings, there was little to no communication among the network members. Finally, network meetings were informal in the sense that rules for participation were not defined, and members were not obliged to join or actively participate in all the meetings. All in all, AutoNet thus provides a suitable setting to study how orchestrators foster commitment in a sluggish setting in which the stakes are high for the orchestrators, but in which members show little commitment to the network activities.

Data sources
Our goal was to capture how orchestrators in low-commitment multi-stakeholder networks try to generate collective innovation projects from such settings. We collected data longitudinally with a real-time data collection period from November 2017 to January 2019. In terms of data sources, we draw on interviews, documentary evidence and observations, which allowed for converging lines of inquiry and the corroboration and triangulation of our data (Yin, 2018).
Interviews: Our first source of data consists of semi-structured interviews with AutoNet's orchestrators and the network members. As mentioned, AutoNet was managed by three orchestrators, of which one was the main orchestrator for AutoNet and the other two operated in the background. However, the network management agency employs two other network orchestrators that we also interviewed as these are experienced in orchestrating networks generally. In the first round of interviews, we interviewed all five orchestrators. The three orchestrators mainly responsible for AutoNet were interviewed a second time at a later stage for purposes of triangulation.
We also interviewed the network's members, for which we maximised the number of actors from different sectors to increase validity of our data. Although some of the network members were not willing to be interviewed, we covered the key informants in the case as judged by the orchestrators: three out of six SMEs, three out of seven large firms and two out of four research institutes. Even though we did not interview everyone, in our interviews certain practices were repeatedly mentioned up to a point where no new ones emerged, suggesting data saturation (Guest et al., 2006). An overview of our interviews can be found in Table 1. Moreover, as we elaborate below, we used a substantial amount of background documents to retrieve as much information as possible about the network and its members (Tracy, 2010). All our interviews are semi-structured enabling informants to report their own experiences and interpretations of events (Flick, 2018) with an interview length between 30 to 60 minutes. Interviews were transcribed and fed back for additional comments or explanations where necessary.
Non-participant observations: A second type of data was gathered through nonparticipant observation during the four network meetings organised for the network, which each lasted between three and four hours. The first author attended the first two meetings, the second author attended the fourth meeting. For the third meeting that we were not able to attend, we collected detailed information from the network orchestrators to verify attendees and relevant information. Secondly, the first author attended two meetings held for the orchestrators only, during which they discussed their activities and future course of action. During all these events, we took extensive field notes, which we used to understand and identify key moments in the orchestrating process and how they influenced the network's evolution. Documents: Finally, we analysed several documents including general project information (9), meeting protocols (6), Powerpoint presentations developed by orchestrators (4) and Powerpoint presentations developed by network members (6). The general project information included documents about the network management agency and the project steps, which were helpful in developing an understanding about the case's context. In a similar vein, the presentations developed by the network members, in which they presented their organisations to their fellow network members, were helpful in understanding these members' profiles, expertise and expectations. Finally, the presentations that the orchestrators prepared for their network meetings helped us make sense of the orchestrating process and the network's evolution, as their content mirrored the network's development.

Data analysis
To understand how orchestrators sought to foster collaboration in a sluggish network setting, we used an inductive approach which is inspired by grounded theory (Strauss & Corbin, 1998) and which is particularly relevant to capture informants' lived experiences. Our analytical process was iterative, as we moved through multiple rounds of coding (Gioia et al., 2013). In a first round of coding, we engaged in open coding to understand the collaborative nature of AutoNet and what this meant for the orchestrators. Being immersed in the case had made clear to us that we were observing a specific network environment that caused particular orchestrating challenges and practices. Hence, the first round of coding focused on data passages that elaborated the network's collaborative nature, the challenges this caused for the orchestrators as well as those passages that showed how they responded to them. From these data passages we created a set of first order codes that were close to the raw data and mostly in vivo.
In the second step, we went through these data passages to determine whether and to what extent they follow certain patterns. In this round of coding, we ventured back and forth between the literature on network orchestration (e.g., Dhanaraj & Parkhe, 2006) to understand what made our case special. In doing so, we discovered that the specific character resulted from an asymmetry between the commitment of the orchestrators and the commitment of the network members. Although it is not a theoretical concept as such, we started to think about this specific environment as 'sluggish', which we believe adequately captures the particular collaborative dynamic in AutoNet. In addition, we came to realise that the sluggish nature caused a commitment challenge for orchestrators as they needed to put substantial time and energy into motivating members to see the network's value. In addition, we noticed how the orchestrators' actions to cope with these low-commitment levels revolved around three main practices. These included the development of a road map, the fact that they triggered interactions by inviting external guest or organised speed dating sessions, and the purposeful deviations from their initial orchestration plan.
In the final step, we started to thicken our analysis around these three practices of orchestration and make sense of them in greater detail. For example, we started to compare the orchestrators' development of a road map to other studies on roadmapping (Möllering & Müller-Seitz, 2018;Phaal et al., 2004). In doing so, we noticed that the roadmapping process in our data differed from extant studies, in the sense that the first version of the road map developed by AutoNet's orchestrators was purely fictional and served mainly as a springboard to develop an actual road map. This insight led us to label this practice 'fictional roadmapping'. In a similar vein, our second order codes showed how orchestrators invited a set of external guests to the network meetings and organised speed dating sessions, which at first glance seemed unrelated to commitment building. Upon closer analysis however, we found that the guests could boost commitment by getting network members to 'think outside the box' and subtly remind them of the importance of the network's topic. In a similar vein, the speed dating sessions served to foster commitment by showing members interesting partners they might end up working with in an environment highly conducive to knowledge sharing. This practice we labelled 'triggering exchanges'. Finally, we noticed how orchestrators started to deviate from their orchestrating programme to make sure that the few engaged members remained engaged, the practice we labelled 'tailoring activities and timelines'. Taken together, these practices of orchestrating sluggish networks emerged as the most important building blocks of our analysis that underpinned our further theorising. To validate our analyses, we presented our findings to AutoNet members and orchestrators in the final stage of the research process and used their feedback to ensure the plausibility of our findings (Tracy, 2010).

The sluggish nature of AutoNet
In AutoNet's early days, the orchestrators managed to gather various SMEs, research institutes and larger firms around the topic of 'innovative applications for the future of the automotive industry'. While the process of getting actors interested and assembling them around this highly promising theme was rather unproblematic, the challenge for the orchestrators started once they wanted to develop concrete innovation proposals with the network members. While AutoNet's members were interested in the future of automotive generally, they did not necessarily expect an innovation proposal to result from the first year of networking, nor were they extremely eager to develop one. As one of them mentioned: 'For us it would not be necessary to develop a concrete project. Getting to know the latest trends and actors in industry would already be valuable' (Representative large firm). Several of the members indicated how they used this first year to explore what value it would bring them to join, if any at all. 'Let's wait and see what comes from this' (Representative SME). At the same time, even though they had only few expectations, many of them knew that if something big would result from the network's activities, they would benefit greatly from being part of it. As one of the academic members explained: 'As academic partners we have little to lose here. The investment is small, but it might be a great opportunity. It would be stupid not to participate' (Representative research institute).
Consequently, most network members were interested and joined the network meetings, but they did not put any additional effort in the networks' activities outside of these. One of the orchestrators, who was responsible for communicating with network members mentioned: 'Some of our participants like to be informed, but usually, most of them don't answer my emails. This is simply not their most important job' (Orchestrator 2). For the orchestrators on the other hand, the stakes of participation were much higher: they operated in a government sponsored programme, which implied that they had exactly one year to develop one or more innovation proposals with the network. If evaluated positively, these proposals would grant the orchestrating agency funding which they could use to coordinate the network for another two years. As such, the fact that network members were not overly engaged from the start put the orchestrators in a precarious position. As one of them elaborated: These programmes have tight deadlines you know, we can't just wait for things to happen because there are only certain periods where we can apply. Before we started working with AutoNet we (orchestrators) had already made a time horizon for ourselves that shows when we must do what to apply for the funding (Orchestrator 1).
In addition to network members' hesitant attitude, the orchestrators themselves were no experts in the automotive field, which reinforced the challenge for them to secure members' commitment as they were unsure how to motivate them. As one of them described: 'We are definitely not the technical experts, and the future of this industry is very uncertain' (Orchestrator 2). While the orchestrators were clearly challenged by this low-commitment nature, they also realised that they had limited power to influence it because while members paid a small fee to be part of the network, participation was not contractually enforced. As one of the orchestrators explained: The only thing we can do is accept it. Because all of them, but especially the large firms, if they just want to come to the meetings and listen, there is little for us to do about it. They paid their membership fees and with that we can in turn finance some things with that. And we must accept it, because if we force them, they would be dissatisfied and probably leave. And for the moment they are happy to just be there, listen and talk to some of the other members (Orchestrator 1).
Despite this challenging collaborative environment, the orchestrator's aim was to foster a degree of commitment among network members that would allow them to develop one or more innovation proposals in the first year of collaborating. In the following, we describe the practices AutoNet's orchestrators engaged in to realise this.

Practices of orchestrating sluggish networks
Fictional roadmapping A first way in which AutoNet's orchestrators tried to trigger commitment was by engaging in a practice that we call 'fictional roadmapping'. Fictional roadmapping implies that the orchestrators developed a road map for the network members, i.e., a plan outlining technological subthemes that network members could focus on for their project proposals as well as a timeline showing when these should be finished. Yet, as the orchestrators were no technological experts in the field of Automotive, they were not able to define these topics right away. At the same time, they realised that if they would not present the members with a direction in the early stages of the network, they would not be able to motivate them and secure their commitment. To solve this problem, the orchestrators decided to start to proactively visit all the network members to discuss their interests, competencies and the expectations, if any. Based on these meetings and outcomes they formulated six very abstract potential future directions for the network, which simply reflected broad and abstract concepts like 'light' or 'materials'. Even though the exact meaning and feasibility of these directions were not clear yet to the orchestrators, they presented them during the first network meeting as the final version of the road map, labelled 'AutoNet's key technological directions' (PowerPoint presentation 1). Subsequently, the orchestrators divided the network into three subgroups to engage in preliminary brainstorming about the potential of these technological directions.
While the orchestrators had thus presented the road map as if it were finalised already, they merely did so to gather input and feedback during this first meeting to understand which topics would motivate network members in going forward. Thus, visualising potential (albeit hypothetical) outcomes of the network triggered a discussion among the network members, which the orchestrators used to identify the promising themes, i.e., 'hot topics', which helped them understand how they could increase members' commitment to the network's activities. As one of the orchestrators put it: 'When we present a technological direction and no one asks a question about it, we know enough already' (Orchestrator 4). At the very end of this first roadmapping session, the orchestrators were able to narrow down the set of four potential technological directions and formed four project groups around them which they used in going forward. Although the orchestrators communicated that network members could still switch between groups if desired, the orchestrators' intention was clearly to give them an established status so they could proceed with these groups during the next meeting. An excerpt from an email sent by one of the orchestrators to his colleagues shows this notion: 'The groups are seen as established; we should communicate them clearly from now on!' Members themselves also appreciated the roadmapping exercise and the direction it provided. As one of them indicated after this first session: 'I didn't know what to expect from this first meeting really, but I think they [orchestrators] did a good job at narrowing things down for us. I do have a better idea now of where we're going with this, and that's a good thing'. (Representative SME).
In sum, by developing and presenting a fictional road map, the orchestrators were able to probe and understand with which topics they would be able to motivate network members and strengthen their commitment to the network's activities.

Triggering exchanges
A second way for the orchestrators to foster commitment in AutoNet was through a practice that we call 'triggering exchanges'. Triggering exchanges implied that they used different ways of stimulating interaction between network members with the aim to convince them of the value of AutoNet and the actors that were present in it. The orchestrators realised that if they wanted to bolster commitment among members in AutoNet, they needed to find a way to make the network meetings less passive in nature and to get members to interact. To do so, the orchestrators invited external speakers from different (academic) disciplines and organisational sectors to the network meetings. At first glance, these guest speakers had little or nothing to do with the field of automotive mobility, yet they were purposefully chosen by the orchestrators to discuss this topic from their field or perspective. For example, during one of the meetings the orchestrators invited a psychologist, who delivered a keynote on the perception and experience of autonomous driving among various age groups. In a similar vein, the orchestrators invited a sociologist who presented socio-demographic factors relevant in the broader context of innovation.
By inviting these guest speakers, the orchestrators tried to provide members with a different perspective that would help them to 'think outside the box' and consequently stimulate a discussion. As one of them explained: 'We try to constantly give our members a different perspective, let them see some things that they probably have heard of, but never actively thought about' (Orchestrator 2). Network members indicated how the presentations held by external speakers helped them shift perspective and look at the network differently: 'It was very exciting for me to hear a psychologist talk about my daily work. We tend to think in such technical terms, so understanding the psychological side of it definitely gives me food for thought' (Representative large firm). While these external presentations were thus predominantly meant to stimulate a discussion among network members, they also served to underscore the importance of the overall topic of autonomous driving, which the orchestrators hoped would also foster commitment among network members. For example, the sociologist presented autonomous driving as one of the most 'significant demographic trends' of our time (PowerPoint presentation 3), something which the orchestrators had actively asked for.
A second way in which the orchestrators tried to trigger exchanges was by organising speed dating sessions for the members. Like romantic speed-dating, they organised short sessions in which network members were able to chat bilaterally and informally for a short time. To see whom to match during the sessions, the orchestrators sent around an Excel file where each network member could tick six potential partners they wanted to meet. Rather than fostering interaction among members collectively, the orchestrators hoped that the speed-dating sessions and the ability to interact bilaterally would help network members realise the potential of the partners present in the network. Initially, some of the network members were rather sceptical about this idea as it meant they had to become active and start interacting with one another. As one of AutoNet's members explained: At first, I was really critical about the speed-dating thing, but then I really appreciated how it enables me to quickly talk about two or three things only and immediately get a sense of whether the person sitting across from me also likes these ideas. (Representative research institute) One of the SME representatives also indicated: 'I would never share my ideas in the big group, but when it's just the two of us it gets much better'. Hence, the speed dating proved a fruitful way to foster interaction between network members who might not have been willing to interact in a larger group setting. For the orchestrators, the speed dating sessions again helped understand the network better as it allowed them to probe 'who hit if off' (Orchestrator 1) and detect matches between members that they would not have come up with. In conclusion, a second way through which AutoNet's orchestrators tried to foster commitment was through triggering exchanges, which they did by inviting surprising guest speakers as well as organising short speed dating sessions.

Tailoring activities and timelines
A final way for the orchestrators to encourage AutoNet's members to become more committed was by customising their orchestration timelines to members' individual engagement levels. Although the general mode in AutoNet was still rather passive, after a while the orchestrators started to pick up that the SMEs were overall more engaged than the other members and were keen to start developing innovation proposals. As one of the orchestrators explained: 'We now see that we have a few champions who are starting to push this network. Right now, these are the SMEs for whom a potential innovation is most critical' (Orchestrator 3). While the orchestrators had initially developed a schedule for their activities which foresaw a fully collective orchestrating process, they realised how they could not stick to this plan if they wanted to safeguard the increasing engagement of these few critical members. At the same time, the voluntary nature of the network also made it impossible to disregard the other members. What is more, a valid innovation proposal needed the input from all three member groups, not just the SMEs. As shown by the following quote, these conditions created a severe temporal challenge for the orchestrators: 'We now clearly have different velocities that we are orchestrating here, but we can't really throw anyone out and go ahead with the SMEs only' (Orchestrator 1).
To address this temporal challenge, orchestrators decided to adapt their orchestration timelines in such a manner that the more active members and the more passive members were both 'served'. This implied that for the more active members, orchestrators would start developing certain activities faster than for the passive ones. For example, one of the orchestrating activities entailed the development of project ideas parallel to the network members, which served to provide network members with some additional food for thought. While orchestrators had planned to present these additional proposals in the final phase of the orchestrating process, they decided to present them to the more active members somewhat earlier in the hope that it would reinforce these first signals of commitment. As one of the orchestrators explained: Originally, we had planned to wait with our own project proposals until September, but we realized that we had to present them earlier to keep some of our faster members interested. It is not ideal, but right now, we have different speeds in the network that we need to accommodate. (Orchestrator 3) In a similar vein, the orchestrators started to treat the more active members somewhat differently from the passive members in the sense that they were favoured during networking activities. For example, the orchestrators were planning a second speed dating session for the network members. To see whom to match during the sessions, the orchestrators again sent around an Excel file where each network member could tick six potential partners they wanted to meet. Yet, in finalising this schedule, they made sure that the more active network members were matched to their preferred partners, while the others were only matched to all their desired partners if possible. Naturally, they did not neglect the other network members, but at the same time they realised that if the SMEs would lose their engagement, motivating the other members would become more challenging, too. As one of the orchestrators put it: 'We have to make sure that the SMEs get their wishes granted; we can't let them down at this point' (Orchestrator 4). Thus, by tailoring their timelines and activities to members' individual commitment levels, orchestrators tried to sustain and further increase those.
All in all, AutoNet's orchestrators thus engaged in three practices to generate commitment among their network members. First, the orchestrators presented a fictional road map which allowed them to understand what topics would motivate network members and which they could use for the development of the innovation proposals. Second, they actively triggered exchanges between network members to help them realise the value of the network generally and the different actors that were part of it. And finally, they customised their orchestration timelines in such a way that the increasing commitment of a few members was safeguarded also in the hope that these would be able to motivate less active network members. We summarise these practices in Table 2. After the end of the first year, AutoNet's orchestrators managed to finalise and submit two innovation proposals for funding. The network moved on to the second phase in which the orchestrators and members collaborate to develop these proposals into tangible innovation outcomes.

Discussion
In this study we set out to explore how network orchestrators try to foster collaboration in multi-stakeholder networks that are sluggish in nature, i.e., in which orchestrators are highly committed, but in which members are rather passive initially. In the following, we discuss the theoretical implications of our findings for the literature on orchestrating multi-stakeholder innovation.

Orchestrating to foster commitment in sluggish network settings
A first implication offered by our study is a more refined understanding of how network orchestrators foster members' commitment in settings where this is not a given. While studies on network orchestration have highlighted the need for orchestrators to convince members of the value of the network and motivate them (Levén et al., 2014;Paquin & Howard-Grenville, 2013) empirical insights on how orchestrators can do this to date remain scarce. Our findings reveal a set of specific practices that orchestrators can use to foster members' commitment in sluggish settings. In AutoNet, the orchestrators first tried to motivate and incentivise members through a practice we call 'fictional • Orchestrators invite speakers from different fields trigger discussions among members and emphasise the societal relevance of the network.
• Orchestrators organise short speed dating sessions to foster bilateral interactions.
• Orchestrators adapt their orchestration timelines to members' individual engagement levels.
• Orchestrators customise the network's activities to correspond to members' individual engagement levels.
roadmapping', which implied that they prepared and presented a fictional set of technological topics for the network to work on. Roadmapping is a technique used in collaborative settings that helps actors identify relationships between existing and developing technologies, products, and markets (Phaal et al., 2004). While such maps generally have a fictional character, they can be very helpful to provide direction in times of uncertainty (Möllering & Müller-Seitz, 2018). The aim of AutoNet's orchestrators, however, was not to provide members a direction as to reduce uncertainty, but rather to get an idea what collaborative topics might lead them to become more committed and develop innovation proposals. Another way for the orchestrators to incentivise was by 'triggering exchanges', i.e., activate members' exchanges by inviting unexpected speakers, such as psychologists, to talk about the topic of autonomous driving from their perspective and by organising short speed-dating sessions for members. Finally, orchestrators started to customise their orchestrating activities and timelines to sustain the engagement of a few motivated actors which they hoped could serve as champions for the less committed members.
Prior studies that focus specifically on the question how commitment can be created in innovation networks mostly focus on safeguarding knowledge appropriation, for example, by setting up formal agreements (Leten et al., 2013;Nambisan & Sawhney, 2011). Studies from other networks settings, such as strategic alliances, have also emphasised the value of implementing contractual safeguards (Perry et al., 2004). In supply chain settings, studies have also highlighted the importance of effective and transparent governance infrastructures to foster commitment among partners (Fawcett et al., 2006). Opposed to these studies, which mostly emphasise the value of pre-defined agreements or structural factors as precursors for commitment, our study draws attention to the fact that commitment can also be fostered through subtle practices that increase members' commitment by convincing them of the network's value and binding them emotionally. As multi-stakeholder collaboration with sluggish characteristics is common in the context of grand challenges and innovation to foster the participation of multiple and dispersed stakeholders (Ferraro et al., 2015), we believe that fostering commitment represents an increasingly important task for orchestrators and our findings thus have both theoretical as well as practical relevance.

Network orchestrators as temporal brokers
A second implication offered by our findings is the refinement of the idea of network orchestrators as 'temporal brokers' (Reinecke & Ansari, 2015), who sequence and align the different temporalities inherent in innovation networks. While studies on orchestrating have started to pick up the notion of temporal dynamics inherent to orchestration (e.g., Giudici et al., 2018), a fine-grained understanding of how network orchestrators cope with competing temporal orientations is still absent. Temporal orientations, i.e., the different 'rhythms' actors use to structure and pace their work (Ancona & Chong, 1996), have shown how the presence of these different orientations can become a source of conflict for organisations and their members. For example,  showed how different temporal horizons between managers and scientists in complex innovation project resulted in severe tensions between these two actor groups. In a similar vein, Reinecke and Ansari (2015) showed how conflicts between market and development temporalities within Fairtrade International led to organising tensions that hampered collaboration between actors from the North and the South. Other studies have shown similar temporal conflicts arising between top and middle managers (McGivern et al., 2018) or managers and venture capitalists (Gersick, 1994).
In our case, these competing temporalities manifested between the orchestrators on the one hand, and the network members on the other. While the orchestrators were operating under time pressure due to the tight deadlines set by the funding programme, for most of the network members these deadlines were not as critical which created a source of temporal conflict for the orchestrators. Prior studies provide some insight into how actors can align temporal conflicts in such a manner that collaboration evolves despite their presence. For example, Reinecke and Ansari (2015) have shown how actors can do so through a continuous process of questioning, articulating, and reconsidering those differences. In addition, these authors emphasise the benefits of highlighting positive interdependencies between actors' ultimate goals in achieving temporal alignment. In their study on different pacing preferences in complex innovations,  highlight how actors should iterate between different pacing styles depending on the innovation trajectory they find themselves in. In addition, these authors draw attention to the importance of establishing a set of learning outcomes as performance metrics rather than relying on temporal performance metrics only. Finally, studies in the domain of strategy making have shown actors overcome temporal conflicts by settling on accounts that link interpretations of the past, present and future in ways that appear coherent and plausible (Kaplan & Orlikowski, 2013).
In our study, however, the orchestrators tried to align network members' temporalities with their own in two unusual ways: First of all, orchestrators tried to fastforward members' interactions, i.e., to make these proceed at a more rapid speed than they would otherwise occur in. To do so, they first of all provided network members with a hypothetical timeline which was part of their fictional roadmapping strategy. Even though the orchestrators were not certain of the accuracy of this timeline, they used it to create network members' awareness of the fact that the clock was ticking, and their timely input was required. In addition, the orchestrators made it a habit to start their network meeting by showing the network's timeline and spending a few minutes on it. A second way for orchestrators to influence network members' temporal orientations was by bending their own timelines in such a way that it triggered members' engagement. For example, the orchestrators either postponed or moved forward certain orchestrating activities in the hope that this would increase network members' commitment to the collaborative process.
In sum, we observed how network orchestrators tried to align competing temporal orientations of themselves and their network members by trying to fast-forward members' interactions as well as by adapting their own timelines. Contrary to extant research that shows how actors align conflicting temporalities by finding common ground between different orientations or by foregrounding a shared meta goal Reinecke & Ansari, 2015), our findings reveal a different mode of alignment. In AutoNet, rather than articulating these different orientations or finding a common denominator, orchestrators tried to subtly manipulate their members' timelines with the aim of bolstering commitment levels. They did so by trying to reinforce the notion that time was limited as well as by bending their own timelines in such a way that would support network members to become active. As such, our findings provide a fresh perspective on the enactment of temporal brokerage in an orchestration setting, in which temporal conflicts are not constituted by members' diverging stakeholder demands (e.g., Henry et al., 2022), but by competing (temporal) demands between network members and their orchestrators. Sensitising for different options of temporal brokering in this way increases the chances of preventing innovation networks from fading out without tangible outcomes.

A dramaturgical perspective on network orchestration
A final contribution offered by this study is the provision of a dramaturgical perspective on network orchestration. A dramaturgy lens in organisation studies largely draws upon the early work by Goffman (1959) and highlights the staged and performed nature of management and organisational life in general. It builds on the idea that organisations run on predefined scripts and performances (Clark & Mangham, 2004) and emphasises how interactions between actors are 'constructed, sustained, and managed' (Schreyögg & Höpfl, 2004, p. 693) to mobilise audiences and deliver organisational outcomes (Gardner & Avolio, 1998). Studies adopting such a perspective have shown for example, how management gurus draw on constructed narratives to shape managerial identities and consequently foster organisational change (Clark & Salaman, 1998). In a similar vein, Khoury et al. (2021) showed how grass-roots organisations engage in practices of strategic role-playing practices to mobilise important audiences with the aim to advance their overall credibility.
In the case of AutoNet, the dramaturgical character of orchestration was particularly visible in the sense that orchestrators acted as directors with the aim to motivate and incentivise network towards becoming committed to the network's activities. To do so, they strategically pre-defined their activities and staged network members' interactions to maximise their perception of the network's value and their engagement to its activities. For example, the chronological order of the network meetings and the specific activities that would take place during these meetings was carefully planned by the orchestrators in order to steadily increase members' commitment levels. By first presenting the network with interesting topics (i.e., the road map) followed by interesting partners (i.e., speed dating sessions) and external speakers, orchestrators planned to build members' commitment step by step and in the most logical manner. Secondly, the orchestrators strategically staged the exchanges between network members and external speakers. First of all, these speakers and their seemingly 'off-topic' expertise were chosen in such a manner that they had a particularly surprising effect on network members and would help them trigger a discussion among network members that would enthuse them to become active. Secondly, the orchestrators made sure that the speakers used particular narratives to evoke the impression of the networks' activities as having great societal relevance. Regardless of whether orchestrators knew this to be true themselves, they deliberately ensured that external speakers highlighted the link between AutoNet's activities and societal developments such as climate change. Hence, the orchestrators carefully and strategically directed their members interactions and exchanges with the aim to increase their commitment to the network's activities.
At the same time though, our findings also show how orchestrators' dramaturgical efforts could only be shaped to a certain extent, and that sometimes orchestrators had to deviate from their initial script in order to react to ad-hoc developments in the network. For example, when they noticed how some members became more active and committed than others, they adjusted their overall orchestrating timelines and brought forward certain activities to ensure these more active members would not lose their commitment. The most prominent example of this pertains to the speed-dating sessions: While all network members were asked whom they wanted to meet during those session, the orchestrators allowed themselves to flexibly match partners according to their engagement levels. Especially in the second speed-dating session, orchestrators deliberately planned which network members would be matched to safeguard the growing commitment of the more active members. Hence, while in many settings orchestrators might largely rely on pre-defined scripts (e.g., Khoury et al., 2021) the findings of our study sensitise us to the fact that sluggish networks require a more flexible dramaturgy that also relies on improvisation and adaptation. Although more research is necessary to fully grasp the phenomenon of dramaturgy in orchestrated innovation settings, we believe our study provides a fruitful springboard to discuss the overall relevance of a dramaturgical lens for innovation contexts. Further scrutinising the dramaturgy behind orchestration seems not only helpful to enhance our knowledge of orchestrating as such, but also to understand why some orchestrators reach certain outcomes with their network members (e.g., create innovative solutions) while others fail to do so.

Limitations and outlook
As any empirical study, our work also has its limitations: First of all, although AutoNet is an exemplary case of a sluggish innovation network, it is a single case study. While this design allows for developing a fine-grained understanding of empirical phenomena, it may potentially limit the generalisability of our findings (Eisenhardt & Graebner, 2007;Siggelkow, 2007). Nevertheless, we believe our findings do have relevance for similar innovation settings in which orchestrators are keen themselves to develop innovative outcomes but have little formal power to realise this. Studies on orchestration have increasingly started to emphasise how orchestration is becoming less dominant (Reypens et al., 2021) and, instead, performed on an ad hoc basis (Giudici et al., 2018). As such, we believe the relevance of the network setting we study in this paper as well as the practices of creating commitment have relevance beyond the current context. Furthermore, while we interviewed all actor groups in AutoNet, we still have a relatively small number of interviewees. Note though that almost half of the network members and all orchestrators have been interviewed, giving a high degree of saturation for the AutoNet case. Moreover, through other data sources we substantiated and triangulated our interviews as much as possible. Despite these limitations, our findings should be of interest to orchestrators in similar settings who want to extract and create value for their network members. We encourage scholars to conduct further research on similar network settings to validate our findings.
Finally, we believe an important avenue for future studies lies in assessing temporality and the broader notion of 'temporal work' (Kaplan & Orlikowski, 2013) in other orchestration settings. While our case provides some interesting insights on the question of different temporal orientations in sluggish multi-stakeholder networks, future research should focus on discovering the influence of different sources of temporal conflicts in networks with other dynamics, too. Especially for dominating orchestrators who coordinate networks in a rather formal and top-down manner (Reypens et al., 2021), temporal orientations are likely to become highly relevant and may call for a particular set of orchestrating practices. A promising future methodological avenue to address this aspect lies in conducting network ethnography (Berthod et al., 2017), which includes the use of both quantitative (social network analysis) and qualitative data, collected and analysed both in a separate and parallel fashion. Applying such a methodological approach to sluggish innovation networks would allow capturing practices of network orchestration and of creating commitment in even more richness than relying on qualitative data only. In addition, it would allow to track commitment levels, e.g., through assessing actors' tie structures (Gilsing et al., 2008) and as such give an additional measure of how commitment levels develop over time and their contribution to innovative efforts.

Conclusion
Orchestrators are increasingly involved to coordinate networks and foster the development of tangible outcomes. Yet, while most of these networked settings pertain to highpowered networks in which members' commitment is contractually enforced, this is not always the case. Sometimes, orchestrators find themselves in a setting we refer to as a sluggish: a collaborative environment in which members devote little time and effort, but in which the orchestrators are highly committed. Our study shows the challenges this environment results in and illustrates how orchestrators try to navigate this setting through three different orchestration practices. Our main message is that sluggish innovation settings might result in innovative outcomes over time, but that it involves specific challenges and requires a set of different orchestrating practices to realise this and prevent the waste of innovation potential.

Disclosure statement
No potential conflict of interest was reported by the authors.