AI-Mediated Participation and People Sustainability: A Socio-Technical Case Study in Healthcare Shift Scheduling

Abstract

Artificial intelligence (AI) is increasingly reshaping organizational dynamics, not only through efficiency gains but by influencing how work is structured, interpreted, and experienced. In healthcare, where professional team stability is crucial, this transformation intersects with structural issues such as persistent nurse turnover. This study presents an exploratory case study of a private accredited hospital in Italy that introduced an AI-enabled shift scheduling system (“Dream-Shift”) in response to perceived inequities and workforce instability. The system was embedded in a participatory architecture that included a Nursing Practice Council and HR dashboards to visualize staffing patterns. Drawing on theories of Sustainable Human Resource Management (SHRM), algorithmic management, and people sustainability, the study examines how AI-mediated transparency and participation affect fairness perceptions, predictability, and organizational climate. Using administrative data, ethnographic observations, internal documents, and informal feedback, the study finds that the algorithm did not eliminate all inequities but made decision constraints visible and debatable. It redistributed the emotional burden of scheduling and enabled more structured conversations about work. Managers transitioned from unilateral decision-makers to facilitators of collective interpretation. The results suggest that when integrated into participatory infrastructures, AI can foster organizational transparency, support relational stability, and act as a socio-technical enabler of people sustainability rather than as a tool of control.

1. Introduction

Artificial intelligence (AI) is entering organizational life with a speed and depth that few other technologies have exhibited in recent decades. Yet, despite the growing sophistication of predictive models and scheduling algorithms, the central tension for contemporary organizations does not lie in the computational capacity of AI, but in the way it intersects with the social fabric of work—particularly with participation, fairness, and the deep structures of trust that sustain organizational life [1]. This tension becomes especially visible in healthcare settings, where work is inherently relational, emotionally intense, and dependent on stable professional teams. Across many health systems, nurse turnover has assumed a structural character. It is no longer an episodic phenomenon linked only to salary disparities or local labor market contingencies, but the outcome of a broader erosion of the conditions that make work sustainable over time [2]. Organizations increasingly struggle to maintain continuity of care, rebuild cohesive teams, and preserve the knowledge that circulates informally through routines, conversations, and shared clinical practice. This fragility becomes even more pronounced when the work environment is highly regulated, time-sensitive, and exposed to growing pressures of accountability, as in hospitals [3].

Within this scenario, the expectations of newer generations entering the healthcare workforce reveal a profound cultural transition. Younger professionals demand a different equilibrium between private life and organizational obligations; they expect transparency in scheduling criteria, equitable distribution of burdensome shifts, and a participatory voice in decisions that affect their daily routines [4]. Traditional managerial approaches—often based on hierarchical allocation of shifts and implicit bargaining—are increasingly perceived as opaque or qualitatively unfair. These perceptions, even when not explicitly articulated, gradually erode trust and amplify intentions to leave [5].

At the same time, AI-based systems are beginning to transform organizational processes that had long resisted formalization. Workforce scheduling is emblematic in this respect. The planning of shifts is not a mere logistical task; it is the concrete point where personal lives, professional identities, and organizational constraints collide. The inherent tension between individual preferences, regulatory requirements, and clinical needs turns scheduling into a dense arena of negotiation, often governed by informal adjustments, implicit reciprocity, or managerial mediation. When such processes lack transparency, they easily generate suspicion, perceived favoritism, and disengagement [6].

The hospital examined in this study faced an increasingly unsustainable dynamic: over several years, nurse turnover reached a level that threatened the stability of care units and undermined the continuity of services. The organization recognized that this pattern was not reducible to remuneration or contract type alone. Instead, it reflected a deeper sense of unpredictability, inequity, and lack of voice in daily work decisions—particularly around shift allocation. In response, the leadership chose not to treat AI as a purely technical solution to automate scheduling, but as an instrument around which to redesign the participatory architecture of the organization [7,8].

The initiative combined three integrated components: an AI-enabled scheduling system capable of processing individual preferences and generating draft rosters; a Nursing Practice Council (NPC) functioning as a structured arena for shared governance, collective interpretation, and negotiation; and a set of data dashboards designed to make workforce patterns visible over time—night shift distribution, turnover trends, staffing fluctuations, and early indicators of engagement [9,10,11].

The resulting configuration is not a technological innovation in isolation. It is a socio-technical experiment in which AI, professional norms, and participatory structures co-evolve. Unlike many accounts that frame AI as a mechanism of managerial control—or, conversely, as a miraculous solution to inefficiency—this study adopts a more nuanced perspective: AI becomes meaningful not because of its predictive capacity, but because it reshapes the conversational and institutional spaces in which everyday decisions are made.

This paper therefore addresses three interconnected questions:

• How does AI reposition participation within the organizational system when introduced through transparent and negotiated processes?
• What mechanisms link AI-enabled visibility to perceived fairness, predictability, and the emotional climate of work?
• How can these mechanisms contribute to people sustainability in healthcare, understood not as a static outcome but as an evolving socio-technical capability?

By examining these questions, the study contributes to a growing body of literature that seeks to understand how AI can be integrated into complex human systems not as an external governing force, but as an enabling device that supports transparency, dialog, and more equitable organizational practices.

2. Theoretical Background

The literature on organizations operating at the intersection of AI, work, and sustainability can be interpreted through at least three traditions that, while developing somewhat independently, are increasingly converging: Sustainable Human Resource Management (SHRM), algorithmic management in its socio-technical evolution, and emerging perspectives on people sustainability. What these strands share is a persistent tension between technical efficiency, procedural equity, and the relational resilience of the organization—particularly visible in healthcare but relevant far beyond it [12,13].

SHRM originates from the idea that HR management cannot be reduced to maximizing short-term performance. The perspective opened by Ehnert (2009) [12] introduces a deeper notion of sustainability: not simply compliance with external constraints, but a process of preserving, regenerating, and developing human resources over time. This approach departs from traditional HRM models, often centered on linear efficiency, and embraces a paradoxical logic: organizations must guarantee productivity and continuity while at the same time protecting the physical, cognitive, and emotional energies of their workforce [13].

In health professions, this reading is particularly central. Research shows that sustainability is not only about recruitment, but about the capacity to create work environments that are equitable, predictable, participatory, and psychologically safe—conditions that reduce the risk of professional flight and strengthen organizational identification [2,8]. Studies on the Magnet Recognition Program confirm this interpretation: hospitals that value autonomy, professional voice, and shared governance tend to exhibit higher levels of retention, engagement, and quality of care [14,15,16,17]. In other words, human sustainability is not achieved through extrinsic incentives alone, but through the quality of relationships, participation, and transparency.

A second relevant tradition concerns the transformation of organizational coordination in the digital era. Early research on algorithmic management—especially in platform-based work systems—highlighted significant risks: opacity of decision processes, reduction in autonomy, and forms of subtle control that rationalize human activity to an extreme degree [18]. Algorithms were interpreted less as tools of support and more as “black boxes” that could withdraw agency from workers and weaken traditional channels of voice.

More recent work has begun to shift this interpretive center of gravity. Studies show that the impact of algorithms depends crucially on the social and institutional context in which they are introduced [19]. In systems characterized by participatory governance and strong professional communities, algorithms can take on a different function: not as instruments of imposition, but as cognitive and negotiation supports that help groups manage complex trade-offs. This is the core of the hybrid human–AI decision-making perspective, in which AI does not replace human judgment but complements it, making constraints more legible, consequences of choices more transparent, and distributive logic more understandable [20,21].

In the case of healthcare rostering, where constraints are numerous and roles are interdependent, algorithms that clarify rules and boundaries can reduce ambiguity, enhance the legitimacy of decisions, and lower the emotional charge associated with everyday coordination. Technology, in this sense, can become an “object around which the organization thinks,” rather than a device that dictates outcomes. The emerging literature on people sustainability adds a further interpretive layer. Aust et al. (2020) [22] argue that well-being at work cannot be treated as a static indicator but as an emergent process arising from the alignment between structural factors (resources, processes, policies) and relational ones (trust, equity, voice). What makes a work environment sustainable is not the absence of workload, but the capacity of the organization to design spaces for dialog, sharing, and psychological safety through which professionals can interpret and give meaning to the inevitable tensions of work.

A complementary stream of research—often located in operations management and health services research—examines AI and advanced analytics as decision-support tools for healthcare operations (e.g., staffing and rostering, patient flow, capacity planning, and resource allocation). Much of this work is oriented toward optimization, feasibility, and efficiency, whereas fewer studies examine how these systems reshape participation, legitimacy, and professional voice in day-to-day coordination. This gap is particularly salient for shift scheduling, a domain in which operational constraints intersect directly with work–life sustainability and perceptions of justice [6].

From this standpoint, AI is not seen as a purely technical instrument but as part of the infrastructure that can contribute—if intentionally designed—to making work more equitable, predictable, and comprehensible. People sustainability emerges from the interaction between algorithmic transparency, professional participation, shared governance structures, and collective sensemaking. It is precisely this interaction—not the technology alone—that constitutes the analytical core of the case presented here.

3. Materials and Methods

3.1. Study Design and Research Paradigm

We adopted an exploratory single-case study design to understand a socio-technical change process rather than to test a technology in isolation. The study is grounded in an interpretivist/constructivist paradigm, assuming that the effects of AI-enabled scheduling emerge through collective interpretation, negotiation of constraints, and evolving perceptions of procedural justice. Consistent with a sensemaking perspective, the analytic goal was theoretical (not statistical) generalization: to identify mechanisms through which AI-mediated transparency and participation may support people sustainability in healthcare organizations.

3.2. Setting and Intervention

The case concerns a private accredited hospital in Italy (192 beds) that experienced recurrent nursing turnover and escalating conflict around shift allocation. In response, the hospital introduced an AI-enabled rostering system (“Dream-Shift”) designed to generate draft schedules by processing staffing constraints and stated staff preferences. Importantly, the technology was embedded in a participatory infrastructure comprising (i) a Nursing Practice Council (NPC) as a structured arena for shared governance and negotiation, and (ii) HR dashboards used internally to visualize staffing patterns (e.g., night-shift distribution, staffing fluctuations) over time. The study focuses on the early implementation phase, during which the organization was actively stabilizing procedures and roles around the new system.

3.3. Data Sources and Collection

Data collection unfolded alongside implementation and drew on four complementary sources. (1) Administrative monitoring: the research team reviewed aggregated workforce indicators already produced by the hospital for managerial purposes (e.g., turnover series, night-shift distribution summaries, staffing fluctuations, and engagement indicators such as Gallup Q12 items). (2) Non-participant observations: between March and September 2025, the authors attended six NPC meetings (60–90 min each), taking structured fieldnotes on topics discussed, decision points, and interactional dynamics (tone shifts, contestation, and moments of agreement). (3) Internal documents: we analyzed protocols, guidelines, official communications, draft rosters, minutes, and project materials that documented how the organization described and institutionalized the change. (4) Informal feedback: short conversations and spontaneous comments in wards and handovers were recorded as anonymized jottings and used as an emotional barometer of the process. No patient-level data were accessed; all materials were handled in aggregated or anonymized form (Table 1).

Table 1. Data sources and their role in the analysis.

Source	Material	Analytic Focus	Use in This Paper
Administrative monitoring	Aggregated HR/ward dashboards and historical series	Contextual trends and distributional patterns	Descriptive backdrop; triangulation with qualitative evidence
NPC observations	Fieldnotes from council meetings	Sensemaking, negotiation, and hybrid decisions	Primary evidence for participation and fairness mechanisms
Internal documents	Protocols, minutes, communications, draft rosters	Formalization of rules and roles	Evidence of institutional framing and procedural stabilization
Informal feedback	Anonymized jottings from wards/handovers	Everyday perceptions and emotional climate	Illustrative perceptions supporting themes; not treated as standalone evidence

3.4. Data Analysis

Analysis followed an iterative, abductive logic. We began with a sensitizing framework derived from Sustainable HRM, algorithmic management, and people sustainability (e.g., voice/participation, fairness, predictability, and relational climate). Fieldnotes and documents were coded in cycles: an initial round to identify emergent themes and a second round to refine a shared codebook and map themes to mechanisms. Both authors independently coded a subset of materials to calibrate the codebook; differences were discussed until consensus was reached, and the remaining corpus was coded using the agreed definitions and an audit trail of decisions. Administrative indicators were used descriptively to contextualize and interrogate perceptions (e.g., whether perceived improvements aligned with monitored patterns), without claiming causal attribution. Where illustrative quotations are reported, they are anonymized and, where necessary, lightly edited for clarity.

3.5. Trustworthiness and Reflexivity

To enhance credibility and trustworthiness, we triangulated themes across multiple sources (observations, documents, and informal feedback) and maintained an audit trail of coding decisions and analytic memos. We conducted informal member checking by discussing preliminary interpretations with key organizational interlocutors (e.g., NPC coordinators) to verify that mechanisms identified reflected their experience. Reflexivity was addressed through explicit acknowledgement of researcher positioning: the authors are not employed by the hospital and had no managerial role in the intervention; however, their disciplinary backgrounds (organizational/health economics and occupational medicine) may influence attention to specific mechanisms (e.g., fairness, wellbeing). We mitigated this risk through joint interpretation sessions, systematic comparison across data sources, and careful separation of observed patterns from normative claims.

4. Results

4.1. From Opacity to Legibility: Fairness as Explainability

Across observations, documents, and informal feedback, the most immediate shift concerned the visibility of constraints. Before the intervention, rostering decisions were often interpreted through interpersonal lenses (“favoritism” or “being sacrificed”), because criteria were difficult to reconstruct from the outside. After deployment, the algorithm did not eliminate trade-offs, but it rendered them discussable: staff could see which constraints (coverage requirements, skill mix, contractual rules) limited specific requests and why certain allocations were difficult to change. This shift from guessing to understanding reduced the intensity of fairness disputes, even when outcomes remained imperfect.

As one nurse put it during an NPC discussion: “It’s not that everything is perfect now. But at least I know why I have that shift. Before I was just guessing. Now I understand.”

4.2. Predictability and the Ability to Plan Life Outside Work

Predictability emerged as a second recurring theme. In NPC discussions and informal feedback, nurses framed scheduling not merely as logistics but as a condition for organizing family responsibilities, rest, and recovery. The combination of earlier visibility of draft rosters and the possibility to request justified adjustments through a structured process reduced the sense of last-minute uncertainty described in the pre-intervention period. Even when preferences could not be fully satisfied, participants reported that the process felt more stable and less contingent on informal bargaining.

4.3. Redistribution of Emotional Burden and Changing Managerial Roles

A salient organizational effect concerned ward managers. Prior to deployment, managers were positioned as the sole arbiters of competing requests and thus became the focal point of dissatisfaction. During NPC observations, managers increasingly framed their role as facilitative: explaining constraints, translating system outputs into negotiable options, and supporting a shared interpretation of what was possible. In this sense, AI did not “decide” in place of managers; rather, it provided an external reference point that helped shift conflicts from personal to institutional terms.

4.4. Hybrid Human–AI Decision-Making in Practice: An Illustrative Trade-Off

The hybrid human–AI decision-making dynamic became visible when the NPC had to arbitrate between competing values rather than optimize a single objective. A recurring trade-off concerned balancing individual requests for specific weekends off against equitable rotation of night shifts and minimum-coverage requirements. In practice, the algorithm generated a draft roster that satisfied most stated preferences but revealed that accommodating a specific weekend request would concentrate night duties on a smaller subset of nurses in the subsequent weeks. The NPC used this transparency to negotiate an alternative: a temporary swap combined with an explicit “compensation” rule (e.g., priority in the next cycle) to maintain distributive balance while responding to a time-sensitive personal need. This episode illustrates how the algorithm functioned as a constraint-visualization device, while legitimacy was produced through collective deliberation and agreed rules.

4.5. Descriptive Administrative Monitoring as Contextual Evidence

Finally, internal administrative monitoring provided a descriptive backdrop to the qualitative themes. HR dashboards and historical series were reviewed to contextualize staff perceptions (e.g., monitored dispersion of night shifts, staffing fluctuations, and engagement items). Because these indicators were produced for operational purposes and not as a research dataset, we interpret them cautiously and avoid causal claims. Nonetheless, the monitored patterns were consistent with the qualitative evidence in that the organization was able to discuss distributional issues with greater precision (e.g., identifying units or periods where imbalance persisted) and to track scheduling-related concerns in a more structured way over time. Future longitudinal work is needed to quantify whether these early signals consolidate into stable changes in turnover and engagement.

5. Discussion

The findings suggest that the most significant impact of the AI-enabled scheduling system was not a narrow gain in technical efficiency but a shift in relational and interpretive dynamics. As illustrated by the NPC trade-off around weekend requests and night-shift equity (Section 4.4), the algorithm contributed primarily by making constraints visible. Legitimacy, however, was produced through deliberation: the NPC translated system outputs into collectively acceptable rules (e.g., temporary swaps and compensation logic), showing that hybrid human–AI decision-making is less about delegating choices to a machine and more about improving the quality of collective reasoning.

This challenges some dominant narratives about algorithmic management that emphasize opacity and control. While such risks remain salient in many contexts, the case supports the view that outcomes depend crucially on the institutional environment in which AI is embedded. Here, participatory governance acted as a sensemaking infrastructure: the NPC provided a stable arena for contesting, interpreting, and refining the algorithm’s outputs, preventing the system from becoming a “black box” and instead turning it into an object of negotiation.

From a people sustainability perspective, the mechanisms observed align with the idea that wellbeing and retention are shaped by process qualities—voice, predictability, and procedural justice—rather than by workload reduction alone. Fairness improved in a specific sense: not as perfect symmetry, but as explain ability and shared visibility of constraints. Predictability functioned as a form of organizational care by enabling staff to plan life outside work with greater confidence, which is particularly consequential in shift-based healthcare.

The case also helps reconceptualize managerial work under AI. Ward managers did not lose responsibility; instead, responsibility became less personalized and more institutional. With an external reference point (the draft roster and its constraint logic), managers could move from defensive bargaining to facilitative leadership, coordinating dialog and making trade-offs explicit. This resonates with shared governance traditions (e.g., Magnet-inspired practices) in which leadership is exercised through enabling professional voice rather than through unilateral allocation.

Several limitations should be noted. First, this is a single-case study in a specific organizational setting, and findings aim at theoretical rather than statistical generalization. Second, administrative indicators were not collected as a research dataset; consequently, quantitative claims are intentionally cautious. Third, although we addressed researcher bias through reflexive practice and triangulation, interpretive work necessarily reflects the authors’ theoretical lenses. Future research should pursue longitudinal and comparative designs to test whether the observed mechanisms translate into sustained improvements in turnover, engagement, and clinical outcomes.

6. Conclusions

The case analyzed in this paper shows that an AI-enabled scheduling system, when introduced alongside robust participatory governance, can contribute to people sustainability not by replacing human decision-making but by enabling more transparent, predictable, and dialogical processes.

The algorithm did not simplify the complexity of the hospital; it made it thinkable. By clarifying criteria and constraints, it allowed nurses and managers to reframe scheduling from a source of suspicion and stress into a shared organizational problem open to collective reasoning. Fairness improved because the rationale behind decisions became visible; predictability increased as professionals could plan their lives with greater confidence; ward managers moved towards a more facilitative and strategic role. This suggests that the key question is not whether AI can improve efficiency but whether it can help organizations become more transparent, equitable, and thoughtful in how they make decisions that affect every day work. When AI is aligned with participatory governance and professional values, it can reinforce—rather than erode—the human foundations of organizational life. As with any single-case study, findings should be interpreted with caution. The hospital examined has specific characteristics—in terms of size, culture, and professional history—that shape the way the intervention unfolded. The data also capture an early phase of the experience; while improvements in relational dynamics and administrative indicators are visible, their long-term stability remains to be assessed. Moreover, the introduction of AI was not an isolated event: the existence of the Nursing Practice Council, the openness of leadership, and the maturity of the nursing staff all played crucial roles that are not easily disentangled. Future research could develop this work along at least three directions. First, longitudinal analyses would help determine whether the positive dynamics observed here consolidate, evolve, or face new forms of resistance as familiarity with the system grows. Second, comparative studies across different hospitals, regions, or national contexts would make it possible to understand which elements of the model are context-specific and which may be generalizable as robust socio-technical configurations. Third, deeper investigation of the relationship between AI and professional identity—particularly in nursing—could shed light on how autonomy, responsibility, and clinical judgment are redefined when algorithmic systems become part of everyday work.

These questions point towards a broader research agenda: understanding how the future of healthcare work will be shaped not only by technological advances, but by the capacity of organizations to weave together technology, governance, and relationships in ways that sustain people rather than exhaust them.