Exploring HR Professionals’ Experiences with Generative AI in Recruitment: A Phenomenological Study in Public Hospitals in Uganda

Abstract

The growing integration of generative artificial intelligence (GenAI) in human resource management is transforming recruitment by enhancing decision-making efficiency and candidate screening worldwide. However, little is known about how HR professionals in resource-constrained settings, such as Uganda’s public health sector, experience and interpret these technologies. Therefore, this study aimed to explore the lived experiences of HR professionals using GenAI tools in recruitment within selected public hospitals in Uganda. The study employed a qualitative phenomenological approach; in-depth semi-structured interviews were conducted with 12 HR officers from regional referral hospitals utilizing AI-assisted recruitment or digital screening systems. Thematic analysis identified five key themes: (1) navigating GenAI with limited formal training, (2) concerns about fairness and algorithmic bias, (3) balancing recruitment speed with human intuition, (4) ethical dilemmas and transparency gaps, and (5) the evolving role of HR professionals in an AI-augmented workplace. Participants expressed a complex mix of optimism about efficiency gains alongside skepticism rooted in infrastructural challenges, inadequate policy frameworks, and digital literacy barriers among applicants. Findings underscore the need for tailored AI training, robust ethical guidelines, and policy reforms addressing the contextual realities of low-resource environments. This study contributes valuable insights to the emerging discourse on AI adoption in HR management by foregrounding the human experiences behind technological change in Sub-Saharan Africa’s public sector. It advances understanding of how AI tools are socially negotiated in complex institutional settings, offering practical implications for enhancing AI integration and governance in healthcare recruitment.

Plain Language Summary

Exploring HR Professionals’ Experiences with Generative AI in Recruitment

The growing integration of generative artificial intelligence (GenAI) in human resource management is transforming recruitment by enhancing decision-making efficiency and candidate screening worldwide. However, little is known about how HR professionals in resource-constrained settings, such as Uganda’s public health sector, experience and interpret these technologies. Therefore, this study aimed to explore the lived experiences of HR professionals using GenAI tools in recruitment within selected public hospitals in Uganda.

Introduction

In recent years, the rapid rise of generative artificial intelligence (GenAI) has transformed human resource management (HRM) globally (Budhwar et al., 2023; Chowdhury et al., 2024; Namatovu & Kyambade, 2025a; Rane, 2024). From automating repetitive tasks such as resume screening and interview scheduling to drafting job descriptions and personalized candidate communications, GenAI is fundamentally shifting how organizations attract and select talent (Nyberg et al., 2025). This technological evolution has garnered significant scholarly attention worldwide across multiple sectors, including healthcare, education, and finance (Song et al., 2025; Weber et al., 2024). Yet, in the African context particularly in Uganda’s resource-constrained public health sector, research on HR professionals’ experiences with GenAI remains minimal. The uptake of GenAI in recruitment processes within public hospitals is of growing interest, as these settings face unique challenges such as persistent workforce shortages, limited infrastructure, and regulatory constraints. Our study positions “HR professionals’ experiences with GenAI in recruitment” as the focal point, exploring how their perceptions, adoption behaviors, and lived experiences shape the effectiveness and adoption of these tools in practice.

This study hinges on a phenomenological approach to capture the subjective experiences of HR personnel immersed in GenAI-enabled recruitment. Generative AI refers to algorithmic systems capable of creating human-like text, images, or structured outputs through models like GPT-4 (Feuerriegel et al., 2024; Gupta et al., 2025), while recruitment experience covers the spectrum of HR tasks from sourcing and screening to candidate engagement and decision-making (Purohit & Banerjee, 2025). Numerous factors are known to influence recruitment outcomes, including organizational culture, candidate volume, and compliance needs. GenAI interacts with these elements by promising increased efficiency, objectivity, and improved candidate fit; multiple studies report up to a 30% to 40% increase in hiring efficiency and reductions in bias through AI screening (Lievens & Dunlop, 2025). However, concerns related to bias amplification, transparency, data privacy, and the necessity of human oversight remain prominent (Phillips-Wren & Virvou, 2025).

In Uganda, public hospitals grapple with severe HR shortages with only about one doctor per 25,000 citizens and frequent understaffing in rural facilities (Mutumba et al., 2025). The recruitment workload is intensified by high turnover, regulatory compliance (Namaganda et al., 2015), and the absence of digital infrastructure. These challenges underscore the potential value of GenAI tools in streamlining hiring and lightening administrative burdens. Yet, the lack of supportive policies, digital literacy limitations, and cultural misunderstandings pose barriers that need to be understood through the lived experiences of frontline HR staff.

This study is grounded in the Technology Acceptance Model (TAM) and complemented by Sociotechnical Systems Theory and the Unified Theory of Acceptance and Use of Technology (UTAUT) to interpret HR professionals’ experiences with GenAI in recruitment. TAM, originally proposed by Davis (1989), posits that two primary factors like perceived usefulness and perceived ease of use influence an individual’s decision to adopt a new technology. In the context of GenAI in recruitment, these constructs help explain how HR professionals form attitudes toward AI tools, especially when balancing efficiency gains with ethical or operational concerns (Cao & Duan, 2022; Kyambade & Namatovu, 2025a; Mutumba et al., 2025; Vrontis et al., 2022). UTAUT further expands this model by integrating constructs like social influence and facilitating conditions, which are particularly relevant in hierarchical, resource-constrained public institutions like Uganda’s hospitals (Dwivedi et al., 2021). Meanwhile, Sociotechnical Systems Theory emphasizes the interplay between technology and human actors within organizational settings, highlighting the importance of aligning AI systems with organizational values, job roles, and existing infrastructure (Pasmore et al., 2020; Sarpong & Rees, 2021). This theoretical triad enables a nuanced exploration of the complex, dynamic, and context-specific interactions that shape the adoption and impact of GenAI in recruitment. It is particularly valuable in Uganda’s public healthcare setting, where structural limitations and cultural expectations influence both the perceived utility and the social acceptance of emerging technologies.

Despite a growing body of global research on artificial intelligence (AI) applications in human resource management (HRM; Budhwar et al., 2023; Chowdhury et al., 2024; Halid et al., 2024; Namatovu & Kyambade, 2025a; Rane, 2024), there remains a notable paucity of studies that explore the lived experiences of HR professionals, particularly within public health sectors of low- and middle-income countries (LMICs) like Uganda. The vast majority of existing research tends to employ quantitative methods such as surveys and large-scale data analyses focusing on measurable outcomes like efficiency, accuracy, or bias reduction (Pan & Froese, 2023; Zong & Guan, 2025). This emphasis often overlooks the rich, subjective perspectives of frontline HR practitioners who engage directly with generative AI (GenAI) tools.

Furthermore, the current literature is heavily concentrated in developed Western contexts and, to some extent, Middle Eastern countries (Budhwar et al., 2023; Chowdhury et al., 2024; Halid et al., 2024; Rane, 2024), with scant attention paid to Sub-Saharan Africa, despite its unique socio-economic and technological landscapes. This geographical skew limits the generalizability of findings and neglects contextual factors such as infrastructural constraints, digital literacy, and cultural attitudes toward AI that are critical in African settings. In addition, much of the extant research on generative AI’s role in HR has concentrated on sectors like education and corporate recruitment (Aguinis et al., 2024; Ardichvili et al., 2024; Budhwar et al., 2023; Chowdhury et al., 2024; Kyambade & Namatovu, 2025c; Kyambade, Namatovu, & Male Ssentumbwe, 2025), while healthcare especially public health institutions in resource-limited environments, have been largely overlooked. Given the critical importance of effective recruitment in healthcare settings and the distinct challenges faced by public hospitals in LMICs, this gap is particularly striking.

Our study addresses these deficiencies by employing a phenomenological qualitative approach to investigate how HR professionals in Uganda’s public hospitals experience, interpret, and adapt to GenAI tools in recruitment. By foregrounding the voices of HR practitioners within Sub-Saharan Africa’s healthcare sector, this research offers unique insights into the sociotechnical integration of AI in an underexplored yet vital context. This study, therefore, addresses a unique research question;

What are the lived experiences of HR professionals using generative AI in recruitment within Uganda’s public hospitals?

Literature Review

Generative AI (GenAI) refers to artificial intelligence systems capable of creating new content like text, images, audio, or video, based on learned patterns from large datasets. In the context of human resource management (HRM), tools such as GPT-4, ChatGPT, HireVue, and automated CV-screening bots have revolutionized recruitment practices by automating core tasks including resume analysis, job description generation, candidate pre-screening, and interview scheduling (Suen et al., 2020; Van Esch & Black, 2021). These tools harness natural language processing (NLP) and machine learning to mimic human judgment, accelerating decision-making while reducing manual labor. Platforms like HireVue, for example, leverage video analytics and AI-based scoring to evaluate candidates’ language, facial expressions, and tone, with reported efficiency gains of up to 40% in early-stage screening processes (Purohit & Banerjee, 2025). ChatGPT and similar large language models have also been adopted for drafting tailored communication, automating FAQs, and offering real-time interaction with applicants (Lu & Arndt, 2023).

However, the rapid deployment of GenAI in HR has raised critical ethical, psychological, and operational concerns. Scholars warn of algorithmic bias, lack of contextual understanding, and the potential to reinforce existing social inequities, especially when these tools are trained predominantly on Western-centric datasets (Binns, 2020; Mehrabi et al., 2021). A candidate from a non-traditional background, for example, may be unintentionally penalized due to mismatched educational nomenclature or unfamiliar institutional rankings (Ntoutsi et al., 2020). Moreover, candidates often receive little to no feedback from automated systems, which erodes transparency and challenges principles of procedural fairness in hiring (Budhwar et al., 2023; Chowdhury et al., 2024). The “black box” nature of AI decision-making where the logic behind an outcome is hidden, further complicates accountability, especially in regulated sectors such as public health and education (Floridi & Cowls, 2021; Kyambade, Kagere, et al., 2024; Pasquale, 2020).

In developing regions, especially sub-Saharan Africa, the adoption of AI technologies in HRM faces complex challenges. These include insufficient digital infrastructure, low internet penetration, and a lack of technical skills among HR personnel (Kyambade, Kagere, et al., 2024; Ndung’u & Signé, 2020). A study by Chilunjika et al. (2022) and Namatovu and Kyambade (2025c) highlights how public institutions often lack the capacity to integrate AI tools effectively, relying on donor-funded pilots or externally built platforms with minimal localization. Even when digital tools are introduced, poor digital literacy limits their usage and amplifies dependency on external vendors (Chakravorti et al., 2022).

Ethical concerns are further exacerbated in these contexts due to the absence of robust regulatory frameworks governing data privacy, algorithmic fairness, and AI accountability (Adjekum et al., 2021; Kyambade, Kisseka, & Namatovu, 2025). For example, while countries like Uganda have enacted data protection laws (e.g., the Data Protection and Privacy Act, 2019), enforcement remains weak, leaving public HR systems vulnerable to misuse of candidate data. Despite these constraints, localized innovations show promise. For instance, Afriwork, a platform operating in East Africa, utilizes AI-driven matching through low-bandwidth applications such as Telegram and WhatsApp to connect job seekers with employers. Yet, disparities in access, algorithmic transparency, and inclusion persist, raising questions about equitable adoption of AI tools (Mutimukwe & Olutola, 2023).

While global literature on AI in recruitment is expanding rapidly, most studies focus on corporate settings in high-income countries, overlooking how HR practitioners in resource-constrained environments experience these technologies (Budhwar et al., 2023; Chowdhury et al., 2024; Halid et al., 2024; Rane, 2024). Moreover, the bulk of research emphasizes outcomes such as efficiency, accuracy, or candidate perceptions rather than the inner experiences of recruiters navigating these systems in real time (Chien & Chen, 2022; Suen et al., 2020). In Africa, very few studies address the interaction between GenAI and HRM in public institutions, let alone in mission-critical sectors like healthcare. A recent review by Adepoju et al. (2023) identified a lack of empirically grounded research on human-AI collaboration in government-led HR functions across the continent.

Additionally, while candidate experiences with AI are often studied, the perspectives of HR professionals, the implementers and mediators of AI systems remain largely neglected (Margherita & Bua, 2021; Mutimukwe & Olutola, 2023). This study thus addresses a timely and critical gap by exploring how HR professionals in Uganda’s public hospitals perceive, adapt to, and make ethical judgments about GenAI in recruitment. In doing so, it aligns with recent scholarly calls for more context-sensitive, phenomenological research into AI’s integration within public service delivery (Chakravorti et al., 2022; Floridi & Cowls, 2021). Specifically, there is a lack of evidence on how GenAI is experienced in African public health institutions, where both institutional capacity and cultural expectations differ from the settings in which these tools are typically developed. This study addresses this gap by exploring the lived experiences of HR professionals in Uganda’s public hospitals, providing insights into how global technologies are interpreted, challenged, or adapted in local contexts (Adepoju et al., 2023; Vrontis et al., 2022).

Theory

This study is theoretically grounded in the Technology Acceptance Model (TAM), complemented by the Unified Theory of Acceptance and Use of Technology (UTAUT) and Sociotechnical Systems Theory, to offer a multidimensional understanding of how HR professionals in Uganda’s public hospitals experience and interact with generative AI (GenAI) in recruitment processes. TAM, developed by Davis (1989), remains one of the most widely applied frameworks for explaining technology adoption in organizational settings. It asserts that perceived usefulness and perceived ease of use are the two core determinants influencing users’ attitudes toward, and subsequent adoption of, new technologies. In the context of AI-enabled recruitment, these constructs are particularly relevant, as HR professionals often weigh the benefits of automation such as speed, accuracy, and reduced workload, against usability concerns, unfamiliar system logic, and ethical apprehensions (Cao & Duan, 2022; Kyambade et al., 2023; Vrontis et al., 2022). For example, when GenAI tools are perceived as opaque or difficult to customize, even their potential efficiency gains may not translate into enthusiastic or sustained use.

To enrich the explanatory power of TAM, this study also draws from the Unified Theory of Acceptance and Use of Technology (UTAUT), developed by Venkatesh et al. (2003) and later expanded in UTAUT2. UTAUT adds critical variables such as social influence, performance expectancy, effort expectancy, and facilitating conditions. These constructs are especially relevant in public health institutions in developing countries, where hierarchical organizational structures, limited training resources, and rigid bureaucratic norms often shape technology uptake. For instance, if hospital leadership champions AI systems and provides sufficient infrastructural and technical support, HR professionals may be more willing to embrace GenAI tools despite initial reservations (Dwivedi et al., 2021; Namatovu & Kyambade, 2025b). Conversely, if there is a perceived lack of peer support or managerial clarity, adoption may stagnate, regardless of the tool’s technical capabilities.

Sociotechnical Systems Theory provides a further layer of understanding by emphasizing the interdependence between social (human) and technical (machine) components in organizational systems (Pasmore et al., 2020; Sarpong & Rees, 2021). This perspective moves beyond user-centric models like TAM and UTAUT to consider how GenAI tools must align not just with individual perceptions, but also with institutional workflows, values, and constraints. For instance, a GenAI system that does not accommodate local hiring norms or fails to process regional data formats may introduce friction into recruitment workflows, prompting informal “workarounds” by HR staff or diminishing trust in the technology. This theory is particularly germane in Uganda’s public hospitals, where HR departments often operate in complex, under-resourced environments and where technology must be socially embedded to be effective. It highlights that the success of AI in HR is not solely a function of system design but is contingent on how well the technology is integrated with the human systems it intends to support.

These three theoretical frameworks offer a comprehensive lens to analyze GenAI adoption and use. TAM foregrounds individual-level cognitive and behavioral responses; UTAUT incorporates broader organizational and social determinants; and Sociotechnical Systems Theory situates these within dynamic institutional contexts. This triadic approach is particularly valuable for understanding GenAI in Uganda’s healthcare sector, where structural limitations, cultural expectations, and professional norms collectively shape the adoption landscape. Applying these theories allows this study to go beyond simplistic evaluations of “success” or “failure” and instead explore the nuanced and often contested process through which emerging technologies are understood, negotiated, and embedded in everyday HR practice.

Methodology

Research Design

This study adopted a qualitative phenomenological research design, drawing on foundational frameworks by Moustakas (1994) and Van Manen (2016), to explore the lived experiences of HR professionals who used GenAI tools in recruitment within Uganda’s public hospitals. Phenomenology was considered appropriate because it aimed to uncover the subjective, first-person perspectives of individuals as they navigated the novel and complex phenomenon of AI integration in human resource management in a resource-constrained healthcare environment (Kyambade & Namatovu, 2025b). Unlike quantitative methods that focus on measurement or prediction, phenomenology focused on capturing the rich, nuanced meanings that participants assigned to their experiences, revealing how they perceived, interpreted, and emotionally responded to the introduction of GenAI systems. This approach was critical for understanding the complexities of technology adoption, which involved cognitive, emotional, and ethical dimensions not easily captured through numerical data alone (Creswell & Poth, 2018). By emphasizing HR practitioners’ narratives, the study sought to illuminate both the practical challenges and benefits they encountered as well as the deeper meanings and assumptions influencing their engagement with AI recruitment tools.

Furthermore, phenomenology allowed the researchers to bracket preconceived notions and biases, thereby enabling an open exploration of participants’ authentic experiences (Moustakas, 1994). This was particularly important in the Ugandan public hospital context, where GenAI adoption was relatively new and influenced by unique cultural, infrastructural, and organizational factors. Through systematic, in-depth interviews and thematic analysis, the study captured the essence of how GenAI affected recruitment practices, decision-making, trust, and ethical concerns as experienced by HR professionals. Overall, the phenomenological design provided a robust framework for generating rich, contextualized insights into human-AI interaction in public healthcare recruitment, a crucial step toward informing tailored policies, training, and ethical guidelines for AI integration in similar low-resource settings.

Research Setting

The fieldwork for this study was conducted across five public hospitals in Uganda, each of which had integrated some level of digital recruitment technology, including generative AI (GenAI) tools. These hospitals comprised Mulago National Referral Hospital, Uganda’s largest and most specialized public health facility and three regional referral hospitals located in Mbarara, Bushenyi, and Mbale. These sites were purposively selected based on prior evidence of using digital systems to manage recruitment processes, either through automated CV screening, AI-assisted interview scheduling, or other GenAI-enabled functionalities. The choice of these hospitals was strategic to capture a diversity of experiences across different geographic and administrative contexts, representing both urban and semi-urban settings within Uganda’s public health sector. Mulago Hospital, being a national referral center, often led in adopting new technologies due to greater resources and exposure, whereas the regional referral hospitals faced more pronounced infrastructural and human resource challenges. This variation provided a richer understanding of how GenAI tools were experienced in different resource environments.

For conceptual clarity, this study distinguishes between generative artificial intelligence (GenAI) tools and AI-assisted recruitment technologies more broadly. GenAI refers specifically to systems capable of generating novel textual or communicative content, such as AI-generated job descriptions, automated candidate correspondence, or interview question prompts (e.g., GPT-based applications). In contrast, AI-assisted recruitment tools encompass traditional machine learning and natural language processing (NLP)-based systems used for functions such as automated CV screening, applicant ranking, and interview scheduling, which support recruitment decision-making without producing original content. Participants in this study reported experience with a combination of both generative AI tools and AI-assisted recruitment systems; accordingly, the term “GenAI” is used in this study to reflect AI-enabled recruitment practices that incorporate generative functionalities within broader digital recruitment infrastructures in Uganda’s public hospital context.

Participants and Sampling

The study engaged 12 human resource (HR) professionals, comprising HR managers and recruitment officers, who were purposively selected based on their direct and sustained involvement in AI-supported recruitment processes within Uganda’s public hospitals. Purposive sampling was employed to ensure that only those with firsthand experience using generative AI (GenAI) tools in recruitment activities were included, thereby aligning with the phenomenological aim of capturing deep, lived experiences (Palinkas et al., 2015). Eligibility criteria required that participants had at least 6 months of active engagement with GenAI tools. These tools included, but were not limited to, automated curriculum vitae (CV) screening systems, AI-assisted job description generators, applicant tracking platforms incorporating natural language processing (NLP), and AI-powered interview scheduling or chatbot interfaces. This ensured that participants had sufficient exposure to reflect meaningfully on the impact of GenAI in their day-to-day professional roles.

To enhance transparency regarding sample diversity, the 12 participants were drawn from five public hospitals, comprising three urban facilities and two semi-urban regional hospitals. Participant representation included HR managers (n = 5) and recruitment officers (n = 7), reflecting varying institutional contexts. GenAI exposure also varied: while all participants met the minimum requirement of 6 months of use, their experience ranged from 6 to 18 months, and the tools they interacted with differed across sites including automated CV-screening modules, NLP-enabled applicant ranking systems, AI-assisted job description generators, and chatbot-based candidate communication. This distribution ensured that the sample captured a broad spectrum of operational realities and levels of GenAI integration across Uganda’s public hospital ecosystem.

Data saturation was achieved by the tenth interview, at which point no new themes, categories, or substantive insights were emerging from the data, and subsequent interviews yielded repetition rather than conceptual expansion. Two additional interviews were nevertheless conducted to confirm saturation and ensure the robustness and adequacy of the thematic structure. This sampling decision aligns with established phenomenological research guidelines, which suggest that samples of 10 to 15 participants are sufficient to achieve saturation when participants share rich, relevant lived experiences (Creswell & Poth, 2018; Guest et al., 2006).

Data Collection

Data were collected using semi-structured, in-depth interviews, which provided the flexibility to explore individual experiences while maintaining consistency across participants. Each interview lasted approximately 45 to 60 min, depending on the depth and flow of the conversation. Interviews were conducted either face-to-face or via Zoom, depending on participant availability and institutional COVID-19 protocols in place at the time. All interviews were audio-recorded with prior informed consent to ensure accurate transcription and analysis, while also upholding ethical standards related to privacy and confidentiality. The interview guide was designed based on principles of phenomenological inquiry, with a focus on drawing out participants’ lived experiences and subjective interpretations of generative AI (GenAI) tools in recruitment. The questions were open-ended and exploratory, allowing participants to steer the conversation based on their individual narratives. Sample questions included “Can you describe your first experience using generative AI in recruitment?”“What are the most significant challenges you encounter when using GenAI systems?”“How do these tools fit within your institutional hiring processes or recruitment goals?”“Have your perceptions of AI changed over time? If so, how and why?” and “Can you share a moment where you felt the AI worked well or failed, in your recruitment experience?”

These prompts were intended to elicit rich, descriptive accounts of interaction with GenAI tools, while also capturing contextual and emotional dimensions. Follow-up and probing questions were used to clarify meanings, uncover underlying values or tensions, and allow participants to reflect more deeply on their evolving relationship with AI-enabled systems. Interviews were transcribed verbatim shortly after each session and cross-checked for accuracy. Field notes were also taken to document non-verbal cues, emotional responses, and contextual observations that could enrich interpretation. In cases where participants used technical terms or local references (e.g., specific software names, job role codes), clarifying information was added during transcription to preserve meaning.

Data Analysis

The data analysis process was guided by Moustakas’ (1994) phenomenological method, which was selected for its focus on uncovering the essence of lived experience. All interviews were first transcribed verbatim to ensure that every spoken word was preserved for detailed analysis. Transcriptions were then imported into NVivo 12, a qualitative data analysis software, which was used to organize, code, and retrieve data efficiently. The first step in the analysis involved epoché or bracketing, where the researchers consciously set aside personal biases, preconceptions, and prior experiences with AI in recruitment to approach the data with an open and reflective mindset. This was essential for maintaining phenomenological integrity and allowing the participants’ voices to remain central throughout the analysis process (Creswell & Poth, 2018). Next, the transcripts underwent horizonalization, a stage in which each significant statement made by the participants was given equal weight and considered meaningful. These statements were then reviewed and clustered into meaning units, which represented core aspects of the participants’ shared experiences. For example, comments related to the opacity of GenAI decisions, concerns about data privacy, or reliance on intuition were grouped into thematic clusters that captured recurring patterns across multiple interviews.

The process continued with the development of textural descriptions which captured what the participants experienced and structural descriptions, which focused on how those experiences occurred within specific institutional and technological contexts. This step allowed the researcher to interpret not only the content of the experiences but also their situational and emotional nuances. Finally, a composite description was constructed that synthesized both the textural and structural dimensions into a cohesive understanding of the overall phenomenon. This synthesis revealed several rich and nuanced themes, such as “trust and transparency,”“efficiency paradox,”“ethical dilemmas,” and “human versus machine judgment.” These themes were not treated as discrete categories but were understood as interconnected layers of a complex, dynamic interaction between HR professionals and GenAI tools in public hospital recruitment settings. Throughout the analysis, iterative coding and constant comparison were employed to refine interpretations and ensure that emerging insights were grounded in participants’ original language and lived realities. Member checking and peer debriefing were used to enhance the credibility and confirmability of the findings, supporting the study’s overall trustworthiness.

Researcher Positionality and Reflexivity

The research team comprises scholars with backgrounds in human resource management, public sector governance, and digital transformation in low-resource settings, with prior academic and professional engagement in technology-enabled organizational practices. We acknowledge that these experiences may carry assumptions regarding the potential benefits and risks of AI adoption in recruitment. To manage potential bias, reflexive practices were employed throughout the study, including bracketing of prior assumptions, maintaining reflexive analytic memos, and engaging in regular peer debriefing during data analysis. These practices ensured that interpretations remained grounded in participants’ lived experiences rather than researchers’ preconceptions.

Trustworthiness and Rigor

To ensure methodological rigor and trustworthiness, this study adhered to Lincoln and Guba’s (1985) established criteria for qualitative research: credibility, dependability, transferability, and confirmability. These criteria provided a robust framework for evaluating the quality and authenticity of the findings, particularly in a phenomenological inquiry that depends heavily on participants’ subjective narratives. Credibility was achieved primarily through member checking, a process where participants were invited to review emergent themes and preliminary interpretations to confirm the accuracy and resonance of the findings. This step ensured that the representations of participants’ experiences were faithful to their intended meanings and minimized researcher misinterpretation. Additionally, prolonged engagement with the data and repeated immersion in transcripts further enhanced internal validity. Dependability was maintained by creating a detailed audit trail, documenting every step of the research process from participant recruitment and interview procedures to coding decisions and theme development. Furthermore, peer debriefing was conducted among the research team to critically reflect on analytical choices, examine assumptions, and enhance consistency in interpretation. This process helped expose potential bias and improved the transparency of methodological decisions.

To address transferability, the study employed rich, thick description of the research setting, participant demographics, and contextual factors surrounding GenAI use in public hospitals. By providing nuanced insights into Uganda’s healthcare recruitment environment such as the digital divide, resource limitations, and institutional governance, the findings offer relevance to similar low- and middle-income contexts where AI integration in HR is emerging. Confirmability was strengthened by triangulating interview data with Supplemental materials such as institutional policy documents, particularly HR manuals detailing recruitment procedures, and the use of a reflective research journal maintained by the primary researcher. This journal captured analytic memos, methodological reflections, and decisions made during the study, providing evidence that findings emerged from participants’ accounts rather than researcher bias or imagination. These strategies ensured that the study’s findings were credible, traceable, and contextually grounded, enhancing the reliability and integrity of insights into HR professionals’ lived experiences with generative AI in recruitment. Participants received clear information about the study, gave written consent, and were assured of confidentiality and the right to withdraw at any time. Identifiers were anonymized, and data were stored securely following university ethics protocols.

Findings

Thematic analysis of interviews with 12 HR professionals from five public hospitals revealed four dominant themes characterizing their experiences with generative AI (GenAI) in recruitment. These themes include: Learning and Adapting to GenAI Tools, Trust and Algorithmic Bias, Efficiency versus Human Judgement, and Ethical and Transparency Concerns. Direct quotations from participants (e.g., HR01, HR02) are used to illustrate themes.

Theme 1: Learning and Adapting to GenAI Tools

Participants in this study consistently described the adoption of generative AI tools in recruitment as a challenging and often frustrating learning process, marked by a steep learning curve and a lack of formalized training or ongoing technical support. Many HR professionals narrated how initial training sessions were insufficient for the realities they faced when working with real candidate data, which frequently exposed the limitations and rigidities of the AI systems.

For example, one participant (HR03) reflected on the complexity of the system’s “logic,” stating, “I had to learn quickly because the system kept rejecting candidates, I thought were qualified, it took me weeks to figure out its ‘logic’.” This highlights not only the technical challenge but also the cognitive effort needed to decode the AI’s decision-making criteria, which were often non-intuitive and opaque. The difficulty was compounded by the AI’s inability to correctly interpret local educational qualifications or terminologies, as HR07 explained: “We started with a one-hour training, but nothing prepared us for real data: the AI struggled with degree names like ‘BCom’ vs. ‘Bachelor of Commerce’.” This mismatch between the AI’s programming, often based on standardized or global datasets and local linguistic or educational nuances created persistent barriers to smooth implementation.

Furthermore, HR professionals described a tedious cycle of trial and error, with HR01 noting, “Every time I uploaded a candidate shortcut, I was surprised. The AI would flag an unusual graduation year and reject the file.” Such technical glitches forced HR officers into quasi-technical roles, requiring them to become “part-time AI trainers,” as HR05 put it, engaging in constant tweaking and keyword adjustments to coax the system into functioning properly. The participants’ need to manually rewrite job descriptions or reformat candidate information to fit the AI’s parsing parameters (HR08) reveals a significant gap between the technology’s design assumptions and the practical realities of recruitment in Uganda’s public hospitals.

This theme underscores that the integration of GenAI in HR recruitment is not a plug-and-play solution but rather a process of co-adaptation between users and technology. It reflects the informal and iterative nature of technology adoption in contexts with limited digital infrastructure and training. Importantly, it highlights the resilience and resourcefulness of HR professionals who, despite minimal support, developed workarounds and learned through experience to make GenAI tools operational. These findings resonate with broader studies in LMICs, where digital innovations require localized adaptation and continuous learning to achieve meaningful impact.

Ultimately, this theme illustrates that successful GenAI adoption in recruitment is contingent not only on the capabilities of the technology itself but also on the readiness, support, and ongoing capacity development of its human users, critical factors that need to be addressed through targeted training, user-centered design, and sustained technical assistance.

Theme 2: Trust and Algorithmic Bias

A pervasive concern among participants was the lack of trust in the generative AI system’s ability to fairly and accurately process locally contextualized data. HR professionals expressed doubts about whether the AI could properly interpret Ugandan naming conventions, educational credentials, and regional diversity, leading to suspicions of embedded biases, many of which remained unverifiable due to the opaque nature of the algorithms.

For instance, one participant (HR04) voiced skepticism about the AI’s capacity to handle culturally specific information, stating, “Sometimes I wonder if it understands Ugandan naming patterns or educational pathways, it might be biased against regional names or diploma holders.” This concern highlights the system’s potential lack of cultural competency, where AI trained predominantly on global or Western-centric data sets may misclassify or undervalue qualifications and names that do not fit predefined templates. Similarly, HR10 observed an instance where a candidate with a legitimate degree, Uganda’s premier institution, was downgraded because the AI system failed to recognize the qualification’s format: “I’ve seen it downgrade a candidate with a valid university qualification because it didn’t match international formats.” This example underscores the risks of AI systems reinforcing structural inequities by privileging credentials or profiles aligned with international standards while marginalizing locally relevant qualifications.

Participants also reflected on the paradox of AI systems replicating, and sometimes amplifying, the very biases that human-led recruitment efforts seek to eliminate. As HR06 remarked, “Our concern is that it reinforces the same biases we’re trying to correct, maybe even more so.” This points to the phenomenon of algorithmic bias, where training data containing historical inequities leads AI to perpetuate discriminatory patterns, a risk well documented in the literature. The inability to interrogate or audit the AI’s decision-making process often described as the “black box” problem, further undermined confidence in the system’s fairness. HR11 shared, “Some staff argued: ‘The AI selects only Kampala-based grads,’ but we can’t prove it because decisions are hidden.” Such opacity creates a trust deficit, particularly acute in public institutions where recruitment decisions must be transparent and accountable.

Regional disparities also emerged as a subtle yet troubling consequence of the AI’s filtering. HR02 noted, “Candidates from western Uganda seem underrepresented in shortlist, AI might favor candidates with familiar-styled names.” This suggests that the algorithm may inadvertently prioritize candidates from certain regions or ethnic groups, disadvantaging others due to uneven representation in the training data or culturally biased pattern recognition. The concern that GenAI might unintentionally marginalize rural or underrepresented populations echoes broader warnings in AI ethics literature, emphasizing the importance of context-sensitive design and rigorous bias audits.

This theme illustrates a fundamental challenge in deploying AI-driven recruitment tools in contexts with diverse cultural and educational landscapes: the necessity of building algorithmic transparency, fairness, and adaptability. Trust in AI systems is not only technical but deeply social and cultural. Without mechanisms to explain, audit, and correct bias, HR professionals may remain hesitant to fully embrace GenAI technologies, potentially limiting their transformative potential in Uganda’s public health recruitment.

Theme 3: Efficiency Versus Human Judgement

A prominent theme emerging from the study was the perceived tension between the increased efficiency brought by generative AI tools and the diminishing role of human judgment in recruitment decisions. While participants appreciated the considerable time savings offered by AI, allowing them to process large volumes of applications quickly, they expressed concern that this speed often came at the cost of overlooking important qualitative factors that humans traditionally rely on.

One HR professional (HR09) reflected this trade-off by saying, “The AI is faster, but I feel like I’ve lost my intuition, things that used to ‘feel right’ are now overlooked.” This statement captures the essence of the loss many recruiters feel when relying heavily on algorithmic decisions that prioritize keyword matching or data patterns over the nuanced, experiential knowledge developed through years of human interaction with candidates. Another participant (HR03) lamented, “We used to spot leadership by traits in a CV; now we rely on text patterns, and miss those signals.” This points to the difficulty AI has in capturing intangible qualities such as leadership potential, emotional intelligence, or cultural fit, which often emerge through subtle cues beyond explicit data.

The volume-handling capability of GenAI tools was widely praised, with HR12 noting, “We can process 400 CVs in minutes, but some of the most promising candidates get skipped by the AI.” This suggests that while AI excels at screening for objective criteria quickly, it may fail to detect hidden gems, applicants who do not perfectly fit the algorithmic profile but possess strong potential or unique skills. The paradox of gaining efficiency at the expense of missing talent was succinctly summarized by HR05: “The paradox is clear: speed gains COULD be undermined by missed potential if AI isn’t well-tuned.” This observation underscores the critical need for continual refinement of AI tools to balance quantitative efficiency with qualitative insight.

To address these concerns, participants described adopting hybrid recruitment approaches that combine AI-driven initial screening with subsequent human review. HR07 stated, “We eventually added a manual check at the end, our human touch still matters.” This approach reflects an emerging consensus in the HR technology literature, where AI is viewed as augmenting rather than replacing human recruiters, enhancing capacity while preserving critical human oversight. The hybrid model allows HR professionals to intervene when the AI’s limitations become apparent, especially in evaluating soft skills, contextual factors, or non-traditional career paths that algorithms may undervalue.

This theme highlights the ongoing challenge in integrating AI within recruitment practices, especially in mission-critical sectors like healthcare where human judgment and empathy are crucial. It reinforces findings from prior studies indicating that AI’s efficiency gains must be tempered with strategies that preserve recruiter autonomy and relational insights. In Uganda’s public hospitals, where recruitment decisions impact not only workforce capacity but also patient outcomes, maintaining a balanced human-AI collaboration is particularly vital.

Theme 4: Ethical and Transparency Concerns

A significant theme that emerged from the participants’ narratives was deep unease regarding the ethical dimensions and transparency of generative AI (GenAI) systems used in recruitment. Respondents expressed anxiety about how these AI tools make decisions and who holds accountability for those decisions, underscoring a broader institutional concern about governance in AI deployment.

One participant (HR06) captured this uncertainty by stating, “We don’t know who developed it or how it makes decisions, we cannot challenge it legally or ethically.” This quote highlights the frustration with the opaque nature of AI algorithms, commonly referred to as the “black box” problem. HR professionals receive lists of shortlisted candidates generated by AI without clear explanations of the criteria or rationale behind these selections. As HR01 emphasized, “It feels like a black box: we feed CVs, get lists. But who is accountable for a bad hire?” This lack of transparency raises questions about responsibility, especially when recruitment decisions affect organizational performance and fairness.

Data privacy was another critical concern raised by participants. Given that many AI systems used in Ugandan public hospitals are often provided by foreign vendors or hosted on external servers, HR staff feared potential data breaches or misuse of sensitive candidate information. HR11 voiced this apprehension, saying, “There are privacy worries: candidate information is on foreign servers, what if data leaks?” The transfer and storage of personal data across borders without clear safeguards challenge existing legal and ethical frameworks in Uganda, where data protection laws are still evolving. HR10 further stressed this by noting, “No one explained whether it respects candidate consent or local data laws.” This lack of clarity on consent mechanisms and legal compliance undermines confidence in GenAI adoption.

The ethical issue of fairness was also front and center. Respondents feared that hidden biases within AI algorithms might perpetuate discrimination, yet the inability to audit or scrutinize the AI systems led to blind acceptance of their outputs. HR08 said, “We worry about bias, fairness, yet we can’t audit the system, so all decisions are accepted blindly.” This points to a governance gap where institutions lack the technical capacity or regulatory frameworks to ensure AI accountability. Without transparent auditing mechanisms, there is a risk that AI may reinforce existing inequities, contrary to the ethical mandates of public healthcare recruitment.

This theme reflects wider institutional anxieties that resonate with global debates about AI ethics and governance, especially in contexts where data protection laws and AI regulation are still. The uncertainty about vendor accountability, algorithmic fairness, and data sovereignty echoes findings from other low- and middle-income countries where AI adoption often outpaces ethical frameworks. In the Ugandan public health sector, these concerns are amplified by the critical nature of recruitment decisions affecting healthcare delivery and public trust. The findings underscore the urgent need for transparent AI governance models, comprehensive stakeholder engagement, and localized ethical guidelines to safeguard candidates’ rights and institutional integrity as AI tools become more prevalent.

Discussion

This study explored the lived experiences of HR professionals using generative AI (GenAI) in recruitment within Uganda’s public hospitals. Through a phenomenological lens, the findings illuminate how HR practitioners interact with, adapt to, and critique GenAI technologies in resource-constrained environments. Four main themes emerged: learning and adapting to GenAI tools, trust and algorithmic bias, efficiency versus human judgement, and ethical and transparency concerns. These findings align with and extend existing global and regional literature.

The findings affirm much of the existing research on AI adoption in HR, such as the challenge of rapidly upskilling HR professionals and addressing algorithmic bias (Binns, 2020; Van Esch & Black, 2021). However, this study also highlights contradictions, notably in the persistent mistrust due to cultural misalignment of AI models, which contrasts with more optimistic views of AI efficiency gains reported in high-income countries (Purohit & Banerjee, 2025). The Uganda-specific digital divide and infrastructural limitations further compound these challenges, as echoed by Kayanja et al., (2025), Chilunjika et al. (2022), and Ndung’u and Signé (2020). These contextual realities underscore that the integration of GenAI in recruitment is not a purely technical innovation but is deeply shaped by socio-economic and institutional factors.

Learning and Adapting to GenAI Tools

Participants consistently highlighted a steep learning curve and the absence of structured support for onboarding GenAI tools. This mirrors findings from global studies, where HR professionals reported that GenAI implementation often outpaces training and digital capacity-building efforts (Suen et al., 2020; Van Esch & Black, 2021). The adaptation process in Uganda was largely informal, relying on trial and error, reflecting broader concerns about digital readiness in Sub-Saharan Africa (Chilunjika et al., 2022). This resonates with Ndung’u and Signé (2020), who argue that while Africa is undergoing rapid digital transformation, the absence of AI-specific training hinders meaningful integration.

Furthermore, the struggle to understand and calibrate GenAI tools reflects the “technological anxiety” reported in other low- and middle-income country (LMIC) contexts, where AI systems are often introduced without contextual adaptation (Chakravorti et al., 2022). These challenges suggest that successful GenAI integration requires more than deployment, it demands localized capacity development and ongoing digital support.

In similar low-resource contexts, structured capacity-building initiatives have been shown to significantly ease GenAI adoption. For example, Rwanda’s community-led digital literacy programs for public health staff have improved confidence and reduced the trial-and-error learning burden (Hémono et al., 2024). Likewise, Kenya’s county health departments have begun integrating short GenAI orientation modules into routine HR training sessions, demonstrating that low-cost, context-specific training can accelerate adoption even in constrained environments. These examples highlight that Uganda’s informal learning approaches could benefit from more systematic, localized digital upskilling models.

Trust and Algorithmic Bias

Participants raised concerns about whether GenAI systems could interpret Ugandan data accurately, particularly in terms of naming patterns, educational institutions, and regional identities. This echoes findings from Binns (2020) and Mehrabi et al. (2021), who highlight the risks of algorithmic bias in AI systems trained predominantly on Western datasets. Studies show that GenAI can reinforce existing inequalities by favoring profiles aligned with global norms, thereby marginalizing candidates from underrepresented regions (Ntoutsi et al., 2020; Zliobaite, 2020).

In this study, HR professionals’ suspicions of cultural misalignment such as systematic underrepresentation of candidates from rural areas, underscore a critical ethical dilemma. These perceptions of bias erode trust in AI systems and can hinder adoption, consistent with prior research showing that algorithmic opacity can weaken user confidence, particularly in public sector institutions (Wirtz et al., 2019).

Experiences from other LMICs suggest that localized algorithm refinement can reduce cultural misinterpretation and strengthen trust. In Kenya and India, for instance, HR departments retrained AI screening models using local applicant data to improve recognition of indigenous names, rural educational backgrounds, and regional dialect patterns, resulting in lower false rejection rates and enhanced perceived fairness (Mamuli et al., 2025; Sharma et al., 2025). These cases demonstrate that contextual calibration is both feasible and impactful, offering a practical benchmark for Uganda’s health sector.

The potential regional and socio-cultural biases embedded in GenAI tools represent a significant ESG-related social risk. As sustainability scholarship demonstrates, unmanaged social risks can undermine organizational legitimacy, employee morale, and long-term performance (Kyambade, Bartazary, & Namatovu, 2024; Kyambade et al., 2025a, 2025b; Nguyen et al., 2025). If GenAI systems disproportionately exclude candidates from certain regions or socio-economic backgrounds, public hospitals risk eroding their social license to operate and compromising equity commitments central to public health. Addressing algorithmic bias is therefore not only an ethical requirement but a strategic necessity, directly linked to institutional sustainability and community trust.

Concerns about unequal treatment and exclusion raised by participants resonate with broader evidence on systemic inequities within Uganda’s public-sector institutions, particularly those related to fairness, representation, and participation in governance structures (Kyambade et al., 2026; Tushabe et al., 2023, 2025). These studies similarly highlight how institutional cultures and structural inequalities shape perceptions of fairness, paralleling the skepticism HR professionals express toward AI-driven screening.

Efficiency Versus Human Judgement

The findings suggest a dualistic experience: HR professionals appreciated the speed and volume-handling capacity of GenAI but lamented the loss of human intuition in assessing nuanced candidate attributes. This aligns with Purohit and Banerjee (2025), who found that while AI boosts recruitment efficiency, it often fails to evaluate qualitative dimensions such as interpersonal skills, adaptability, and leadership potential. Similar findings were reported in the healthcare context, where recruiters emphasized the value of intuition and relational judgment in assessing candidates’ suitability for emotionally demanding roles (Krishna et al., 2021; Kyambade & Namatovu, 2025e).

Hybrid models where GenAI tools are supplemented with manual review, emerged in this study as coping strategies to maintain human oversight. This reflects what Tambe et al. (2019) describe as “AI augmentation” rather than “AI replacement,” particularly in mission-critical sectors like healthcare where judgment and empathy are indispensable.

These recruitment challenges also carry sustainability implications. Uganda’s public hospitals already face chronic workforce shortages, making effective recruitment a critical determinant of long-term system viability. Insights from sustainable operations research indicate that organizations increasingly adopt digital tools to mitigate systemic human-resource risks (Abu Afifa et al., 2024). GenAI should therefore be understood not only as an efficiency innovation but as a strategic tool for sustaining workforce pipeline stability, reducing vacancy cycles, and supporting service continuity in resource-constrained healthcare systems. This positions GenAI adoption within broader debates on sustainability-oriented human resource management.

Ethical and Transparency Concerns

Participants voiced deep ethical concerns regarding decision-making transparency, data privacy, and legal accountability of GenAI systems. These apprehensions mirror the global literature highlighting the “black box” problem of AI, where users cannot easily trace or challenge algorithmic decisions (Floridi & Cowls, 2021; Pasquale, 2020). In public institutions, this concern is amplified by heightened responsibility to ensure fairness and compliance with national data protection laws (e.g., Data Protection and Privacy Act, 2019).

Lack of clarity on data storage, informed consent, and audit mechanisms exacerbates resistance to GenAI, especially when systems are sourced from external vendors. Similar trends have been observed in other LMIC healthcare settings, where AI tools deployed without stakeholder consultation raised trust and legitimacy issues (Adjekum et al., 2021; Kyambade & Namatovu, 2025d; Whitelaw et al., 2020). This reinforces the need for ethical frameworks and participatory governance in AI adoption in HR processes.

Comparable public-sector health systems in countries such as Bangladesh and Nepal have adopted “human-in-the-loop” hybrid recruitment models, where GenAI-generated shortlists are combined with manual review to mitigate bias and preserve human judgement (Arifuzzaman et al., 2023; Dhakal et al., 2023). These innovations align closely with coping strategies reported by Ugandan HR practitioners, underscoring that hybrid oversight offers a realistic pathway for responsible GenAI use in low-resource health institutions. Lessons from these contexts reinforce the value of co-designing AI processes with users to enhance legitimacy, transparency, and cultural fit.

The transparency and accountability gaps reported by HR professionals align with broader governance challenges observed in developing economies, where weak disclosure mechanisms undermine institutional performance. As Abu Afifa and Nguyen (2024) and Mugambwa et al., (2024) show, disclosure quality is a central pillar of organizational governance, directly shaping public trust and operational effectiveness. Viewed through this lens, the “black box” nature of GenAI in Ugandan public hospitals reflects not just a technical issue but a governance and disclosure deficit, where the absence of clear auditability and algorithmic transparency mirrors long-standing challenges in financial and operational reporting. Thus, demands for transparent AI systems can be understood as a new frontier of non-financial disclosure, essential for maintaining legitimacy in public service delivery.

Structural Challenges and Governance Failures

The experiences described by HR professionals cannot be fully understood without situating them within Uganda’s broader digital governance landscape. The challenges reported limited training, infrastructural gaps, opaque algorithms, and weak accountability, reflect not only internal institutional constraints but deeper structural and policy failures shaping AI implementation in the public sector. Uganda currently lacks a comprehensive national AI strategy, and the broader digital transformation agenda operates through fragmented sector-specific policies, leaving public institutions without clear guidance on AI procurement, data governance, or algorithmic accountability. These structural gaps reflect patterns identified in other Ugandan institutional studies, particularly research on gendered governance and underrepresentation within public universities, where entrenched institutional norms and capacity constraints limit equitable participation and decision-making (Tushabe et al., 2023, 2025). The same institutional weaknesses that undermine gender-inclusive governance also constrain the ability of public hospitals to negotiate transparency, accountability, and contextual alignment in AI procurement and deployment. This policy vacuum creates conditions where public hospitals adopt GenAI tools in an ad hoc, reactive manner, often without standardized safeguards or capacity-building frameworks.

Power imbalances between public hospitals and international AI vendors also exacerbate governance weaknesses. With many AI tools being imported, donor-supported, or externally developed, hospitals have limited influence over system design, data use, or transparency mechanisms. This dynamic aligns with broader critiques of donor-driven digital development, where technology adoption is shaped more by external agendas than local needs or regulatory oversight. As a result, hospitals face “black box” systems whose inner workings are inaccessible, and data privacy responsibilities fall disproportionately on end-users who lack technical or legal leverage. These structural constraints amplify the mistrust, dependency, and workarounds observed among HR professionals, demonstrating that micro-level frustrations are rooted in macro-level governance gaps rather than individual resistance or lack of skills. Addressing GenAI adoption in Uganda’s public sector therefore requires systemic reforms including clearer national AI governance frameworks, strengthened institutional bargaining power with vendors, and locally grounded digital sovereignty approaches that prioritize contextual relevance and accountability.

The findings from Uganda also challenge several universalist assumptions embedded in TAM, UTAUT, and Sociotechnical Systems Theory. While these models prioritize perceived usefulness, ease of use, and system–task fit, our data show that institutional trust, cultural alignment, and digital neocolonialism concerns often override these classical predictors. Participants relied on manual oversight even when GenAI tools were efficient, suggesting that in low-resource public sectors, perceived reliability and cultural legitimacy may be stronger determinants of acceptance than usefulness alone. Likewise, the informal workarounds and improvisational practices documented in this study extend sociotechnical theory by illustrating how “fit” must be reconceptualized in contexts marked by resource scarcity, fragmented infrastructure, and entrenched bureaucratic norms. These insights point toward the need for a Contextualized Technology Acceptance Model for low-resource public health systems, where socio-cultural trust, institutional capacity, and local relevance shape adoption more strongly than traditional TAM/UTAUT constructs.

Participant-Generated Recommendations

Although participants were not explicitly asked to provide recommendations during the interviews, many offered practical suggestions spontaneously while reflecting on their lived experiences with AI-enabled recruitment in Uganda’s public hospitals. These participant-generated recommendations emerged organically from challenges encountered in day-to-day practice rather than from researcher prompting, thereby reflecting experiential insights grounded in operational realities.

Several participants emphasized the need for structured and context-specific training to enable HR professionals to effectively and ethically use AI-supported recruitment tools. As one participant noted, “Most of us are learning by trial and error. Formal training would help us understand both the strengths and the risks of these systems” (HR04). Others highlighted the importance of clear national and institutional guidelines to address concerns around fairness, accountability, and algorithmic bias, particularly in public-sector recruitment where transparency is critical. A recruitment officer remarked, “Without clear policies, it is difficult to explain or justify AI-based decisions to applicants” (HR09).

Participants also recommended maintaining strong human oversight in AI-assisted recruitment processes, cautioning against over-reliance on automated outputs. They stressed that AI tools should complement rather than replace professional judgment, especially when contextual factors such as local labor market conditions and applicant diversity are considered. Collectively, these recommendations underscore participants’ call for balanced AI integration that combines technological efficiency with ethical safeguards, human discretion, and institutional accountability.

Overall, our study confirms several trends in AI adoption such as gains in efficiency, emerging trust deficits, and ethical ambiguity, but uniquely contextualizes them within Uganda’s under-resourced health sector. The findings affirm that technological adoption is not just a technical process but also a human, ethical, and cultural one. They call for localized training, contextual adaptation of algorithms, co-design with users, and stronger regulatory frameworks to ensure equitable and transparent AI use in public recruitment.

Implications of the Study

This study offers significant practical and theoretical implications for both policy and practice in the adoption of generative AI in HR recruitment within resource-constrained public health settings. Practically, the findings highlight the urgent need for tailored digital literacy and AI-specific training programs for HR professionals to bridge the existing skills gap and facilitate smoother integration of GenAI tools. Policymakers and hospital administrators should prioritize investments in infrastructure that support reliable digital recruitment systems and ensure ongoing technical support to mitigate “technological anxiety” among staff. Moreover, addressing ethical concerns around algorithmic bias and transparency must be a core component of AI governance frameworks, particularly in sensitive sectors like healthcare where recruitment decisions have profound social consequences. The study also underscores the importance of hybrid human-AI recruitment models that balance efficiency gains with critical human judgement, suggesting that AI augmentation rather than full automation should be the operational goal. Theoretically, this research enriches the phenomenological understanding of technology adoption by revealing how cultural, contextual, and ethical dimensions shape HR professionals’ experiences in LMICs, thereby informing broader discourses on AI democratization and equity in digital health systems.

Limitations and Directions for Future Research

While this study offers rich phenomenological insights into HR professionals’ experiences with GenAI in public hospital recruitment, several limitations warrant further consideration. First, participants’ pre-existing attitudes toward AI may have influenced their interpretations and reflections. Although a neutral, non-leading interview guide was employed and bracketing techniques were applied during analysis to enhance reflexivity and reduce interpretive bias, the possibility of residual subjectivity cannot be entirely eliminated. Future research could complement qualitative accounts with mixed-method designs, incorporating survey-based measures of AI readiness, digital literacy, or technological anxiety to better control for attitudinal predispositions and strengthen analytical triangulation.

Second, the findings are context-specific, shaped by Uganda’s socio-cultural, institutional, and public-sector healthcare environment. While this contextual grounding enhances depth and authenticity, it may limit transferability to private-sector hospitals, other LMICs, or technologically advanced healthcare systems. Comparative cross-country or cross-sector studies would provide valuable insights into how regulatory frameworks, digital infrastructure maturity, and organizational cultures moderate AI adoption in recruitment. Such research would enhance theoretical generalizability and allow for contextual boundary conditions to be more clearly defined.

Third, although the purposive sample size was appropriate for phenomenological inquiry, it inherently limits statistical generalization. Future studies could expand the empirical scope to include a broader range of healthcare institutions such as private, faith-based, and specialized referral hospitals, and potentially adopt multi-site or national-level sampling frames. A sequential explanatory mixed-method design could further strengthen robustness by quantitatively testing patterns identified qualitatively.

Fourth, this study centered exclusively on HR professionals’ perspectives. Recruitment ecosystems are multi-actor systems involving candidates, senior hospital administrators, procurement committees, and IT personnel responsible for system deployment and oversight. Future research should adopt a multi-voiced approach to capture the interplay between strategic decision-making, technological implementation, and user experience. Such an approach would deepen understanding of relational dynamics and governance structures surrounding AI integration.

Fifth, this study did not systematically collect statistical data on vacancy volumes, applicant-to-vacancy ratios, or recruitment cycle duration across participating hospitals. Recruitment intensity may influence how HR professionals perceive the necessity, efficiency, or risks of AI-enabled tools. Future research could integrate organizational metrics to examine how structural workload pressures moderate attitudes toward AI adoption.

Finally, ethical concerns particularly regarding fairness, bias, transparency, and accountability, emerged prominently in participant narratives. Future research should move beyond perception studies to design, implement, and empirically evaluate AI governance frameworks tailored to healthcare recruitment in LMICs. Longitudinal designs would be especially valuable in assessing how trust, fairness perceptions, and regulatory compliance evolve as GenAI technologies mature and digital infrastructure improves. By adopting comparative, multi-method, and longitudinal approaches, future studies can build a more comprehensive and theoretically refined understanding of AI-enabled recruitment in resource-constrained healthcare systems.

Conclusion

This phenomenological study sheds light on the complex and nuanced experiences of HR professionals integrating generative AI tools into recruitment processes within Uganda’s public hospitals. The findings reveal a dynamic interplay between enthusiasm for AI-driven efficiency and persistent concerns around bias, transparency, and the erosion of human judgement. Crucially, the study situates these experiences within the broader socio-technical challenges of digital readiness and ethical governance in a resource-limited healthcare context. By foregrounding the lived realities of HR practitioners, this research contributes to a deeper understanding of how emerging AI technologies are negotiated on the ground in LMIC settings. It advocates for a cautious yet proactive approach to AI adoption, one that prioritizes capacity building, human oversight, and ethical transparency to harness the full potential of generative AI while safeguarding fairness and inclusivity in healthcare recruitment.

The study therefore offers a theoretical contribution by demonstrating that technology acceptance in low-resource public sectors cannot be adequately captured using conventional TAM or UTAUT pathways. Instead, factors such as institutional trust, cultural resonance, perceived foreignness of AI systems, and reliance on manual verification emerge as core determinants of acceptance. These findings lay the groundwork for a Contextualized Technology Acceptance Model for LMIC public institutions, emphasizing that successful GenAI integration requires attention to socio-cultural legitimacy and systemic constraints, not only usability or perceived benefits.