Digital Shifts in Human Resource Management; Artificial Intelligence Tools in Workforce Strategies in Government Agencies

1. Introduction

Artificial Intelligence (AI) has become a major driver of digital transformation across organizations, significantly reshaping Human Resource Management (HRM) practices. In modern HRM, AI is increasingly used for automating administrative tasks, improving decision-making, enhancing employee management systems, and supporting data-driven workforce strategies. These applications help organizations improve efficiency, accuracy, and responsiveness while reducing manual workload. In government agencies, however, the adoption of AI remains limited despite ongoing digital transformation efforts.

Based on the findings of this study, AI tools are generally least utilized across HR functions such as employee management systems, decision-making and analytics, cybersecurity, communication, and service delivery. While basic applications like data entry, attendance monitoring, and email management are present, advanced AI systems such as predictive analytics, AI dashboards, fraud detection, and automated document processing are still minimally implemented. This indicates that AI integration in public sector HRM is still at an early stage of development. Despite low utilization, workforce strategy results show that AI is perceived as moderately effective in several areas, including service delivery improvement, employee adaptability, decision-making support, and work process efficiency. However, effectiveness remains limited in cybersecurity integration, training and skill development, and full organizational readiness.

Challenges such as insufficient training, limited budget allocation, lack of clear policies, cybersecurity concerns, and organizational resistance continue to hinder full AI adoption. These findings suggest a gap between the potential of AI and its actual implementation in government agencies. Thus, this study aims to examine the level of AI utilization in HRM and evaluate its impact on workforce strategies in selected government agencies, highlighting the need for stronger institutional support and digital capability development.

2. Materials and Methods

The study employed a descriptive research design to systematically examine the current condition of digital shifts and the utilization of Artificial Intelligence (AI) tools in selected government agencies. The focus was on understanding how these tools affect HR functions like recruitment, training, and performance evaluation within the Metropolitan Manila Development Authority (MMDA) and the Department of Public Works and Highways (DPWH). The respondents consisted of 56 personnel, with 26 (46.43%) from MMDA and 30 (53.57%) from DPWH. Purposive sampling was used in selecting the administrators, while convenience sampling was applied to select the staff members. Demographic data showed that most respondents were female, single, belonged to the 25–34 age group, and had rendered 1 to 10 years of service.

Data were gathered using a validated, structured questionnaire designed to measure the respondents’ perceptions regarding the implementation, impact, and challenges of AI-enabled HR systems. The instrument utilized Likert-scale statements to assess the extent of AI utilization, workforce strategies, and operational barriers such as budget constraints and data privacy concerns. To ensure validity and reliability, the questionnaire underwent expert review by specialists in Human Resource Management, Public Administration, and Information Technology. Reliability testing through a pilot test yielded Cronbach’s Alpha coefficients of 0.82 for AI Tools Utilization and 0.86 for Workforce Strategies, indicating high internal consistency and strong instrument reliability.

The researchers followed systematic and ethical procedures by securing formal permissions from the government agencies and obtaining informed consent from the participants before distributing the instruments. Strict confidentiality and anonymity of the respondents were maintained throughout the process. The gathered data were tabulated and analyzed using specific statistical tools, including percentage for demographic distribution, and weighted mean for ranking the level of AI utilization and challenges. Furthermore, the Mann-Whitney U Test was used to determine significant differences between the assessments of administrators and staff, while the Pearson Product-Moment Correlation was applied to identify the relationship between AI utilization and workforce strategy effectiveness.

3. Results and Discussions

1. What are the Artificial Intelligence tools used in government agencies in terms of the following sub-variables?

Legend:

Range Scale Verbal Interpretation Symbol

5 4.20-5.00 Highly Utilize HU

4 3.40-4.19 Utilized U

3 2.60-3.39 Moderately Utilized MU

2 1.80-2.59 Least Utilized LU

1 1.00-1.79 Very Least Utilized VLU

This table presents the assessment of administrators and staff regarding the use of AI-driven automation tools, revealing that integration is currently more visible at the oversight level than in daily operational execution. The highest-ranked tool is the "Use of AI for automated data entry and record keeping" with a composite weighted mean of 2.47, verbally interpreted as Less Utilized (LU). This is closely followed in second rank by the "Use of AI for report generation and analytics" with a composite mean of 2.24 (LU). These rankings indicate that while automation is beginning to touch administrative workflows, it remains in its foundational stages, primarily focused on routine data management and analytics tasks.

A clear gap exists between the two respondent groups, as administrators evaluate tools like automated data entry (3.00) and chatbots (2.70) as Moderately Utilized (MU), while staff members perceive these exact same tools to be either Less Utilized (LU) or Very Less Utilized (VLU). In fact, chatbot inquiries scored a low 1.76 among staff, while the "Use of AI for automated document processing" emerged as the lowest-ranked tool overall (Rank 5) with a composite weighted mean of 2.09 (LU). Ultimately, the overall composite weighted mean sits at 2.24, falling firmly under the "Less Utilized" bracket and highlighting that operational staff experience minimal daily interaction with AI workflows.

This low-to-moderate baseline of AI adoption aligns with recent findings on digital transformations within organizational frameworks. The realities shown in the data echo the readiness challenges noted by Campued et al. (2023), who emphasized that evaluating the awareness and readiness of agencies is a critical prerequisite to successful AI adoption. Furthermore, the emphasis on automated data entry and report analytics as top-ranked tools reflects trends identified by Rosario et al. (2025), where the digital transformation of cloud data transactions utilizes AI-based authenticators to streamline quality management practices. While local human resource practices are actively redefining the future of work through emerging tools (Gines, 2025; The SDP Team, 2025), the data suggests that organizations must address training and infrastructure gaps to shift from localized administrative tasks toward optimized workforce automation as envisioned by Quadri et al. (2024).

As shown in Table 2, the assessment of administrators and staff regarding the use of AI as a decision-support system illustrates that administrative leaders perceive a higher level of integration than operational personnel. The highest-ranked capability is the "Use of AI to analyze large volumes of agency data" with a composite weighted mean of 2.38, verbally interpreted as Less Utilized (LU). This is closely followed in second rank by the "Use of AI predictive analytics for planning and forecasting" with a composite mean of 2.36 (LU). These results demonstrate that while the organization recognizes the value of data-heavy processing and long-term projection tools, their current application remains limited and localized at the higher tiers of management.

A noticeable discrepancy is evident between the two groups of respondents, as administrators view almost all indicators—such as analyzing large volumes of data (2.80) and AI-generated recommendations (2.73)—as Moderately Utilized (MU). On the other hand, the staff consistently perceives these decision-support features as Less Utilized (LU) or Very Less Utilized (VLU), with the use of AI dashboards for management decisions receiving the lowest staff score of 1.76. Consequently, this dashboard feature emerged as the lowest-ranked indicator overall (Rank 5) with a composite mean of 2.18 (LU). This pattern culminates in an overall composite weighted mean of 2.28 (LU), reinforcing the finding that frontline staff are largely disconnected from the AI metrics and dashboards shaping institutional directions.

This restricted baseline of decision-support adoption matches broader literature regarding data-driven frameworks. The gap between administrative oversight and operational usage aligns with the observations of the Bangko Sentral ng Pilipinas (2025), which noted that emerging technologies in human resource management offer substantial opportunities for strategic forecasting but also present ongoing implementation challenges. As organizations navigate these early stages of analytical deployment, establishing clear lines of data transparency and training remains essential for turning localized administrative oversight into meaningful institutional choices.

The data presented in Table 3 details how administrators and staff evaluate the deployment of AI within employee management systems, illustrating an overall restricted level of integration. The top-performing indicator is the "Use of AI for attendance and timekeeping" which secured a composite weighted mean of 2.14, verbally interpreted as Less Utilized (LU). Following in second place is the "Use of AI for task assignment and monitoring" with a composite mean of 2.00 (LU). These results indicate that automated employee management is primarily confined to standard administrative tracking and operational oversight, rather than driving core personnel developmental tasks.

A stark division emerges when comparing the evaluations of the two respondent groups across all criteria. Administrators rate the use of AI for attendance and timekeeping at a moderate 2.60 (MU), whereas staff members overwhelmingly perceive every single indicator as Very Less Utilized (VLU). This widespread disconnect is highlighted by the low staff scores for performance evaluations (1.68) and attendance tracking (1.73). Consequently, the "Use of AI in recruitment and applicant screening" ranks lowest overall (Rank 4) with a composite mean of 1.95 (LU). This pattern culminates in an overall composite weighted mean of 2.01 (LU), proving that operational employees experience almost no direct, meaningful interaction with intelligent talent management tools.

This limited implementation across recruitment and evaluation systems strongly aligns with specific localized studies on human resource technologies. The low ranking of automated hiring systems directly reflects the findings of Cacatian et al. (2025), who explored the integration of AI in human resource recruitment in Metro Manila and noted that widespread adoption faces structural roadblocks despite initial innovative efforts. As organizations navigate these early stages of adoption, bridging the gap between administrative intent and operational reality remains essential for transforming basic automated systems into truly collaborative human resource management tools.

Looking at the data in this table, the assessment of administrators and staff concerning the deployment of AI in security and monitoring features reflects a heavily foundational stage of adoption. Tied for the highest rank are the "Use of AI for detecting security threats" and the "Use of AI for data privacy protection," both securing a composite weighted mean of 2.27, which translates to a verbal interpretation of Less Utilized (LU). These are immediately followed by the "Use of AI for monitoring network activities" with a composite mean of 2.24 (LU). These top scores indicate that the organization places its primary, albeit limited, technological focus on defensive operations and protecting information assets rather than inner system maintenance.

A familiar perceptual divide persists between the two respondent classes across all security indicators. Administrators evaluate threat detection (2.60), network monitoring (2.60), and data privacy protection (2.67) as Moderately Utilized (MU), signaling a confidence in the safeguards established at the infrastructure level. Conversely, the operational staff perceives these protections as Less Utilized (LU) or even Very Less Utilized (VLU), particularly grading fraud detection at a low 1.78. This pushes the "Use of AI for fraud detection" to the bottom overall position (Rank 4) with a composite mean of 2.12 (LU), shaping an overall composite weighted mean of 2.21 (LU) and proving that frontline staff see minimal daily integration of automated security protocols.

This low-to-moderate baseline of AI security implementation maps onto broader discussions regarding institutional trust and data safety. The emphasis on data privacy protection as a top administrative priority, contrasted with lower operational execution, closely mirrors the concerns highlighted by Meshram (2023), who noted that integrating artificial intelligence into organizational ecosystems requires a highly balanced approach to secure personal data while training users to manage systemic vulnerabilities. Consequently, until the general workforce is actively engaged in these protective workflows, the organization’s digital defense mechanisms will remain confined to an oversight utility rather than serving as a pervasive, fully integrated security infrastructure.

This table details how both administrators and staff evaluate the deployment of AI within communication and service platforms, indicating that these digital tools are still in the early stages of practical implementation. The highest-ranked indicator is the "Use of AI in email and message management" which achieved a composite weighted mean of 2.31, verbally interpreted as Less Utilized (LU). This is closely followed in second rank by the "Use of AI for online service processing" with a composite mean of 2.25 (LU). These findings demonstrate that while the organization has begun testing the waters with automated sorting and basic transaction processing, these digital channels are not yet deeply embedded into the institutional communication structure.

A sharp divergence in viewpoint remains obvious between the two groups of respondents across all service metrics. Administrators consistently view these capabilities with higher confidence, rating online service processing at 2.73 (MU) and message management at 2.67 (MU), indicating a belief that these channels are moderately functional. On the other side of the spectrum, staff members perceive these exact same mechanisms as Less Utilized (LU) or Very Less Utilized (VLU), giving public information dissemination their lowest evaluation at 1.75. As a result, the "Use of AI to improve response time to clients" rests at the very bottom of the list (Rank 5) with a composite mean of 2.11 (LU), feeding into an overall composite weighted mean of 2.20 (LU) and proving that operational personnel see little to no day-to-day impact on client interaction efficiency.

This restricted utilization of client-facing AI tools reflects current public sector discussions on tech-driven transparency and public communication. The low baseline of staff involvement and public dissemination tracking aligns with observations by Lebonick (2024), who emphasized that integrating AI into public operations requires addressing major hurdles in infrastructure and open governance transparency before these tools can effectively serve the general public. Without expanding these digital communication channels and training frontline personnel to leverage them, the technology will continue to serve as an administrative convenience rather than an active tool for improving institutional response times and community outreach.

2. How do the administrators and staff assess the workforce strategies in using AI tools in government agencies in terms of the following?

Legend:

Range Scale Verbal Interpretation Symbol

5 4.20-5.00 Highly Evident HE

4 3.40-4.19 Evident E

3 2.60-3.39 Moderately Evident ME

2 1.80-2.59 Least Evident LE

1 1.00-1.79 Very Least Evident VLE

Based on the results in Table 6, the scores for work process efficiency show a different trend from the other data, as the staff actually gave slightly higher marks for efficiency than the administrators did. Standing out at the top position is the statement that "AI tools help speed up daily work tasks," which achieved a composite weighted mean of 2.61, verbally interpreted as Moderately Efficient (ME). This is followed in the second rank by the idea that "AI improves overall productivity" with a composite score of 2.57, verbally interpreted as Less Efficient (LE). These numbers indicate that while both groups feel that AI helps speed up their daily tasks, it has not yet scaled up enough to deeply transform the overall productivity of the organization.

Unlike the previous tables where administrators were more optimistic, this data shows that the staff feels a stronger practical benefit in their actual routines, rating overall efficiency as Moderately Efficient at 2.60 (ME). Meanwhile, administrators gave a lower overall mean of 2.54, interpreting it as Less Efficient (LE). This gap is most visible in the indicator "AI improves accuracy of work outputs," which dropped to the bottom position (Rank 5) with a composite mean of 2.27 (LE) because staff gave it a strict 2.58 (LE) evaluation. This pattern shows that while the technology successfully cuts down the time needed to finish tasks, both groups still have doubts about the precision and reliability of the automated work.

This focus on operational speed paired with concerns over output accuracy mirrors ongoing discussions within regional technology adoption studies. The finding that AI accelerates tasks but requires careful quality verification corresponds with research by Sishi et al. (2025), who investigated the impacts of artificial intelligence on human resource management practices and highlighted that automated efficiency gains must be balanced against systemic implementation limits and oversight demands. Ultimately, the data implies that to move from localized time-saving benefits toward comprehensive operational excellence, the organization must refine its current software configurations to ensure that automated speed does not come at the expense of data integrity and work quality.

As shown by the ratings in Table 7, the feedback on skill development and training reveals a very similar view from both sides, indicating that current learning efforts are still quite limited. Standing out at the top position is the statement that "Training programs improve AI knowledge," which achieved a composite weighted mean of 2.59, verbally interpreted as Less Efficient (LE). This is followed in the second rank by "Employees receive training on AI tools" with a composite score of 2.58 (LE). These numbers suggest that while the initial training sessions are somewhat effective at building basic technical awareness, the current programs are not deep enough or frequent enough to make a major impact on the workplace.

The breakdown between the two groups shows that administrators are slightly more positive about the early stages of training, rating both knowledge improvement and tool training as Moderately Efficient at 2.67 (ME). On the other hand, the staff rated every single indicator in the table as Less Efficient (LE). Tied for the lowest positions (Rank 4) are the items "Employees are encouraged to attend AI seminars" and "Training programs help employees adapt to AI," both finishing with a composite mean of 2.51 (LE). This pattern reveals a clear organizational weakness, showing that simply introducing a tool is not enough if employees feel they do not have the ongoing guidance or encouragement to actually adapt to the technology.

This clear need for stronger training and educational support mirrors themes found in broader human resource studies. The low evaluation of AI adaptation and seminar encouragement directly links to the findings of Vadivel and Rasswanth (2024), who noted in their study on artificial intelligence in human resource management that proper training frameworks are absolutely vital for successful workplace integration. Without expanding these learning programs and building a continuous training culture, the organization will likely struggle to get its staff past the early learning curve and fail to unlock the full collaborative potential of automated systems.

The results highlighted in Table 8 focus on employee adaptability and acceptance, revealing an interesting trend where the staff actually shows a higher level of acceptance toward these systems than the administrators do. Taking the top spot overall is the item "Use of AI dashboards for management decisions," which earned a composite weighted mean of 2.79, verbally interpreted as Moderately Efficient (ME). This is followed in second place by the "Use of AI to analyze large volumes of agency data" with a composite score of 2.64 (ME). These numbers show that the workforce is open and ready to accept big-data visual layouts and tracking tools within their daily operations.

Looking closely at the two groups, the staff members consistently expressed higher adaptability, giving an overall weighted mean of 2.76 (ME) and scoring every single category as Moderately Efficient. In contrast, the administrators were much more conservative, rating almost all indicators as Less Efficient (LE) at 2.47, with only the management dashboard escaping that low grade at 2.83 (ME). Because of this administrative hesitation, the "Use of AI-generated recommendations for decision making" finished at the very bottom of the rankings (Rank 5) with a composite mean of 2.59 (LE). This gap suggests that while the operational staff is eager and willing to adapt to AI features, leadership remains cautious about relying on automated suggestions for critical decisions.

This dynamic of high staff openness versus cautious leadership aligns with specific regional findings on technology acceptance. The willingness of employees to adapt to automated data platforms corresponds with a study by Maraver and Villacruel (2024), who explored the acceptance of artificial intelligence in selected local industries and emphasized that worker readiness is highly dependent on how clearly the tools fit into their current operational habits. Because the frontline staff is already comfortable with the software, the main challenge for the organization is no longer overcoming employee resistance, but rather building leadership confidence so that the entire workplace can move forward together.

The data listed in Table 9 looks at service delivery improvement, showing a good amount of agreement from both groups that AI really helps level up public service. Taking the top spot overall is the indicator "AI reduces waiting time for clients," which got a composite weighted mean of 2.76, interpreted as Moderately Efficient (ME). Right behind it in second place is "AI improves the quality of public service" with a composite score of 2.74 (ME). These numbers show that both the bosses and the workers clearly see how automation helps cut down delays and makes serving the public much smoother.

Looking at the two groups, the staff members are actually a bit more excited about these improvements, finishing with a higher overall mean of 2.77 (ME) compared to the administrators' 2.67 (ME). The staff felt happiest about troubleshooting, scoring faster problem resolution at a high 2.85 (ME). On the other side, "AI makes government services more accessible" landed at the very bottom of the list (Rank 5) with a composite mean of 2.66 (ME). This tells us that while the software is great at speeding up things for clients who are already using the system, there is still work to do in using AI to reach more people.

This link between faster processing and better service matches recent local studies on technology in the workplace. The finding that AI cuts down waiting times and keeps clients happy directly supports the work of Cusipag et al. (2024), who studied AI implementation in Philippine institutions and noted that automated systems drastically speed up response times and boost service quality. Since the basic benefits of speed and quality are already working well, the organization's next step is to make these digital tools easier to reach for the wider community.

The details gathered in Table 10 show how the organization handles governance and institutional support for AI, revealing that while plans are moving forward, foundational rules are lagging behind. Taking the top spot overall is "Management supports AI implementation," which earned a composite weighted mean of 2.78, verbally interpreted as Moderately Efficient (ME). This is followed closely in second place by "The agency has plans for AI development" with a composite score of 2.74 (ME). These numbers suggest that the leadership team is highly eager and forward-looking when it comes to adopting new software frameworks.

Looking at the breakdown between the two respondent groups, the administrators are much more confident about the organization's backend backing, posting an overall mean of 2.75 (ME). Meanwhile, the staff felt less secure, giving an overall mean of 2.58, which falls under Less Efficient (LE). This gap becomes most striking at the bottom of the list, where "The agency has clear policies on AI usage" ranks dead last (Rank 5) with a composite mean of 2.53 (LE). This ranking shows that while the company is actively pushing and planning for technology upgrades, it hasn't yet laid down concrete, everyday rules to guide the staff on how to safely use these tools.

This combination of strong management backing alongside weak formal rules mirrors current regional discussions on digital transformation. The reality that institutional planning is outstripping actual rule-making aligns with findings by Almazan (2024), who studied public sector technology management and pointed out that rapid automation often creates a policy gap where tools are deployed before clear guidelines are established. To fix this bottleneck and ease the staff's hesitation, the organization needs to pivot from general planning to writing clear, step-by-step usage policies that protect data and give employees clear boundaries.

3. Is there a significant difference between the assessment of administrators and staff on the workforce strategies including the use of AI tools?

Level of Significance @ 0.05

Table 11 presents the inferential analysis comparing the assessments of administrators and staff on the various sub-variables of artificial intelligence (AI) systems and workforce strategies. An independent sample t-test assuming unequal variances was utilized at a 0.05 level of significance to determine whether a statistical disconnect exists between leadership and frontline employees. The data reveals a clear and statistically significant difference in how the two groups perceive the actual deployment of specific workplace technologies. Significant differences were confirmed for Automation Tools (t = -2.3924), Decision-Support Systems (t = -2.4182), Cyber-Security and Monitoring (t = -2.1834), and Communication and Service Delivery Tools (t = -2.3958). Because the absolute t-values for these four areas surpass their respective critical values, the null hypothesis of no significant difference is rejected. This pattern stems directly from a perceptual gap where administrators consistently estimated a higher baseline of utilization, while frontline staff reported much lower engagement. This disparity strongly implies an administrative disconnect, meaning leadership feels these systems are active and available, but the actual day-to-day operators either lack hands-on access or have not yet integrated them into their routine assignments.

Conversely, the data demonstrates that both respondent groups are statistically aligned regarding the softer, structural dimensions of the technology rollout. No significant differences were found for Employee Management Systems (t = -1.9064), Work Process Efficiency (t = 0.1401), Skill Development and Training (t = -0.2260), Employee Adaptability and Acceptance (t = 0.5438), Service Delivery Improvement (t = 0.2703), and Organizational Support and Readiness (t = 0.4243). For these six indicators, the absolute t-values fall well below the critical values, meaning the analysis fails to reject the null hypothesis. This shared outlook indicates that both administrators and staff experience the exact same institutional constraints on the ground. For instance, both groups gave lower, parallel marks to Skill Development and Training and Work Process Efficiency, showing a mutual realization that training programs are currently insufficient and that everyday operational workflows have not yet been fully optimized by the new software.

On the whole, the t-test confirms that while administrators and staff share an identical, realistic view of the organization’s strategic struggles, they remain sharply divided on how much the technological infrastructure is actually being used on the ground. To resolve this issue, administration must transition away from macro-level system updates and instead establish transparent feedback loops, practical training protocols, and hands-on onboarding windows that actively empower staff to adopt these tools in their daily routines.

4. Conclusion

Based on the statistical findings of the study, the following conclusions are drawn regarding the use of AI tools and workforce strategies in the agency:

1. Low Baseline of AI Utilization: The overall deployment of AI systems in the organization remains at a "Less Utilized" baseline. Key technologies such as automated HR tools and decision-support systems have not yet been fully integrated into the daily administrative habits and tasks of the office.

2. Perceptual Disconnect Between Admin and Staff: There is a clear statistical gap between leadership and frontline employees regarding technology use. Administrators feel that AI tools are active and moderately utilized, whereas the actual staff on the ground report that they rarely interact with or experience these platforms.

3. Shared Training and Workflow Bottlenecks: Both respondent groups are completely aligned when it comes to institutional challenges. Administrators and staff mutually agree that current technology training programs are lacking and that daily operational workflows have not yet been optimized to make their workloads easier.

4. Execution Gap in the Digital Transition: Ultimately, while the agency demonstrates a strong willingness and institutional readiness to adapt to digital shifts, its transformation is heavily slowed down by the lack of hands-on platform onboarding and a lack of clear operational guidelines for everyday users.

5. Recommendations

Based on the conclusions drawn, the following practical interventions are recommended to improve the integration of AI within the agency's human resource strategies:

1. Provide Practical Hands-On Training: Management must move away from theoretical seminars and focus on continuous, step-by-step technical workshops. This will ensure that frontline staff gain actual familiarity and competence in operating automation tools and decision-support databases.

2. Create a Feedback Channel for Staff: Leadership should establish a clear communication line or survey system where frontline employees can directly report technical difficulties, operational anxieties, and specific reasons why they are not using underutilized programs like employee management software.

3. Update Internal Workplace Rules: The agency should revise its standard operating procedures and job descriptions to formally include and recognize the use of AI tools. This will clear up role confusion and officially push employees to make daily workflows faster and more efficient.

4. Monitor Employee Progress Regularly: Project supervisors should implement a quarterly tracking framework to evaluate the staff's technical adjustments and skills over time. This ensures that the agency's tech budget is continually supported by a workforce that is well-trained and ready to use the to