Construction sites were dynamic environments where tools frequently moved between locations and workers, making accurate tracking and inventory management a persistent challenge. Despite advancements in construction practices, many firms still relied on manual methods, such as paper logs or spreadsheets, which were prone to human error, lacked real-time visibility, and often resulted in tool misplacement or theft. These inefficiencies contributed not only to operational delays but also to increased project expenses and heightened safety risks (Hasan et al., 2024; Ahmadi, 2024).

To address these challenges, the construction sector explored the use of emerging technologies in automation and data analytics. Radio Frequency Identification (RFID) systems were widely adopted to enhance asset visibility and monitoring accuracy. Bhadeshiya et al. (2021) discussed the evolution of RFID applications in construction, highlighting their effectiveness in tracking and monitoring materials and resources throughout various project phases. Similarly, real-time locating systems (RTLS) were implemented to improve the tracking of equipment and personnel on construction sites. Szkypaczak and Krygier, (2017) presented the use of Bluetooth-based RTLS for asset tracking, demonstrating its potential in enhancing productivity through lean management strategies.

In response to these challenges, the integration of Artificial Intelligence (AI) and Machine Learning (ML) gained traction for advancing tool tracking systems. AI-powered systems, particularly those incorporating image recognition and real-time data processing, could reduce the need for manual intervention and enhance accuracy in identifying tools and logging usage patterns (Darko et al. 2020). Similarly, Wang et al. (2023) emphasized that AI-based vision systems were particularly advantageous in unstructured environments like construction sites, where traditional tracking systems might fail. These intelligent systems improved tool identification by analyzing images with minimal supervision, adapting well to cluttered and variable workspaces.

Moreover, predictive analytics supported by AI was leveraged to forecast equipment needs and streamline material logistics. Ben-Daya et al. (2019) found that predictive models could significantly enhance supply chain coordination, while Schwabe et al. (2021) reported that AI-driven demand forecasting could optimize tool allocation, ultimately minimizing excess inventory and ensuring that necessary equipment was always available.

Effective communication also played a vital role in real-time tool tracking and management. SMS-based notification systems were increasingly integrated into construction site monitoring platforms, especially in regions where internet reliability was limited. Supporting this, Li et al. (2022) found that SMS systems significantly improved site coordination, allowing managers to receive instant alerts and make timely decisions even in offline or remote locations. Despite the promise of AI-based systems, the success of their deployment in construction was not solely determined by technical capability. User adaptability and ease of integration into existing workflows were equally important. Park and Lee, (2020) stressed that human factors, such as system intuitiveness and responsiveness to worker input, were crucial to sustained usage. Ashrafian et al. (2023) further emphasized the need for AI solutions that were designed with user experience in mind. Their study revealed that systems which aligned closely with the needs and skills of construction personnel tended to achieve higher rates of acceptance and operational integration.

Recognizing these gaps and opportunities, this study focused on the development and evaluation of the ToolGuard AI, a system designed to integrate real-time monitoring, automated tool identification, and SMS-based notifications tailored to the specific demands of construction environments. Given the high demand for accurate and responsive tool monitoring in construction, this research aimed to evaluate the ToolGuard AI across several key dimensions: real-time monitoring capability, tool identification accuracy, SMS notification reliability, and system acceptability in construction settings. Prior studies, including Bolpagni et al. (2021), emphasized that successful adoption of AI technologies in construction was significantly influenced by system functionality and performance, which directly affected acceptance levels among users. Therefore, this study not only assessed the technical performance of the ToolGuard AI but also examined its acceptability, ensuring it addressed the unique challenges of tool management in dynamic construction environments. In alignment with the Sustainable Development Goals (SDGs), particularly SDG 9 (Industry, Innovation, and Infrastructure) and SDG 12 (Responsible Consumption and Production), this study promoted the use of intelligent systems to enhance operational efficiency and accountability in construction. By reducing tool loss, optimizing inventory practices, and improving workflow coordination, the ToolGuard AI contributed to more responsible resource management. 

The main objective of the study was to develop a ToolGuard AI for real-time tool monitoring and inventory management, specifically focused on technical features, real-time performance, sensitivity in tool identification, SMS notification performance, and acceptability.

This study used the Design and Development Research (DDR) method to create a technological solution for tool monitoring and inventory control through real-time feedback and automated alerts. Evaluation was conducted with the participation of 30 respondents, consisting of two site managers, one tool keeper, two safety officers, three construction workers, three civil engineers, five electrical engineers, six IT professionals, and eight academic representatives. These evaluators were selected based on their expertise and availability during the evaluation period.

The development of the ToolGuard AI, a monitoring and inventory management system for the construction industry adopted a developmental research approach, which allowed the system to be continuously designed, tested, and refined to meet specific functional, performance, and user experience requirements. This approach ensured that the ToolGuard AI evolved based on feedback and testing, making it adaptable and responsive to real-world challenges faced on construction sites. By utilizing this iterative process, the system was developed to not only meet the immediate needs of the construction industry but also anticipate future technological advancements.

The evaluation of the ToolGuard AI was conducted through a combination of technical assessments, experimental testing, and evaluator assessments to comprehensively measure the systems performance and acceptability in construction environments. The evaluation focused on four key areas: real-time performance, tool identification sensitivity, SMS notification performance, and system acceptability. To evaluate real-time performance, tool transactions were simulated by placing and removing tools within the detection area. The response time, measured from tool movement to system update, was recorded using a stopwatch. This process was conducted over ten trials to ensure data consistency, and the average response time was computed to assess the systems responsiveness.

For tool identification sensitivity, the system was tested using a standard set of construction tools positioned at various angles and placements. The systems ability to correctly identify tools was documented, and sensitivity was calculated based on the ratio of true positives to the sum of true positives and false negatives, repeated across ten trials.

The SMS notification performance was evaluated by triggering tool status changes and measuring both the success rate and notification delay. The delay was defined as the time elapsed between the systems detection of a tool event and the receipt of the SMS message on a cellphone, recorded using a stopwatch. This manual verification reflected real-world end-user experience. The process was repeated under varying network conditions and across 10 trials to ensure reliable results.

In addition to manual observations, system-generated logs, including the inventory log and analytical system report, were accessed on a laptop to support the analysis of detection accuracy, response time, and notification performance.

The ToolGuard AI was technically composed of a Raspberry Pi 4 Model B as the main processing unit, integrated with a USB camera for tool detection, an Arduino Nano for GSM control, and a SIM800L GSM module for SMS communication. The software framework was developed using Python programming language, incorporating OpenCV for image processing and TensorFlow Lite for AI-based object recognition. The system was enclosed in a protective casing, with an LED indicator, relay module, and power regulation components to ensure safe and stable operation. The overall configuration of the device allowed it to perform autonomous detection, message transmission, and inventory recording with minimal user intervention, making it suitable for integration into active jobsite workflows. The system architecture allowed for seamless coordination between the hardware and software components, enabling automatic tool identification, real-time data logging, and SMS-based status reporting.

The system architecture of the ToolGuard AI is designed to optimize the interaction between the hardware components and the software framework. At the heart of the system, the Raspberry Pi receives real-time data from the USB camera. The camera captures images of the tools, which are then processed by the TensorFlow Lite model running on the Raspberry Pi. Once a tool is detected, the system updates the inventory and triggers notifications to the designated personnel. The communication between the Raspberry Pi and the Arduino Nano ensures that the data flows seamlessly, allowing the SIM800L GSM module to send out SMS alerts to users when needed.

According to Kim et al. (2022), the architecture ensures real-time performance, with minimal delay between tool detection and system updates. From the moment a tool is placed in the detection area, the system processes the image, identifies the tool, and sends a notification - all within less than one second. This response time is critical in construction environments, where prompt updates are necessary to maintain accurate tool inventories and keep operations running smoothly.

The findings disclosed that the ToolGuard AI was highly functional and dependable for real-time tool monitoring and inventory management in construction environments. The device responded consistently when tool movement was detected and automatically updated the inventory report through its programmed system. This reliable performance was a result of its integrated design and well-configured software and hardware components.

The results demonstrated that the ToolGuard AI achieved 100% response rates for the Hard Hat and Hand Drill, indicating the systems capacity to reliably detect these tools without delays in all trials. This performance illustrated the systems high sensitivity and operational speed, especially for tools with clearly defined visual features. Zhao et al. (2018) emphasized that object recognition systems perform most accurately with high-contrast, geometrically distinct items, which likely contributed to these consistent detection results. The system demonstrated real-time performance by successfully updating tool status during actual use. In the conducted 10 trials per tool, the system achieved response rates ranging from 80% to 100%. This confirmed that the ToolGuard AI was capable of immediate system updates upon tool activity, validating its efficiency in real-time operation.

The results demonstrated that the ToolGuard AI performed exceptionally well across most tools, with 100% log generation rates for the Hard Hat and Hand Drill, indicating flawless performance in capturing tool updates. These tools were successfully tracked in all ten trials, showcasing the systems ability to reliably detect and log tool activity. The Pickaxe, Hand Saw, and Welding Inverter followed closely with a 90% response rate, successfully generating logs in nine out of ten trials, suggesting that the system performed well but had occasional minor delays.

For tools such as the Angle Grinder, Metal Float, and Masonry Float, the ToolGuard AI showed a response rate of 80%, meaning eight out of ten trials produced on-time detections. While the system performed well, it experienced two instances of delayed log generation per trial, likely due to challenges in object recognition or tool placement. The Spade, with the lowest response rate of 70%, showed three delayed responses out of ten trials, highlighting areas for optimization.

The Spades lower detection rate is particularly notable, as it points to a critical area for improvement. Tools with irregular shapes or thin profiles, such as the Spade, may pose difficulties for the ToolGuard AIs detection model. As Szeliski, (2020) pointed out, computer vision systems often struggle to recognize objects with non-standard shapes, especially in dynamic or cluttered environments. This detection challenge suggests that the ToolGuard AI could benefit from integrating deep learning algorithms to handle complex shapes and orientations. These algorithms are more capable of adapting to various tool profiles, especially when encountering complex shapes or orientations. Additionally, incorporating adaptive lighting could help improve detection in environments with fluctuating light conditions, further enhancing tool recognition and reducing delays in logging. Similarly, tools such as the Angle Grinder, Metal Float, and Masonry Float exhibited an 80% response rate, which also suggests that the system could benefit from improvements in its ability to differentiate between tools with similar shapes. These tools visual overlap may have caused confusion during detection, leading to delays. The need for refined object recognition algorithms here is clear. Improving the ability to distinguish between similar tools would reduce detection delays and improve the overall performance of the system. Expanding the training dataset to include more detailed features for each tool or adjusting camera calibration could help minimize errors caused by tool proximity, improving both responsiveness and accuracy.

Despite these minor delays, the ToolGuard AI demonstrated that it consistently maintained on-time log generation in at least 70% of trials, proving its reliability for real-time tool monitoring on construction sites. This performance aligns with industry standards for embedded systems, where a response rate of 80% or higher is considered acceptable for deployment in dynamic environments like construction and manufacturing (Kim et al., 2021). The overall performance of the ToolGuard AI not only meets but exceeds these benchmarks, confirming its reliability and readiness for field applications.

While the Spades detection rate emphasized the need for ongoing optimization, especially for tools with non-standard shapes, human intervention plays an important role in counter-checking the systems performance. When the system experiences delayed responses or fails to detect certain tools, manual inventory checks can be implemented. For example, construction supervisors can be alerted when a tools status is not updated in real time, allowing them to verify the tools activity manually. Human intervention in these situations ensures that inventory logs remain accurate and that no tool is unaccounted for, even if the systems detection is delayed or incorrect.

By incorporating these manual verification processes, the system can operate seamlessly while also minimizing human error. Although the systems automatic tool tracking reduces the need for manual oversight, human intervention continues to be essential in ensuring that inventory management is maintained at optimal standards, particularly when tools with irregular shapes or challenging orientations are involved. This ensures that the system remains functional and accurate while continuing to improve through machine learning and other optimizations.

The Spades lower detection rate highlights the importance of continuous system optimization. As Szeliski, (2020) discusses, computer vision systems must continuously adapt to handle objects with varying shapes and orientations. Integrating more advanced deep learning techniques or fine-tuning the existing OpenCV-based detection model would significantly improve the ToolGuard AIs real-time responsiveness and detection accuracy. Implementing adaptive learning mechanisms, where the system learns from new tools and environments, would help minimize detection errors and optimize real-time performance, ensuring the system adapts to future challenges and tool varieties.

Overall, the results confirmed the ToolGuard AIs technical reliability and its ability to autonomously track tool movements and generate real-time logs. The system performed excellently for tools like the Hard Hat and Hand Drill, with 100% log generation across all trials. The Spades detection challenges, however, pointed to areas for further refinement. The ToolGuard AIs ability to constantly evolve through deep learning and adaptive algorithms will ensure its continued success in real-world applications. With ongoing improvements, the ToolGuard AI will remain a powerful, scalable solution for automated tool tracking in the construction industry, improving accountability, traceability, and efficiency (Ghezzi & Cavallo, 2018).

The results showed that the ToolGuard AI achieved 100% identification accuracy for tools such as the Hard Hat, Pickaxe, Hand Drill, and Hand Saw, identifying these tools correctly in all ten trials. This performance underscores the systems high sensitivity and reliability when recognizing tools with clearly defined visual features. These tools, which have straightforward shapes and characteristics, confirmed that the ToolGuard AIs detection algorithms are well-suited for tools with easily distinguishable visual cues. This is consistent with the findings of Zhao et al. (2018), who noted that computer vision systems perform best with high-contrast, geometrically distinct objects in clear viewing conditions.

The Angle Grinder and Welding Inverter demonstrated a 90% identification rate, with successful identification in nine out of ten trials. Although the system performed well, minor inconsistencies in recognition were observed, particularly under different orientations or lighting conditions. These inconsistencies indicate areas where the systems performance could be further enhanced, particularly in dynamic environments where lighting and tool placement can vary frequently. The Metal Float and Masonry Float recorded an 80% identification rate, with two instances of misidentification for each tool. These results highlight that while the ToolGuard AI generally performed well, it faced challenges when distinguishing between tools with similar visual characteristics. Refining the systems recognition model to better differentiate between similar tools would improve its performance in such cases. Chen et al. (2019) similarly observed that real-time object recognition systems often face difficulties distinguishing between visually similar items, especially in environments with background clutter or variable lighting.

The Spade had the lowest identification rate at 70%, with three misidentifications across ten trials. This suggests that tools with irregular shapes or thin profiles, like the Spade, are more difficult for machine vision systems to detect, particularly in real-time settings where the tools may be in motion or partially obstructed. The Spades performance pointed to a critical need for optimization in recognizing tools that deviate from typical shapes.

Human intervention played a significant role when the system faced challenges in identifying tools like the Spade. In instances where the system failed to detect a tool or generated a delayed or incorrect response, personnel could be alerted to manually verify the tools status. This verification process typically involved visual confirmation or manual checking to ensure the tools presence was properly recorded in the system. Such manual oversight served as a critical safeguard, especially for tools that were irregularly shaped, thin, or positioned in ways that obscured their features. This approach helped maintain the accuracy and reliability of the monitoring process. Lee and Park, (2021) emphasized that even in AI-integrated systems, human supervision remains essential for validating uncertain outputs, especially in environments where visual ambiguity, interference, or sensor limitations may impair detection accuracy. Their findings support the continued relevance of human verification as a complementary component in semi-automated tool tracking systems like the ToolGuard AI.

As the system continues to evolve, the integration of deep learning algorithms, such as Convolutional Neural Networks (CNNs), could significantly enhance the detection capabilities of the ToolGuard AI. These algorithms excel at identifying complex patterns, making them particularly effective for detecting tools with non-standard shapes. CNNs would allow the system to adapt better to varying tool profiles, improving its overall accuracy. Additionally, incorporating adaptive lighting techniques would improve the systems ability to handle poor lighting conditions, which often cause detection failures, especially when tools are obscured by shadows or nearby objects.

For tools like the Angle Grinder, Metal Float, and Masonry Float, which share similar features, the ToolGuard AI could enhance its recognition by refining the object recognition algorithms. Expanding the training dataset to include more diverse examples of these tools in various settings, as well as improving camera calibration, would help the system reduce misidentifications caused by visual overlap and tool proximity. This would ensure better overall performance in environments where tools are placed close together or in challenging orientations.

Despite these challenges, the ToolGuard AI consistently achieved identification accuracy rates ranging from 80% to 100% across most of the tools evaluated. This level of performance reflects a strong sensitivity to tool detection and classification, even under varying field conditions. The systems ability to maintain high recognition rates under different orientations, lighting conditions, and tool types highlights its robustness and operational potential. According to Szeliski, (2022) industry standards for tool tracking technologies in uncontrolled environments typically require a minimum accuracy of 80% to be considered viable for real-world deployment. The ToolGuard AI not only met this threshold but also exceeded it in the majority of test scenarios, reinforcing its reliability and technical readiness for integration into actual construction workflows. Its capacity to handle diverse tool types with minimal misidentification suggests that the system is not only functionally capable but also adaptable for broader application in the construction industry.

The performance results highlight that the ToolGuard AI is a capable and reliable system for real-time tool tracking, especially for tools with distinct visual characteristics. While the Spades detection challenges indicate areas for further improvement, the systems overall performance supports its potential for widespread application in construction environments. By incorporating advanced machine learning techniques, expanding the training dataset, and introducing adaptive lighting, the ToolGuard AI could enhance its ability to identify tools with complex shapes and continue to improve its accuracy and responsiveness across a wider range of tools.

As the ToolGuard AI continues to be optimized, human intervention will remain an essential part of the process, particularly when dealing with tools that the system struggles to detect. Manual verification will ensure the systems reliability until it becomes fully autonomous in tool tracking. Moreover, real-time learning algorithms could enable the ToolGuard AI to evolve dynamically by continuously adapting to new tools and environmental conditions. This adaptability would ensure the systems long-term viability, making it an even more robust and scalable solution for tool tracking in the construction industry.

The sensitivity of the ToolGuard AI in terms of tool identification was also evaluated. The results revealed that the device could correctly identify tools with a success rate of 70% to 100%, depending on the tools shape and clarity. Tools with clearer outlines were identified accurately, while those with irregular shapes resulted in slightly lower recognition. Nevertheless, the system was able to maintain consistent identification in most test conditions.

The SMS notification feature of the ToolGuard AI also functioned reliably. The system successfully sent real-time status updates to mobile numbers using a GSM module. Across the 10 trials per tool, it recorded SMS success rates ranging from 80% to 100%, with minor delays attributed to network connectivity issues. These results confirmed that the device was capable of maintaining communication under typical construction site conditions.

The results revealed that the ToolGuard AI achieved a 100% success rate in transmitting SMS notifications for the Hard Hat, Pickaxe, Hand Drill, and Hand Saw, indicating flawless system performance for these tools. All notifications were successfully delivered in every trial, which reflected the reliable connectivity and robust communication infrastructure of the system. These tools, with their stable transmission rates, confirm that the ToolGuard AIs SMS module is highly effective in ensuring tool status updates are communicated in real time.

The Angle Grinder, Metal Float, Masonry Float, and Welding Inverter recorded a 90% success rate, with one instance of failure for each tool. These minor disruptions were likely caused by temporary network fluctuations, occasional signal interference, or short internal delays, as the system occasionally struggled to maintain consistent connectivity. Despite these small interruptions, the system maintained a strong overall performance, demonstrating its ability to operate reliably even when network conditions were less than ideal.

The Spade had the lowest SMS success rate, recording only 80% success, with two SMS failures across the ten trials. This lower success rate could be attributed to connectivity issues or environmental interference, factors that are common on construction sites where signal strength can fluctuate due to obstructions or distance from cell towers. The performance of the Spade suggests the need for further improvements to strengthen the systems ability to transmit notifications in more challenging conditions.

Despite the Spades lower success rate, the ToolGuard AI demonstrated robust performance in SMS notifications, with success rates consistently ranging from 80% to 100% across all tools. These results underscore the systems reliability in transmitting notifications, which plays a crucial role in ensuring tool accountability, operational safety, and workflow coordination on construction sites. Real-time updates are essential in construction environments, where any delay in tool status updates can result in inefficiencies, safety risks, and delayed project timelines.

The results also highlighted the ToolGuard AIs adaptability to varying network conditions. As Akyildiz et al. (2019) noted, construction sites are often subject to signal interference, network congestion, and physical barriers that can hinder connectivity. These issues were likely the cause of the Spades miscommunications, which emphasize the importance of redundant communication channels. To improve system performance, incorporating multi-channel notifications—such as in-app alerts, email notifications, or Wi-Fi-based messages—could help overcome the limitations posed by unstable network signals. Zhang et al. (2021) suggested that multi-modal communication systems could offer more reliable and redundant means of transmitting important updates. 

This reliable SMS performance also contributes to digital coordination on construction sites by bridging real-time field activity with centralized monitoring. As Li et al. (2023) noted, integrated communication features help enhance situational awareness and accountability, making systems like the ToolGuard AI essential in modern, fast-paced project environments.

In line with best practices in real-time inventory management, the ToolGuard AIs SMS performance aligns with principles outlined by Zhou et al. (2020), who emphasized the importance of immediate notifications for maintaining inventory integrity and ensuring occupational safety. Delays or failures in notifications could lead to mismanagement of tools, downtime, and safety hazards. The ToolGuard AI performed well in this regard, meeting or exceeding the 90% industry benchmark for reliable notification systems (Islam et al., 2022). With dynamic network monitoring or automatic retries built into the system, the system could overcome occasional connectivity issues, strengthening communication reliability in environments where signal strength may vary.

The ToolGuard AIs performance in SMS notifications highlights the importance of real-time communication for effective tool tracking and inventory management in dynamic construction environments. While occasional lapses were observed, the systems overall performance supports its integration into industrial workflows, improving both tool accountability and site coordination.

To address the occasional issues in transmission, the ToolGuard AI could integrate redundant communication technologies and further enhance its notification reliability. By improving the network monitoring system and introducing features like retry mechanisms, the tools SMS notification system could adapt to the changing conditions encountered in construction sites, ensuring greater stability and coverage.

In summary, the SMS notification performance of the ToolGuard AI demonstrated its capability to reliably deliver status updates for tool monitoring in real-time. With success rates that met industry standards, the system showed its potential to improve operational efficiency and site safety. To address minor connectivity issues, especially for tools like the Spade, future system refinements should focus on network optimization and multi-channel notifications to ensure even greater reliability under varied construction conditions. Ultimately, the ToolGuard AIs communication performance affirms its suitability for real-world applications and provides a foundation for future enhancements aimed at improving tool tracking accuracy and operational coordination.

The results showed that the ToolGuard AI achieved a 100% on-time delivery rate for tools such as the Hard Hat, Pickaxe, Hand Drill, and Hand Saw, indicating flawless performance with no delays in message transmission. This flawless result highlights the systems capability to maintain uninterrupted communication in ideal conditions, which is vital for fast-paced industrial environments.

For tools like the Angle Grinder, Metal Float, Masonry Float, and Welding Inverter, the ToolGuard AI exhibited a 90% on-time delivery rate, with only one delayed message per tool. These occasional delays were likely caused by network fluctuations or brief internal processing delays that occasionally affected message transmission. These disruptions are common challenges in wireless communication systems, especially in complex environments like construction sites where network conditions are often unpredictable.

The Spade had the lowest on-time delivery rate, recording 80% success and showing two instances of delayed SMS responses. This outcome suggests that connectivity stability or processing lags may have influenced the transmission during certain trials. Such results point to potential weaknesses in network conditions or system optimization that could affect message dispatching under certain circumstances.

Despite these occasional delays, the ToolGuard AI exhibited strong performance in SMS response time, with on-time delivery rates ranging from 80% to 100% for all the tools tested. This result validates the systems ability to maintain timely communication, which is vital for decision-making, incident tracking, and operational continuity in dynamic environments. Timely notifications are a key factor in ensuring that construction sites run smoothly, where delays in information flow could lead to inefficiencies or safety risks. Real-time updates are critical to managing tool availability and minimizing workflow interruptions. The performance of the ToolGuard AIs SMS notification system also demonstrated its ability to adapt to different network signal strengths and intermittent connectivity conditions commonly found on construction sites. As Akyildiz et al. (2019) highlighted, industrial communication systems are often affected by signal interference, network congestion, and obstructions like walls and equipment. The occasional failures observed during the Spades trials can likely be attributed to these factors, suggesting the need for redundant communication channels. Enhancing the system by introducing multi-channel notification systems, such as in-app alerts, email notifications, or Wi-Fi-based messaging, could further improve its performance under challenging network conditions (Zhang et al., 2021).

Moreover, the ToolGuard AIs real-time responsiveness complements best practices in IoT-enabled monitoring systems, which emphasize the importance of low latency and high reliability in delivering notifications. Kumar et al. (2021) stated that in Industry 4.0 environments, effective decision-making depends on the instantaneous flow of data. Any delays in transmitting tool status updates can lead to operational bottlenecks or safety hazards. In this context, the performance of the ToolGuard AI with success rates exceeding 90% meets or exceeds the industry benchmark for real-time notification systems (Islam et al., 2022).

The Spades delays, in particular, emphasized the need for system optimization to enhance communication reliability. Other potential improvements include backup delivery mechanisms or adaptive strategies, allowing the system to adjust to network conditions and maintain reliable message delivery even during fluctuating connectivity (Rashid et al., 2023).

The ToolGuard AI demonstrated high performance in SMS notification delivery, confirming its effectiveness in real-time tool monitoring and operational coordination. Although minor delays were noted—particularly with the Spade - the systems overall responsiveness highlights its practicality for construction environments where timely updates are essential. To further improve communication reliability, future enhancements could include multi-channel messaging, dynamic network diagnostics, and adaptive delivery mechanisms, ensuring broader coverage and greater resilience under varying conditions. With these upgrades, the ToolGuard AI stands to remain a dependable solution for automated tool tracking, promoting operational efficiency, safety, and effective resource management.

The integration of GSM-based SMS communication proved especially suited for field deployment, offering consistent and dependable message transmission even in areas with limited internet connectivity. This approach ensured that tool status updates reached users without interruption, supporting continuous coordination, and minimizing downtime in fast-paced construction settings.

The acceptability of the ToolGuard AI in terms of technical composition, operating performance, safety, user-friendliness, and accuracy in tool identification was evaluated by 30 respondents. Statistical tools such as frequency counts, mean scores, and percentage analysis were applied to interpret the evaluators responses. The results showed that the system was rated “Very Acceptable” in all categories, with the highest mean scores in operating performance, safety, and user-friendliness. The results demonstrated the devices overall capability to meet user expectations in terms of function, reliability, and usability. The grand mean of 4.71 confirmed that the ToolGuard AI was well-received and considered suitable for practical application in construction environments.

These results are supported by contemporary research on AI-powered object detection. Zhao et al. (2018) emphasized the impact of well-trained computer vision models in enhancing inventory reliability when applied to diverse tool sets. Chen et al. (2019) pointed to the importance of detection stability in cluttered environments, which the ToolGuard AI generally achieved. Lee and Park, (2021) stressed that reducing false identifications is key to fostering user trust in AI-integrated monitoring systems, while Martinez et al. (2023) highlighted how real-time detection with high accuracy supports smoother workflows and minimizes disputes over missing or misplaced equipment.

The results therefore establish the ToolGuard AI as a reliable solution for accurate tool identification in construction environments. While its current performance meets the expectations of users and operational demands, ongoing improvements - particularly in addressing detection precision under complex site conditions, would further enhance its functionality and ensure broader applicability in increasingly advanced and data-driven construction practices.

The effectiveness of the ToolGuard AI: A Monitoring and Inventory Management System was demonstrated through its successful integration of AI-driven detection and automated communication technologies. These components enabled the device to monitor tools and update inventory in real time, supporting accountability and improving operational coordination in active work environments. The system consistently fulfilled its intended functions, including the accurate identification of tools, timely generation of inventory records, and reliable SMS notifications. Its modular hardware setup and flexible software design contributed to its stable operation across various testing conditions, highlighting its adaptability for different site scenarios. Furthermore, the ToolGuard AI received strong approval from evaluators in terms of technical composition, operational reliability, safety, user-friendliness, and accuracy in tool identification. These results confirm that the system offers a practical and scalable solution for improving tool inventory processes. The study supports its potential application in construction environments where precision, automation, and real-time response are increasingly valuable. While the ToolGuard AI met its key objectives, future refinements—particularly in enhancing recognition consistency for tools with irregular shapes and ensuring reliable connectivity in low-signal conditions—could further improve its performance. Overall, the system demonstrated readiness for real-world use and contributed meaningful insights into the application of AI and IoT technologies in construction inventory systems.

The author expressed appreciation to those who supported this research, particularly the evaluators and Capiz State University, Main Campus.

The author confirms no conflict of interest.