A Hybrid Model for Enhancing Risk Management and Operational Performance of AEC (Architectural, Engineering, and Construction) Consultants: An Integrated Partial Least Squares–Artificial Neural Network (PLS–ANN) Approach

Abstract

The operational effectiveness of Architectural, Engineering, and Construction (AEC) consultants, whose services have a substantial impact on project execution and results, depends on effective risk management. Using 336 survey responses from professionals in the construction industry, such as consultants, contractors, and employers working on a range of infrastructure and building projects, this study validates a hybrid Partial Least Squares Structural Equation Modeling–Artificial Neural Network (–ANN) approach. In order to ensure both causal analysis and predictive insights for AEC consultant performance assessment, this study combines PLS–SEM and ANN to develop an integrated performance evaluation framework. While ANN ordered their relative relevance in a non-linear predictive model, the PLS–SEM analysis found that the two most important predictors of consultant performance were communication and relationship management (G03) and document and record management (G06). The hybrid approach is a more efficient and data-driven tool for evaluating AEC consultants than traditional regression models since it accurately captures both causal links and predictive performance. These results contribute to a robust and sustainable framework for performance evaluation in the AEC sector by offering practical insights into risk reduction and operational improvement.

1. Introduction

The contractual and technical complications of construction and regulations impose the appointment of a consultant—as employer agent—in order to ensure timely completion of the project, as well as the required quality of project specifications and design standards, and within the allocated budget [1,2,3]. To lead, coordinate, supervise, and manage a project from inception to completion while fulfilling the project’s goals, a public or private organization (a team) is hired as the client’s consultant or representative [4]. The consultant heavily participates in the planning, monitoring, and controlling of the different construction processes, which involve managing quality, cost, schedule, safety, and environmental issues handled by a technical and specialized team [3,5]. Moreover, the consultant’s responsibilities include reviewing shop drawings, responding to the contractor’s requests for information, providing design clarifications and interpretations, and monitoring the contractor’s works in order to ensure the timely operation of the contractor by means of reviewing the program of works; wittiness tests; certifying payment; assessing the financial situation, extension of time claims, and finally preparing reports to the employer [6,7]. Moreover, reliable AEC (Architectural Engineering and Construction) consultants provide support to the project to help remove some aspects of trial-and-error that could otherwise happen due to the contractor [8]. As a result, the professional, managerial, and technical services by the consultant underpin the utilization of its services to the interest of all the project stakeholders [3].

This wide range of tasks requires a high level of management skills, such as competent staff, communication, and reporting skills, and also good management of issues and processes [7]. Past studies revealed that the consultant’s low performance caused various cost overruns, delays, reworks, unnecessary changes, claims, and disputes [9,10,11,12]. Therefore, it is important to know when low performances take place and try to know the underlying causes in order to establish and implement effective mitigation measures by practitioners. Furthermore, the consultant tries to maximize his/her profit while the clients try to reduce possible costs. This causes conflict, hinders client–consultant mutual trust, and causes reduced collaboration [3]. In addition to the aforementioned problems, the presence of asymmetry of information problems in the client–consultant relationship demands additional efforts to control the consultant tasks [4]. The heavy participation of the AEC consultant in construction projects necessitates evaluating the consultant’s performance at the earliest chance possible [13,14]. As a result, regular evaluation of the consultant’s tasks, performance frameworks, and modeling of performance indicators have been developed to reduce the consultant’s low-performance aspects by researchers and governmental bodies [1,11,15].

It should be noted that different terms have been used in the literature to refer to the same function of AEC consultants. These terms include architect, client representative, contract administrator, engineer, engineer representative, project consultant, and supervising engineer [4,16].

The AEC consultant’s assessment has been given some attention in a number of studies, which can be broadly grouped under three categories: (1) the tasks, roles, and responsibilities of consultant [9,17,18]; (2) the establishment of multi-attribute models and frameworks [1,3,7,11,12,15,19,20,21,22,23]; (3) the impact of consultant performance on construction projects [9,24,25].

Ref. [26] performed an extensive assessment of performance measuring methodologies at the project level within the construction sector. Their research underscores the significance of integrating both financial and non-financial key performance indicators (KPIs) for a comprehensive assessment of project performance. Ref. [27] performed a systematic literature review examining the methodologies and selection strategies for consultants in the construction sector. Their research delineates three primary selection methodologies: qualification-based, price-based, and a hybrid of qualification and price-based selection. The study highlights the intricacy of the selection process, which frequently incorporates both direct approaches, grounded in prior experience or reputation, and comparative methods, employing assessments and selection criteria. Ref. [28] presented a conceptual framework for a robotic cyber–physical system designed to automate reality capture and visualization in the monitoring of construction progress. Ref. [29] introduced a hybrid simulation-based methodology for resource allocation and constructability assessment in Roller Compacted Concrete (RCC) pavement initiatives. Their research illustrates the utilization of integrated simulation methods to enhance resource distribution and operational strategy. Ref. [30] examines the impact of stakeholders, such as clients, contractors, and consultants, on construction project performance, with an emphasis on small and medium-sized contractors in Jordan. Ref. [31] presents a systematic literature review of the factors contributing to team effectiveness in construction projects. The research finds leadership, communication, collaboration, trust, and leadership as the primary factors of team performance. Ref. [32] presents a multi-objective optimization model aimed at balancing cost and time in construction projects while considering improvements in productivity. The study presents a novel mathematical methodology for determining optimal compromises, guaranteeing that projects adhere to timelines while remaining within budgetary constraints. Ref. [33] proposes a system for evaluating the rate of Building Information Modelling (BIM) adoption in construction projects. The research develops an extensive instrument of measurement that evaluates variables including competency of the workforce and technical framework organizational readiness. Ref. [34] examines the nascent function of generative artificial intelligence (AI) in the management of construction projects. The research examines the ways in which AI-driven technologies improve decision-making, automate redundant tasks, and optimize the workflow of construction. Ref. [35] investigates the influence of supervising consultants on the performance and efficiency of construction projects.

Recent academic studies emphasize the significance of strategic selection methods, comprehensive performance measurements, and advanced resource planning and technological integration in improving consultant performance in the construction industry. These insights offer essential assistance for practitioners and researchers seeking to enhance project outcomes via effective consultancy methods.

Table 1. Summary of literature review on consultant performance.

Research	RS (AM)	Brief of Study
[9]	Qualitative	This research discussed efficient and effective administration of payment clauses by the client and consultant under the different standard forms of contract though content analysis of contract clauses, comparison between different forms of contract and graphical presentations of payment administration. It developed a guideline and a checklist related to the employer’s and consultant payment obligations.
[17]	Qualitative	This study covered the methods used by consultants to administer variations, labor and material claims, delays and disruptions, planning, and cost control during the construction phase of building contracts. It recommended an appropriate contingency for proper control of project cost and identified factors affecting it.
[18]	Survey Method (EFA)	This research determined 23 performance factors and 7 groups of construction consultants performance affecting achievement of time and quality targets on construction project of Samarinda Baru Airport, Indonesia. It emphasized the importance of having qualified and experienced consultant’s staff.
[11]	Survey Method (Relative Weighting)	This study, which had a sample size of 80, looked at the effectiveness of construction consultants on common fund projects in Ghana using a list of 29 jobs that the Hong Kong Housing Authority had developed to cover the various project phases. The results revealed that the average overall performance of the consultants was around 62% of the maximum score. This indicates the needs for continual improvement and regular performance evaluation of consultant roles.
[19]	Survey Method (RII)	With 72 factors sorted into 10 groups, this study assessed the consultant practices in construction projects. With the input of 42 Ghanaian’s construction practitioners, the findings pointed to the importance of setting policies, rules, and procedures, the importance of effective performance monitoring and implement enterprise resource planning systems to support consultant practices.
[20]	Survey Method (Comparative Analysis)	This study investigated the contract administration practices in the United Nations (UN) system using the inputs from 262 UN employees. The contract management–process assessment model (CM–PAM) was utilized to address the degree of implementation of 10 key contract management categories and 112 associated processes. The study findings reveal different level of implementing contract management in the UN.
[7]	Survey Method (RII)	This study identified six groups and 36 characteristics that have an impact on how well consultants execute on government projects in Ghana, and it ranked those elements in order of importance. They were divided into groups for management, quality, time, cost, environmental considerations, and other reasons. The most important factors included the project’s urgency, the length of time required to complete it, political pressure from higher authorities that affected project delivery, the consultancy team’s ability to make prompt decisions, and the team’s prior project experience.
[21]	Survey Method (SEM)	In order to investigate factors impacting delays in Indian building projects, this study’s SEM model, which was developed, has 19 components and 5 constructs. The SEM results using data from 77 questionnaires found that increased commitment among participants, including consultants, and clear roles and duties can improve performance.
[12]	Pilot and Survey Studies (MR)	Using 62 criteria and 6 groups, this study examined contract administration methods, including consultant roles on federal and state DOT projects in the United States. The study suggested that monthly performance reviews be conducted with the questions from the research questionnaire serving as a starting point.
[23]	Qualitative Survey (APH)	This study defined the contract management process based on a literature review and then assessed the contract management capabilities in Korea through 92 experts using the AHP technique. The results revealed the importance of employing experts with engineering skills to manage the project and properly administrate the contract.
[15]	Interview and Case Studies (CA)	This study comprises a model for contract setting up, document management, risk management, change management, relationship management, and contract closing concerns in the Netherlands wastewater construction sector. It has 93 elements and 7 categories. The study’s 13 interviews highlighted the significance of assigning defined duties and responsibilities, scheduling a kick-off meeting, and thoroughly discussing the contract with the parties. It also revealed the relevance of monitoring measures, developing and maintaining a document management system, and updating records.
[3]	Mixed Methods (CNA&SEM)	By enhancing capacity integration and responsibility delegation in client–consultant collaboration using 13 factor and 4 groups, this study examined the growth of the consulting industry. The findings suggest that consultant capabilities contribute to project process and more delegating of responsibilities increase the consultant control over the project success.
[6]	Mixed Methods (EFA)	This study examined 47 overall factors that contributed to delays in the Portuguese construction industry at various project phases. The delay in reviewing/approving drawings, the delay in quality control, the consultant’s inflexibility, and the wait for test and inspection approval were identified as the poor performance factors of the consultant causing project delay by the consultant through the collection of 94 survey responses, exploratory factor analysis, and focus group discussions. The author suggested seven mitigation measures, including choosing consultants with the right expertise and experience for the project; paying them appropriately; improving the consultant working environment; establishing an effective and efficient communication protocol; establishing change management protocols prior to the project’s start; and developing a supply chain management framework.
[25]	Mixed Methods (AHP)	In this study, 8 groups and 37 delay factors that affect the implementation of sports projects globally were identified and evaluated. The Analytical Hierarchy Process of the multi-criteria decision-making approaches was used to generate the Relative Importance index (RII) of each element from the 101 replies to an online questionnaire. A low level of experience, delays in inspecting, rewriting, and approving papers, errors in the consultant’s issued documentations, and poor communication and coordination were the causes that contributed to consultant delays in that order. After the delays related to the contractor, the delays factors for the consultant came next. The research’s findings were summarized by a recommendation to form a team of technical consultants who are appropriate and knowledgeable for the project type and to use a professional document management system to cut down on the amount of time needed to review or approve the contractor’s submissions.

Note: RS = Research Strategy; AM = Analysis Method; RII = relative importance index; MR = multiple regression; EFA = explanatory factor analysis; SEM = structural equation model; AHP = Analytic hierarchy process; CA = Content analysis; and CNA = Collaboration Network Analysis.

presents some of the recent research studies on consultant performance.

Table 1 lists several approaches, including survey-based models, qualitative analysis, and statistical methodologies, for evaluating the performance of AEC consultants. Although these methods offer insightful information, they frequently fall short in terms of objectivity, generalizability, and predictive power. Many do not capture non-linear interactions since they rely on correlation-based methodologies. This study fills these gaps by combining PLS–SEM with ANN, providing a more reliable, data-driven framework that improves the predictive accuracy and causal analysis of policies regarding the improvement of AEC consultant performance.

1.1. The Model and Hypotheses Development

In assessing the effect of certain key factors affecting AEC consultant performance, a conceptual model that clarifies the relationships between factors that cannot be directly observed (Latent Variables; LVs) and their related variables (Manifest Variables; MV) is essential. In this study, the conceptual model is grounded on the construction contract administration framework (CAPF) developed by [1] with 93 manifest variables (MVs) which were grouped into 11 latent variables (LVs). The definitions of the 93 MVs are presented in the fishbone diagram shown in Figure 1.

1.1.1. G01-Project Governance and Start-Up

The AEC consultant’s performance relies on its ability to establish ground roles and high-level strategies to control the project from the commencement date [19,20,36]. It was found that project governance and start-up play a more important role in controlling the project. Ref. [37] investigated the influence of asymmetric trust on project management efficacy in construction projects and determined that information sharing is a key component in enhancing project success. Ref. [38] examined the implementation of the Critical Path Method (CPM) in consulting projects and determined that efficient project planning enhances consultants’ performance and time management. A study conducted by [39] investigated the efficacy of project management abilities in large-scale engineering projects, revealing that well-organized project management processes significantly contribute to enhancing consultant performance and managing project complexities. Based on the findings of these previous studies, this study hypothesizes the following:

H1: 

There is a significant and positive relationship between project governance, start-up management, and AEC consultant performance.

1.1.2. G02-Team Management

The consultant’s management skills can be reflected by the effectiveness and efficiency of its staff [14]. Therefore, team management expresses the consultant’s capability to assign a qualified and highly skilled team on time [18]. Also, management must provide timely admonition to underperformers. In addition, ref. [4] argued that late resource allocation should be avoided or corrected and that the consultant team’s advice on time and effort influences the project’s overall quality. Additionally, according to [20,40], the amount of experience of the consultant team has a substantial impact on how quickly projects are completed and how well they turn out. Therefore, consultants must make sure that their team members have the necessary training and hands-on expertise to manage the projects [40]. A comprehensive literature review by [31] emphasized the influence of team performance on project success within the construction industry and determined that effectively managed teams enhance the productivity of consultants. Furthermore, ref. [41] investigated the aspects pertinent to the management of dynamic teams, demonstrating that appropriate management strategies and team composition are essential for increasing team performance and, consequently, consultant efficacy. Ref. [42] investigated the psychological impacts of project managers’ work environments in AEC projects, revealing that supportive team management strategies can enhance project success rates and consultant performance. Consequently, the second hypothesis is stated as follows:

H2: 

There is a significant and positive relationship between team management and AEC consultant performance.

1.1.3. G03-Communication and Relationship

As demonstrated by several projects and supported by the agency theory, effective communication, relationships, and collaboration between consultants and clients can be supportive of project success [21]. One of the main reasons is that good communication and relationships will build trust, and the building of trust will greatly reduce transaction costs. A lack of trust, on the other hand, will result in an increase in transaction costs due to such tasks as negotiations of the contract, search for more information, inspections, and close monitoring of works [43]. Another reason is that the wide range of tasks assigned to the AEC consultant puts the acts of the consultant in the center of contact with all parties, and this will lead to the development of more opportunities to drive the project toward success. On the contrary, any adversarial relationships might impact the team closeness and, hence, prevent the desired outcomes [3,14] and miscommunication between clients and consultants might contribute to the necessity to rework the project [40]. A study by [44] demonstrated that managers’ efficient communication methods enhance overall performance by increasing employee work satisfaction. Ref. [45] investigated the external communication strategies of organizations, demonstrating that effective relationship management and strong communication are essential for enhancing efficiency and organizational performance. Additionally, ref. [46] examined the impact of customer relationship management (CRM) on market performance, asserting that strong CRM enhances organizational success and amplifies the efficacy of consultants. Accordingly, the following hypothesis has been formulated:

H3: 

There is a significant and positive relationship between communication and relationship management and AEC consultant performance.

1.1.4. G04-Quality and Acceptance

As part of the consultant’s responsibilities, the team should supervise the works to ensure the contractor’s compliance with the project specifications and standards [47]. The authors of ref. [3], in their study on consultant performance, found that the ability of the consultant to manage the quality of deliverables, reduce the number of accidents, and minimize the overall environmental impacts would lead to greater performance. Ref. [25] revealed that the review time consumed by the consultant team in reviewing the contractor’s submissions would significantly affect the time delivery of the project. Ref. [48] presented a quality management system aligned with ISO 9001 standards for AEC businesses, asserting that this approach enhances administrative processes by increasing quality during the design phase, thereby positively impacting consultant performance. Ref. [49] asserted that the integration of Building Information Modeling (BIM) and the Earned Value Management System enhances the performance of AEC projects in terms of cost, time, and quality and noted that this integration boosts the efficiency of consultants significantly. Furthermore, a study by [50] demonstrated that organizational change management strategies are essential for the adoption of new technologies, enhancing consultant performance through the improvement of quality management systems. Therefore, this study hypothesizes the following:

H4: 

There is a significant and positive relationship between quality and acceptance management and AEC consultant performance.

1.1.5. G05-Performance Monitoring and Reporting

The transmission of a prompt, precise, appropriate, regular, and clear information flow is regarded as one of the keys to the success of the relationship between supply chain members. Efficient information flow will, in turn, enable the growth of the collaborative situation and better relationships [43]. Also, the contractor’s delay may cause several losses to the client; thus, the progress of a project needs to be monitored and controlled by the constant [24] and should be reported periodically [15]. Furthermore, it is essential to keep the client up-to-date with the project issues and progress of the consultant so that the client can make informed and prompt decisions [6]. Additionally, an accountable consultant alerts the contractor of any obligation failures and delays that will take the project back to the contractual dates [4]. A comprehensive review by [51] demonstrated that consistent performance monitoring and reporting significantly enhance project success. Ref. [52] showed that their technique for comparative evaluation of construction project performance effectively enhances project efficiency and identifies opportunities for improvement. A study by [28] highlights that a robotic cyber–physical system facilitating visualization and automatic data collection in construction progress monitoring enhances consultants’ decision-making by optimizing performance monitoring processes. These studies demonstrate that performance monitoring and reporting systems enhance project success rates by improving consultant efficiency. In this regard, this research hypothesizes the following:

H5: 

There is a significant and positive relationship between performance monitoring and reporting management and AEC consultant performance.

1.1.6. G06-Document and Record

Today, document and record management are a vital business strategy for many organizations because it helps to improve the knowledge within the organization and to maintain rights [1]. Organizations with well-established knowledge management processes, according to [43], will rely on a culture of information sharing. It has been shown that a well-established documentation system that is managed with current registrations and that will provide stakeholders with data is one of the key elements for successful document and record management [1,15]. Ref. [53] examined the enhancement of project management through sustainable practices utilizing Information Requirement Matrices and data-driven information models, highlighting that these approaches enhance the efficiency of consultants by optimizing records management processes. Ref. [54] investigated the influence of service innovation capabilities on company performance and showed that the business model mediates this relationship, indicating that efficient record management substantially enhances business operations. A study conducted by [49] demonstrated that the combination of the Earned Value Management System with Building Information Modelling (BIM) enhances project management procedures, enhances consultants’ efficiency, and significantly impacts project performance. These studies demonstrate that document and records management strategies improve project processes, elevate overall business success, and increase the operational efficiency of AEC consultants. Considering this, this study postulates the following:

H6: 

There is a significant and positive relationship between document and record management and AEC consultant performance.

1.1.7. G07-Financial Management

The effectiveness and efficiency of the adopted financial management systems reflect the consultant’s management skills [14]. The consultants would have better performance if they established a financial management system that would, in a timely manner, certify any due payments and track the client’s payment process [1,21,47]. Ref. [6] states that the elimination of bottlenecks and pointless bureaucracies in the payment process is crucial because late payments are one of the major factors in project delays. To score the consultant’s performance, the contractor’s remuneration for any delayed payments and their influence on the contractor’s cash flow must be assessed promptly [9,19,23]. Ref. [49] shows that the combination of the Earned Value Management System (EVMS) with Building Information Modelling (BIM) enhances the performance of Architecture, Engineering, and Construction (AEC) projects in terms of cost, time, and quality while also augmenting the efficacy of consultants. This study demonstrates that financial management techniques facilitate the operational success of consultants and yield sustainable performance enhancement in the AEC sector. Following the discussions, the following theory is put forth:

H7: 

There is a significant and positive relationship between financial management and AEC consultant performance.

1.1.8. G08-Changes and Changes Control

Changes in construction are unavoidable situations, and, whenever a change arises, it requires the amendment of the original specifications and plans accordingly [15,19,47]. Frequently, late changes are associated with rework and subsequently increase the risk of disputes and the possibility of undermining project performance [55]. According to [25], approved change may affect the completion date of the project, and an increased number of scope changes was ranked as the top significant factor causing a project to be delayed. Therefore, the consultant should evaluate, in a timely manner, the contractor’s proposals for possible changes and suggest workable solutions when required. When urgent changes are concluded, the contractor should be notified, and changes should be processed to compensate the contractor in a timely manner [1,22,47]. Ref. [50] investigated the correlation between the effective application of change management strategies and the adoption of new technologies in the AEC sector, revealing that factors such as the efficacy of change agents, the establishment of measurable objectives, and the formulation of realistic timelines are crucial for the successful execution of changes. Ref. [50] identified the factors that facilitate employee adoption of change management processes, asserting that top management commitment, clear communication of change benefits, and the provision of training resources are critical for successful change implementation. Furthermore, a study by [56] demonstrated that employee involvement and efficient management techniques throughout organizational change processes enhance the sustainability of changes and improve consultant performance. Based on these discussions, the subsequent hypothesis has been formulated:

H8: 

There is a significant and positive relationship between changes and change control management and AEC consultant performance.

1.1.9. G09-Claims and Disputes Resolution

The managing of disputes is considered to be a frequent problem in construction projects that requires attention [55]. According to [9,10], an increasing number of disputes is observed in construction, and these mainly arise due to consultant performance. The settlement of any possible claims is crucial during the construction and closeout phases of a project [15,22,23]. Alternative dispute resolution procedures, such as mediation, arbitration, litigation, minitrials, conciliation, and adjudication, may be used to resolve contractual disputes if negotiation fails [55]. Refs. [1,14,55] emphasize that outstanding claims and unresolved disputes not only affect the contractor’s cash flow but also the relationships between various stakeholders. Based on this finding, the authors argue that it is the primary responsibility of the consultant to close these claims in a timely manner and propose an alternative dispute resolution mechanism. Ref. [57] indicated that Alternative Dispute Resolution (ADR) methods, particularly negotiation, mediation, and arbitration, facilitate the efficient and rapid resolution of disputes in construction projects and significantly enhance the performance of consultants. Ref. [58] asserted that the effective management of conflicts in engineering contracts directly influences project success and the efficiency of consultants, demonstrating that dispute management solutions enhance the operational efficacy of consultants. Ref. [59] investigated the integration of Building Information Modelling (BIM) in dispute resolution within construction projects, demonstrating that BIM-based management methods enhance project efficiency by augmenting consultants’ performance. From this aspect, the formulated hypothesis is as follows:

H9: 

There is a significant and positive relationship between claims and dispute resolution management and AEC consultant performance.

1.1.10. G10-Contract Risk Management

Risky events will always occur, and they may cause changes to the originally planned activities, cost, or quality of a project [15,47]. According to the agency theory, the principle (client) is risk-neutral and profit-seeking, whereas the agent (consultant) operates as “risk-averse” and pushes risks onto the primary (client) [4]. Relationships between participants in the supply chain are impacted by risk sharing and allocation [2]. The characteristic of good risk management includes periodic assessment and assignment of responsibilities [15], performing an early review of the design to reduce the employer payments for the additional costs, and preventing possible consequences of late detection of design errors [21]. Monitoring the contractor’s financial situation and potential for bankruptcy will also be the primary steps to prevent project distribution anytime the contractor’s financial stability is in doubt [1]. Ref. [60] investigated the impact of risk allocation and sharing on the performance of mega-industrial projects within the FIDIC 2017 contract model, demonstrating that an equitable distribution of risks is essential for enhancing the efficacy of consultants. Ref. [61] examined the influence of contract management methods on the success of construction projects, highlighting that successful contract management is a critical element in enhancing the performance of AEC consultants. A study by [1] employed a multidimensional fuzzy logic model to assess construction contract management processes, demonstrating that effective contract management methods positively influence project success and consultant performance. In this regard, this research hypothesizes the following:

H10: 

There is a significant and positive relationship between project contract risk management and AEC consultant performance.

1.1.11. G11-Contract Close-Out

Under the professional services contract, the contractor is required to present the capability to the consultant to properly and comprehensively close out the executed contracts. Moreover, the consultants’ duties will remain intact during the defect liability period [19,47]. During this stage, the consultant is required to verify the completion of his/her work as well as review closeout documentation, draw up a list of defects, continue inspection by the technical team [14], settle any unresolved claims, and then process the contractor’s final account by the financial management team [1,14,17]. Ref. [62] investigated the causes of delays in the closing procedures of highway building projects and determined that efficient closure management enhances consultants’ performance by improving project success. Ref. [63] assessed the influence of best-value proposals on project closure performance, demonstrating that effective contract-closing processes lower overall project expenses, thus enhancing consultants’ productivity. Thus, this research hypothesizes the following:

H11: 

There is a significant and positive relationship between contract close-out management and AEC consultant performance.

While recent studies in the literature have identified various factors affecting AEC consultant performance, few studies have analytically evaluated its impact at the project level, and little is known as to how it should be evaluated during the construction phase [1,3]. In addition, the previous studies were formulated to serve a certain geographical area [7,11,12,19,21] or a specific type of project [7,19,20,25] or qualification of indicators [9,17,23]. The main goal of this study is to quantify the most important predictors that have the greatest impact on the performance of AEC consultants. Although PLS–SEM has been used often in construction management research to validate the study hypotheses, there have been relatively few attempts to combine it with other artificial intelligence techniques and even fewer studies of performance measures. Thus, by investigating the integration of PLS–SEM with ANN for AEC performance and then pinpointing the most important predictors, our research fills the gap in the literature. The results of this research help identify the key performance indicators for the issues the consultant identified. The construction industry may concentrate on the most important predictors about which improvement programs can be initiated by ranking the most significant predictors that affect the performance of AEC consultants, and the findings of this research will be especially helpful for decision-makers in their implementation.

In this study, an online questionnaire was adopted to rate the importance of factors affecting AEC consultant performance. The gathered data were analyzed through a multi-analytical approach by integrating PLS–SEM and ANN and then identifying the most significant predictors. The outcomes of this analysis lead to the determination of the key performance indicators for the AEC consultant’s performance-related factors. The construction business can concentrate on the most crucial predictors about which improvement projects can be begun by rating the most significant variables that influence the performance of AEC consultants.

This paper offers a novel method for evaluating the performance of AEC consultants by combining Artificial Neural Networks (ANNs) with Partial Least Squares Structural Equation Modeling (PLS–SEM). By addressing non-linearity in performance predictors and improving predictive accuracy, the hybrid PLS–ANN framework allows for a more thorough examination than prior research, which mostly relies on linear statistical models. Furthermore, by adding new variables that were previously disregarded in the literature, this study improves and broadens the use of key performance indicators (KPIs). These developments give industry experts additional insights to maximize consultant performance in construction projects by offering a more accurate and useful assessment methodology. The results provide a solid, data-driven method for performance evaluation in the AEC industry, helping to close the gap between theoretical research and real-world applications.

2. Methods and Models

Information was gathered about the significance of MVs using an online self-administered survey (Appendix A). The quality and dependability of the data were evaluated. Questionnaires were eliminated according to a number of criteria in order to guarantee the accuracy and dependability of the data. To ensure accuracy, responses containing inconsistent or contradicting data, missing or incomplete answers, and duplicate submissions were eliminated. Only reliable and significant data were incorporated into this study thanks to these standards [10]. The conclusion can also be shortened as the authors describe the summary of the article and indicate those predictors that, after calculation in the model, turned out to be the most important ones. Iterative PLS–SEM was used to evaluate the smaller model size and the twelve study hypotheses and to pinpoint the independent variables (LVs) that significantly affect the performance of AEC consultants. As the fifth step, the significant LVs obtained from SEM (as output) were used as input to rank their normalized importance through ANN, and, subsequently, the results were averaged by ten-fold runs. A combination of SEM and ANN helps to overcome the weakness of the model, which is oversimplifying in SEM and overfitting in ANN.

The SEM uses two different types of models, a measurement model (confirmatory factor analysis, or CFA) and a structural model (path analysis), to ascertain the relationship between LVs [64]. According to [65], it is regarded as a potent tool for estimating conceptual models connecting two or more unobserved variables. Structural Equation Modeling (SEM) simultaneously analyzes independent–dependent relationships in a single model, accounting for measurement errors and latent variables, unlike regression analysis [64]. PLS–SEM is a reliable method for drawing statistical conclusions from data that do not conform to parametric assumptions. The route coefficients and other unknown parameters are estimated iteratively, block by block. According to [66], the PLS–SEM method is effective for non-normal data, formative constructs, and small sample numbers. In management research like marketing [67], strategic management [68], and management information systems [69], the SEM has strong foundations and is widely used. Ref. [70] used the SEM in the construction industry with a sample size of 114 and AMOS software (version 26) to examine a framework and validate the relationship between lean adoption and safety performance.

One method of artificial intelligence is the artificial neural network (ANN). According to [71]’s definition from 2007, an ANN is a massively parallel, distributed processor made up of simple processing units with a neural tendency to store data, learn from experiments, and guarantee the usage of that data. Because ANN features employ a specific number of neurons connected by directed weighted links and are organized in layers, they perform better in predictions than out-of-the-box regression algorithms.

The ANN has been found to be a very novel and useful model when applied to problem-solving and has obvious advantages over other methods of statistics, such as SEM and multiple regression analysis [72]. Firstly, the ANN approach is suitable for describing non-linear pattern problems because it can realize both simple linear and complex non-linear relationships between input and output variables [73]. Secondly, to promote the effectiveness of linear relations and reduce the training time necessary for a large training set, an ANN is superior to other conventional models like logistic, discriminant, and multiple regression analyses [43,74]. Third, an ANN provides more accurate forecasts, and its accuracy can be increased still more by learning from the input. Fourth, an ANN model requires no assumptions about data properties or distribution [72]. Finally, an ANN is more flexible in terms of factor loadings, linearity, sample size, and homoscedasticity [74], and it has good generalizability capabilities [75], making it valuable for regression analysis. However, being a “black box” method sets a limitation on ANN and brands it as inappropriate for the testing of hypotheses and examination of causal relationships.

There are numerous ways to integrate ANN and SEM, according to the literature. For instance, the ANN was used to predict a portion of the responses while the SEM was used to examine another piece [73]. However, ref. [75] recommended that the goodness-of-fit of the study model be reviewed to determine the causal relationship before implementing ANN to predict specific factors. Additionally, the outcomes of an ANN might be used in SEM analysis.

2.1. Subsection

2.1.1. Sampling and Collection of Data

A self-administered, online questionnaire from the CAPF’s 93 consultant tasks as MVs and 11 LVs was used in this study. The questionnaire involved three parts, including questions on respondents and their profiles, ratings of 93 factors, and 11 performance groups. A five-point Likert scale was deployed to rate the importance of the contributing factors and performance groups from not at all important = 1 to extremely important = 5. According to [4], a five-point scale is usually adequate to distinguish the performance differences. The selection of respondents was made randomly from amongst construction practitioners from both the private and public sectors. A total of 366 responses were collected, with a response rate of more than 30%. A total of 30 out of 366 questionnaires were considered unsuitable when tested by Mahalanobis distances (d-squared) and standard deviations because they contained unengaged responses and outlier data. A sample size of less than 200 has been employed in 77.4 percent of the research, according to an analysis of 84 literature papers by [76] on the sample size needed for SEM models in construction management. Like this, ref. [77] conducted a critical review of 139 papers using PLS–SEM between 2002 and 2019 and found that the average sample size was 165. They also found that 58.3% of researchers preferred PLS–SEM over covariance-based SEM because of its ability to handle sample size data and non-normal data. Additionally, ref. [78] verified that PLS–SEM effectively handles small sample sizes. For a trustworthy result analysis, ref. [79] suggested sample sizes ranging from 100 to 200. From now on, a sample size of greater than 300 is adequate for PLS–SEM analysis.

Table 2. Respondents demographics.

Category	Sub-Category	Total	Respondents (%)
Orgnization	Employer	49	14.6
	Consultant/Designer	164	48.8
	Contractor	117	34.8
	Others	6	1.8
Sector	Private	187	55.7
	Public	125	37.2
	Mix	24	7.1
Exprience (years)	<10	74	22.0
	10–15	57	17.0
	15–20	69	20.5
	20–25	66	19.6
	>25	70	20.8
Working Position	Executive Manager	14	4.2
	Department Manager	62	18.5
	Project Manager	71	21.1
	Senior Engineer or Architect	75	22.3
	Quantity Surveyor	30	8.9
	Engineer/Supervisor	73	21.7
	Others	11	3.3
Professional Registration	Not Registered	87	25.9
	Registered	249	74.1

presents the respondents’ demographics.

The number of the contractor participants in this study is 117 (34.8%). Additionally, 164 (48.8%) consultants and 49 (14.6%) clients participated in this study. A total of 187 (55.7%) participants were from the private sector, and 125 (37.2%) were from the public sector. Most of the respondents 262 (78.0%) have working experience from 10 to over 30 years in construction. Additionally, 249 responders (74.1%) hold professional registration. The respondents hold a variety of positions (i.e., field engineers to senior and managerial positions) in their organizations. In conclusion, the respondents’ views reflect public and private sectors and owners, consultants, and contractors. Their opinions are, therefore, based on a variety of viewpoints and are reinforced by sufficient experience.

2.1.2. Analytic Methods and Results

The construction contract administration framework (CAPF) developed by [1] served as the source of the items used in this investigation. Then, as recommended by [43,75], this research integrates iterative PLS–SEM with ANN. A two-step statistical analysis was carried out as per [2,76] within the SEM analysis. The first phase involved using a measurement model, sometimes referred to as a factor confirmatory analysis (CFA), to iteratively evaluate the variables’ validity and reliability. Item loading, reliability and validity testing, and estimation of the outer loadings are some of the different criteria used to evaluate the CFA model. Using path modeling estimation, the study assumptions were then validated [80]. Following the PLS–SEM analysis, an ANN was used to demonstrate the strength and significance of the predictors. The data analysis was carried out using the Smart-PLS 3.3.3 program and the IBM SPSS software platform (version 26).

2.1.3. The Measurement Model

This paper deployed the CFA to conduct factor analysis and by this means ensured the construct validity and convergent validity. Refs. [80,81] suggested eliminating the MVs if elimination would increase the composite reliability (CR). Following this suggestion to increase the CR, the cut-off value from the outer loading was taken as 0.7 and the low-loaded MVs were iteratively eliminated. The results discussed below are relevant to the short model with 44 MVs.

Three criteria were used to evaluate the convergent validity: Cronbach’s alpha, composite reliability, and average variance retrieved [64]. The convergent validity of the MVs for the relevant LV is addressed by the α value, which determines if reliability was attained or not. A cutoff value of 0.70 or higher suggests stronger internal consistency, whereas a larger value of (range = from 0 to 1) indicates lower internal consistency [64,81]. To depict the reliability and internal consistency of MVs, CR is used to analyze composite reliability with a cut-off value over 0.60 [64]. For model confirmation, a value of 0.70 or higher is sufficient [80,81]. However, CR is viewed as a superior indicator of internal consistency because it considers MVs’ outer loadings, whereas α tends to overestimate reliability [81]. AVE analyzes the convergence of a group of items in an LV. It assesses the average commonality for each LV in reflective models and calculates the amount of variance captured by the LV from its related MVs due to the existence of measurement errors. The error variance outweighs the explained variance when the AVE value is less than 0.50 [81]. All the LVs in

Table 3. Results of measurement model (outer loadings, construct reliability, and validity).

Group	Code	Load	p-Value	α	rho_A	CR	AVE
G01	F01_02	0.796	0.000	0.846	0.848	0.847	0.580
	F01_03	0.781	0.000
	F01_04	0.741	0.000
	F01_10	0.726	0.000
G02	F02_01	0.764	0.000	0.844	0.846	0.845	0.577
	F02_02	0.772	0.000
	F02_03	0.779	0.000
	F02_05	0.721	0.000
G03	F03_01	0.764	0.000	0.845	0.845	0.845	0.576
	F03_06	0.783	0.000
	F03_07	0.740	0.000
	F03_09	0.750	0.000
G04	F04_04	0.774	0.000	0.863	0.863	0.863	0.611
	F04_05	0.810	0.000
	F04_06	0.786	0.000
	F04_08	0.756	0.000
G05	F05_03	0.801	0.000	0.868	0.868	0.868	0.621
	F05_04	0.780	0.000
	F05_07	0.785	0.000
	F05_10	0.787	0.000
G06	F06_01	0.806	0.000	0.853	0.855	0.853	0.593
	F06_02	0.800	0.000
	F06_03	0.727	0.000
	F06_04	0.743	0.000
G07	F07_01	0.754	0.000	0.855	0.857	0.855	0.597
	F07_03	0.744	0.000
	F07_04	0.770	0.000
	F07_05	0.820	0.000
G08	F08_02	0.791	0.000	0.872	0.872	0.872	0.629
	F08_03	0.774	0.000
	F08_04	0.811	0.000
	F08_05	0.796	0.000
G09	F09_02	0.752	0.000	0.854	0.855	0.855	0.595
	F09_03	0.778	0.000
	F09_04	0.805	0.000
	F09_06	0.749	0.000
G10	F10_01	0.648	0.000	0.840	0.857	0.843	0.577
	F10_02	0.849	0.000
	F10_03	0.851	0.000
	F10_04	0.665	0.000
G11	F11_03	0.817	0.000	0.894	0.894	0.894	0.679
	F11_04	0.841	0.000
	F11_08	0.838	0.000
	F11_11	0.798	0.000

Note: α = Cronbach’s Alpha; rho_A = Consistent Composite Reliability; CR = Composite Reliability; AVE = Average Variance Extracted.

met the requirements, with α values ranging from 0.84 to 0.894, CR values from 0.843 to 0.894, and AVE values from 0.576 to 0.679.

As a result, the examined parameters exceeded the acceptable limits. Additionally, the discriminant validity (DV) was performed using the Fornell–Larcker criterion. Since each component’s square root of AVE has a larger value than its correlation with any other construct, discriminant validity is confirmed. The variance inflation factor (VIF) was also used to test multicollinearity, which is the presence of two or more independent variables that are strongly correlated. The cutoff value for a well-fitted model is reached when VIF <= 4, which falls within the range of VIF between 1.602 and 3.217 [81]. According to [81], these measurements are regarded as goodness-of-fit metrics. This establishes the validity of the data for further investigation [81,82,83].

The reliability of MVs (expressed as standardized outer loadings of the manifest variables) explains the variance of individual MVs. In other words, the outer loadings express the absolute involvement of the MVs in the definition of its LV. The larger the values of the loadings are, the stronger and more reliable the measurement model will be [81]. According to [82], an outside loading of 0.7 or above is regarded as extremely excellent. As can be shown in Table 3, all MV loadings (except for F10_01 and F10_04) are greater than 0.7, which is a positive sign that CR was achieved and that the constructs were valid [81,82,83]. Additionally, the standardized root means square residual (SRMR) was used to assess how closely the suggested model fits the data. The variation between the implied and observed correlation matrices is measured by SRMR. Given that the SRMR value for the suggested model is 0.032 (<0.080) [81], it fits the data well.

2.1.4. The Structural Model

According to path coefficients and the coefficient of determination (R-Square), the structural model evaluates the relationship between LVs [64]. The R-square displays both the strength of the effect from exogenous LVs to endogenous LVs as well as the degree of explained variance of endogenous LVs. R-square also serves as a gauge of the structural model’s overall effect magnitude. The R-square values in this study ranged from 0.827 to 0.965, indicating that the model significantly influences all LVs [81]. The path model’s outcomes (standardized factor loadings, level of significance, t-values, and standard error) are shown in

Table 4. Path coefficient with values for the structural model.

Group	Description	Path Coefficient (PC)	R-Square	T-Value	Hypotheses
G01	Project Governance and Start-up	0.951	0.904	77.822	Accepted
G02	CA Team Management	0.955	0.912	69.140	Accepted
G03	Communication and Relationship	0.982	0.965	88.082	Accepted
G04	Quality and Acceptance	0.965	0.931	91.010	Accepted
G05	Performance Monitoring and Reporting	0.966	0.933	87.002	Accepted
G06	Document and Record	0.971	0.943	89.672	Accepted
G07	Financial Management	0.935	0.875	59.791	Accepted
G08	Changes and Changes Control	0.951	0.905	85.322	Accepted
G09	Claims and Disputes Resolution	0.939	0.881	59.526	Accepted
G10	Contract Risk Management	0.916	0.839	59.833	Accepted
G11	Contract Close-Out	0.909	0.827	60.830	Accepted

Note: R-square = coefficient of determination; T-statistics > 1.9 of significance at 5%.

and Figure 2.

All the standardized factor loadings have positive values, and all of them are statistically significant (t > 1.96). The larger the factor loadings, the more contribution the factor/group makes to the performance. Accordingly, G03-Communication and Relationship (0.982) was the most important LV that influenced AEC consultant behavior, followed by G06-document and record (0.971), G05-performance monitoring and reporting (0.966), G04-quality and acceptance (0.965), G02-CA team management (0.955), G01-Project governance and start-up (0.951), G08-changes and changes control (0.951), G09-claims and disputes resolution (0.939), G07-financial management (0.9350), G10-contract risk management (0.916), and G11-contract close-out (0.909). Nonetheless, the 11 LVs had a significant impact on AEC Consultant Performance. Also, all the F-square values had a high effect size (ranging from 4.771 to 27.347). Thus, the effect of dropping any LV from the model would be high (Garson 2016 [81]).

2.1.5. Neural Network Analysis

The ANN model was trained by a multilayer perception algorithm. In this study, the output layer had one output variable, such as the AEC consultant performance Evaluation Index (CPEI), whereas the input layer contained 11 LVs, which indicated the significant output from PLS–SEM. The investigation looked at a prototype network with from one to ten hidden nodes because it was impractical to use an exploratory approach to count the number of hidden nodes in the ANNs [43]. This study chose the multilayer perceptron approach per the advice of [43,75] and used Root Mean Square Error (RMSE) to assess the model’s performance through ten validations. A sigmoid function was selected as the activation function. Additionally, the covariant was standardized, and the inputs and outputs were both normalized [84]. The neural network’s architecture is depicted in Figure 3. The root means square error (RMSE) in the training dataset and the testing dataset was used to validate the ANN architecture.

RMSE compares particular datasets to forecast errors and is a scale-dependent indicator of prediction accuracy [85]. The perfect fit is attained if the RMSE value is 0.0, and close-to-zero values are preferred for model fit. The RMSE calculation is shown in Equation (1):

$$RMSE=\sqrt{{\displaystyle \frac{1}{N}}}\sum _{i=1}^n(X_{obs,i}-X_{model,i}{)}^2$$

(1)

where $X_{obs,i}$ represents the observed value, and $X_{model,i}$ represents the model value from the ANN analysis.

The selection of hidden layers and nodes is a difficult task that requires a trial-and-error approach. A common approach is to select a different number of layers and determine the RMSE and then decide about the number of nodes based on the RMSEs. A lower RMSE value indicates an improved data fit and higher forecasting accuracy. As can be seen in

Table 5. Root Mean Squared Error (RMSE) values for the hidden layers of the preliminary network.

Network Layer	RMSE
	Training	Testing
1	0.151	0.122
2	0.163	0.161
3	0.116	0.125
4	0.127	0.130
5	0.102	0.101
6	0.145	0.145
7	0.168	0.184
8	0.096	0.115
9	0.072	0.076
10	0.137	0.112
Average	0.128	0.127
Standard Deviation	0.031	0.031

, there were five hidden nodes, which were sufficient and complicated enough to draw the datasets without adding further errors to the overall model [43].

To prevent the ANN from being biased due to model overfitting, 90% of the data were used for training. The remaining information was used to assess how accurately the trained network predicted the future. The RMSE for the ten runs is shown in

Table 6. Ten-fold analysis of the neural network.

Net	Independent Variable Importance											RMSE
	G01	G02	G03	G04	G05	G06	G07	G08	G09	G10	G11	Training	Test
NN1	0.109	0.131	0.136	0.094	0.102	0.137	0.025	0.130	0.066	0.043	0.026	0.056	0.053
NN2	0.060	0.101	0.156	0.092	0.117	0.113	0.074	0.088	0.087	0.057	0.055	0.045	0.041
NN3	0.067	0.109	0.153	0.089	0.110	0.122	0.072	0.091	0.083	0.047	0.057	0.045	0.035
NN4	0.066	0.104	0.157	0.092	0.114	0.121	0.070	0.089	0.081	0.046	0.058	0.045	0.040
NN5	0.066	0.104	0.157	0.092	0.114	0.121	0.070	0.089	0.081	0.046	0.058	0.043	0.046
NN6	0.066	0.104	0.157	0.092	0.114	0.121	0.070	0.089	0.081	0.046	0.058	0.061	0.051
NN7	0.106	0.140	0.126	0.084	0.078	0.145	0.036	0.126	0.068	0.076	0.015	0.073	0.063
NN8	0.044	0.100	0.152	0.087	0.121	0.118	0.087	0.086	0.088	0.054	0.063	0.045	0.037
NN9	0.112	0.128	0.139	0.092	0.096	0.139	0.027	0.125	0.065	0.062	0.016	0.058	0.056
NN10	0.164	0.012	0.126	0.127	0.058	0.120	0.095	0.165	0.057	0.037	0.039	0.095	0.094
Mean	0.086	0.103	0.146	0.094	0.102	0.126	0.063	0.108	0.076	0.051	0.045	0.057	0.052
ST DV	0.036	0.035	0.013	0.012	0.020	0.011	0.025	0.027	0.011	0.011	0.019	0.017	0.017
NI	59%	71%	100%	64%	70%	86%	43%	74%	52%	35%	31%
Rank	7	4	1	6	5	2	9	3	8	10	11

ST DV = Standard Deviation; NI = Normalized Importance; and RMSE = Root Mean Square Error.

.

The training model’s average RMSE value is 0.057, whereas the testing model’s average value is 0.052. The training and testing model’s standard deviation was 0.017 as well. According to [73,85], accurate predictions were suggested by low levels of RMSE. As a result, the ANN model may be used to determine the numerical relationships between the CPEI and the predictors of AEC consultant performance.

The sensitivity analysis measures how much the network’s model output would be changed with different values of the inputs. Therefore, this paper utilized ten-fold runs to express the impact of changing the importance of the independent variables in forecasting the outputs as shown in Table 6. The percentage ratio of each predictor’s importance value to the largest importance value of LVs was used to calculate the values of normalized importance. In Figure 4, it is evident that G03-communication and relationship is the most significant predictor of AEC performance, followed by G06-document and record, G08-changes and changes control, G02-management of the ca team, G05-performance monitoring and reporting, G04-quality and acceptance, G01-project governance and start-up, G09-claims and disputes resolution, G07-financial management, and G10-contract risk management.

3. Discussion of Results

The AEC consultant heavily participates in planning, monitoring, and controlling the different construction process involving the management of quality, cost, schedule, safety, and environmental issues by a technical and specialized team [3,5]. The PLS–SEM model and ANN results show significant levels of importance for all indicators that influence the AEC consultant’s performance. Spearman Rank Correlation coefficient $(r$), which is a distribution free test, is employed to study the strength of correlation and to compare the different rankings by PLS–SEM and ANN using RII outcomes for each group [86,87]. The correlation coefficient ranges from +1 to −1, where +1 denotes an ideal positive relationship (agreement) and −1 denotes an ideal negative relationship (disagreement):

$$r=1-{\displaystyle \frac{6\sum d^2}{n^3-n}}$$

(2)

where $d$ represents difference between ranks assigned to variables for each cause, and n is the number of pairs of rank (d = 11).

The ranking of Contract Administration Groups using Spearman’s Rank Correlation Coefficient is shown in

Table 7. Correlation between PLS–SEM and ANN results on ranking of Contract Administration Groups using Spearman’s Rank Correlation Coefficient.

Group	ANN		PLS–SEM		d2
	Importance	Ranking	Load	Ranking
G03	0.146	1	0.982	1	0
G06	0.126	2	0.971	2	0
G08	0.108	3	0.951	7	16
G02	0.103	4	0.955	5	1
G05	0.102	5	0.966	3	4
G04	0.094	6	0.965	4	4
G01	0.086	7	0.951	6	1
G09	0.076	8	0.939	8	0
G07	0.063	9	0.935	9	0
G10	0.051	10	0.916	10	0
G11	0.045	11	0.909	11	0
∑d2=					26
r=					0.882
p-value=					0.000

along with the correlation between the PLS–SEM and ANN outcomes. At α = 0.05, n = 11, and a p-value less than 0.05, the Spearman’s rank correlation coefficient is 0.882, which is below the significance level. In addition, the rankings of the latent variables in contract administration show a statistically significant strong positive relationship between the PLS–SEM and ANN outcomes.

Among the AEC performance groups, G03-communication and relationship, and G06-document and record were found to have the most significant relationships with CPEI. For both LVs, the importance was confirmed by SEM and ANN analysis. However, the SEM and ANN results showed that the predictors of G09-claims and dispute resolution, G07-financial management, G11-contract close-out, and G10-contract risk management had the lowest relationship. For the rest of the predictors, the disagreement between PLS–SEM and ANN is referred to as the presence of non-linear relationships between the predictors and dependent variable (CPEI).

The results of this study show that combining PLS–SEM with an ANN offers a more thorough and accurate assessment of the work of AEC consultants. The hybrid PLS–ANN model ensures a more realistic depiction of consultant performance aspects by identifying both linear and non-linear interactions, in contrast to typical regression-based techniques that presume linear connections. The PLS–SEM results support the proposed associations, and the ANN efficiently ranks the most significant predictors, indicating that the two most important aspects influencing consultant performance are document and record management (G06) and communication and relationship management (G03). For project stakeholders, this hybrid method offers a framework for objective, data-driven evaluation that improves decision-making.

Additionally, by providing a quantitative ranking of performance indicators, the suggested approach outperforms earlier research and makes it possible to prioritize crucial areas for development. Building solid relationships with clients and stakeholders can greatly improve project outcomes, according to the ANN analysis, which also indicates that communication management accounts for the largest portion of overall consultant performance. The significance of structured information systems in AEC consulting is further supported by the fact that efficient document and record management lowers errors, disagreements, and project delays. By highlighting particular performance areas that can be enhanced by focused management techniques, our findings not only support earlier theoretical frameworks but also provide industry managers with useful takeaways.

During the construction phase, the consultant carries out several assignments. These assignments are interactive processes that require effective communication and successful relationships between the consultant and other stakeholders [43]. The predictor of G03-communication and relationship management is found to have the most significant effect with path coefficient (PC) = 0.982 and Independent Variable Importance (IVI) = 0.146. The result is consistent with recent studies that have revealed that it is a good indicator of the AEC consultant’s performance and project success [43,88]. Setting up efficient and effective communication amongst the project parties is regarded as one of the few measures of the consultant’s performance [6]. Also, open and effective communication has a positive impact on the supply chain parties [2]. However, formal communication channels are also necessary to disseminate and receive information from various stakeholders through meetings, reports, and other documentation [17]. Therefore, it is not surprising to find that G06-document and record management is the second LV that influences AEC performance with (PC = 0.971 and IVI = 0.126). Also, it is consistent with [43], who revealed that adequate management of this process was a good source of recent market information and knowledge that would help to improve business and performance capabilities. According to [88], information is a key aspect managed by the consultant with their managerial expertise. Refs. [1,16] argue that document and record management improves information sharing, maintains parties’ rights, and keeps all parties involved in the project, and, hence, it can improve communications. The ANN showed that G08-changes and changes control has the third most significant effect (PC = 0.951; IVI = 0.108). A similar level of importance has been presented by [17] when dealing with variations in cost control during the construction phase of the building project. Ref. [6] argues that developing an early change management protocol before the commencement of the project is a good measure for the mitigation of poor consultant performance. The fourth most significant effect is G02-team management. A high level of importance has been presented while dealing with the importance of experienced personnel to run the contract [11] and to be involved at the earliest time in a project [4,16]. Also, ref. [17] stated that allocating adequate and experienced resources when required to ensure effective implementation is important, and [25] concluded that the low level of experience of the consultant team is the top significant consultant-related factor causing project delay. Therefore, consultants should assign the right qualified and experienced team to supervise the project [40]. While the results present a moderate effect of G05-performance monitoring and reporting management (PC = 0.955; IVI = 0.103), controlling the overall project performance appears as one of the top roles of the consultant [4,16]. The results present another moderate level of importance for G04-quality and acceptance management (PC = 0.966; IVI = 0.102), which reflects the technical capability of the consultant. Ref. [3] highlighted the importance of having adequate technical capabilities for successful projects. Similarly, ref. [11] presented a moderate level of importance on the consultant’s performance while dealing with the supervision of contractors and the quality of deliverables. Ref. [25] ranked the consultant’s review and approval time of the project documentation as the second most significant factor causing project delays. A low level of importance has been identified with G09-claims and dispute resolution (PC = 0.939; IVI = 0.076), which matches the same order of ranking for the handling and settlement of claims by [16]. Still, early dispute management is vital for the completion of a successful project and to decrease the costs of contract administration [55]. The results revealed another low level of importance for G07-financial management (PC = 0.935; IVI = 0.063). A similar level of importance was reported in the literature for payment to the contractor and financial management [11,16]. The rank of G10-contract risk management (PC = 0.916; IVI = 0.051) appears to have one of the lowest levels of influence on consultant performance. The study by [11] also found that risk management is one of the least significant strategies that affect AEC performance. The result shows that G11-contract closeout management (PC = 0.909; IVI = 0.045) is the lowest-ranked process group that affects consultant performance. It is consistent with studies by [14], where the post-construction group was ranked as the group with the lowest level of influence affecting the consultant performance, and the study by [11], where final measurement and final account were also ranked low. One possible explanation is that the process comes immediately before the end of the project, and, thereby, any underperformance will not significantly affect the physical completion of the project. Therefore, during the closeout stage, some clients might tend to reduce the consultant’s staff to a minimum.

4. Assessment of Performance Through KPIs (Key Performance Indıcators)

To effectively evaluate performance levels, a key performance indicators (KPIs) system would prevent assessors from using their subjective opinions to interpret the indicators [13]. KPIs represent the precise, qualitative, and quantitative measurement of performance in the form of metrics of several items, which include financial and non-financial dimensions [11]. Therefore, after reaching the final model, a detailed list of the consultant’s KPIs was developed.

Table 8. Key performance indicators of the consultant performance evaluation.

KPI	KPI Definition	KPI Formula
Time Performance Indicator	The difference between the delayed task completion time (${\mathrm{T}}_{\mathrm{d}}$) and the contract’s or agreement’s specified completion time (${\mathrm{T}}_{\mathrm{c}}$) for this task for variable i.	${TP}_i=\left[1-{\displaystyle \frac{T_d-T_c}{T_c}}\right]\times 100$
Compliance Performance Indicator	The ratio of a variable i’s complied tasks (Na) to overall task count (Nt).	${CP}_i=\left[{\displaystyle \frac{N_a}{N_t}}\right]\times 100$
	The ratio of the total number of tasks (Nt) of a variable i to the number of tasks that were not completed because of CCA culpability (Ni).	${CP\prime }_i=\left[1-{\displaystyle \frac{N_i}{N_t}}\right]\times 100$
Customer Satisfaction Indicator	The maximum score rating (Rmax) of a variable i vs. the employers’ rating for CCA services.	${CS}_i=\left[{\displaystyle \frac{R_e}{R_{max}}}\right]\times 100$
Value (Cost) Performance Indicator	The value of tasks completed within budget (Va) in comparison to the total value of tasks (Vt) for a particular variable i.	${VP}_i=\left[{\displaystyle \frac{V_a}{V_t}}\right]\times 100,$
	The value of tasks that were performed outside of budget (Vi) because of CCA culpability versus the overall value of tasks (Vt) for a variable i.	${VP\prime }_i=\left[1-{\displaystyle \frac{V_i}{V_t}}\right]\times 100$

lists time, compliance, customer satisfaction, and value (cost) as performance indicators through six different formulas that can be applied to a majority of the 44 manifested variables as percentages (i.e., from 0 to 100). According to [2], performance is gauged mainly on time, cost, and quality as the main dimensions, which are considered in the proposed KPIs.

Using a predefined range for each performance level could inevitably reduce the assessor’s subjective judgment, maintain the evaluation process’ simplicity, and improve rating consistency across various assessors for other subjective KPIs [13]. Also, the AEC consultant’s performance is better differentiated if cost and schedule measures are included within the assessment criteria.

The suggested PLS–SEM–ANN model enhances conventional performance evaluation techniques by offering AEC consultants a sophisticated way to measure and evaluate KPIs. Through the integration of an ANN for ranking predictor importance and PLS–SEM for causal inference, this approach guarantees a more precise, data-driven evaluation of important performance indicators.

The model’s capacity to measure how important performance parameters affect KPIs is one of its main contributions. The results show that the two areas with the most effects on overall consultant performance are document and record management (G06) and communication and relationship management (G03). Better project outcomes are guaranteed by these two predictors, which have a direct impact on customer satisfaction and compliance rates. Similarly, the third most important aspect, changes and change control (G08), is essential to preserving cost-effectiveness and reducing conflicts.

The suggested PLS–SEM–ANN strategy allows for a predicted review of KPIs, allowing consultants and project managers to concentrate on the most significant areas for development. This is in contrast to standard KPI evaluation approaches, which frequently rely on manual tracking and subjective assessments. For instance, businesses can improve compliance and cut down on delays brought on by inadequate documentation management by streamlining documentation procedures after identifying document and record management as a high-impact predictor. Similar to this, better client–consultant connections can be achieved through the development of efficient communication techniques, which will raise customer satisfaction levels.

Instead of responding to performance issues after a project is over, this improved KPI assessment approach guarantees that AEC consultants can proactively address them. In construction consulting, the integration of AI-driven ranking and structural modeling enables a dynamic, adaptable framework for performance evaluation that enhances stakeholder engagement, project efficiency, and long-term sustainability.

5. Conclusions and Limitations

The performance of a consultant is influenced by a variety of elements; therefore, it is important to pinpoint the ones that are most important. The research model covered in this paper was developed through a questionnaire and is based on earlier research on project management, contract administration, expert judgment, and construction management. The proposed model was investigated using a sequential, multi-analytical technique that combined the study of the artificial neural network (ANN) and partial least square–structural equation modeling (PLS–SEM). The causal association between the AEC consultant performance and the AEC consultant performance assessment index was hypothesized and established using data from 336 industry surveys and PLS–SEM analysis. With very strong factor loading and a minimum value of 0.909 at a 0.05 significance level, the SEM analysis validates the eleven hypothesized positive correlations between performance groups and the performance of AEC consultants. The ANN has been used to help the determination of the most significant predictors of AEC consultant performance based on the significant input variables from the PLS–SEM analysis, as the judgments to assess AEC consultant performance may not always be linear. By using the Spearman Rank Correlation coefficient $(r$), the AAN shows general agreement and a strong correlation with PLS–SEM for ranking latent variables.

The results revealed that, for the AEC consultants to successfully adopt a good level of performance, it would be important to establish a communications and relationship management process and to ensure that document and record management are well designated and implemented. Regardless, the AEC consultant will have in place the other project management practices that include changes and changes control, performance monitoring, and reporting to be managed by a competent team. With the finding of this research and the significant importance observed through different indicators, it is recommended to make the AEC consultant the focal point of contact and a single point of responsibility to ensure the success of a project. Additionally, it is advised that the top management of consultants concentrate on enhancing project relationships and communication, thereby developing adaptable and efficient document management systems and efficiently using skilled individuals from a variety of backgrounds.

This research contributes to the construction management and project management areas in several ways. Although there has been some literature on performance assessment models/frameworks for construction projects, this research is one of the few studies that have established this performance at the project level from the perspectives of contractors, consultants, and clients. Secondly, unlike the previous studies that analyzed the proposed models by either a single or descriptive analysis approach, this research employed a multi-analytical approach (i.e., integrated PLS–SEM and ANN) to reach a sound conclusion on the most significant predictors affecting AEC performance. The SEM explains and validates the linear causal relationships in the research model. Also, the most significant predictors that have a substantial impact on operational performance in the AEC sector are found using PLS–SEM regression while an ANN is intelligent enough to predict both linear and non-linear relationships and the trained model can predict operational performance using the predetermined predictors. The ANN model is evaluated and tuned using cross-validation techniques to ensure its accuracy and generalization ability. The advantages of this integrated approach include the ability to handle complex and high-dimensional data, the ability to engage in accurate prediction, and the ability to identify the most important predictors. Therefore, the SEM–ANN approach can improve reliability and validity and provide an accurate prediction of the results while handling complex data. The findings can offer operational strategies to support decision-makers in improving AEC consultant performance and relevant process implementations.

This study’s limitation is that it only collected the specified indicators from traditional contract types and the post-awarding stage of building projects. Future works may focus on adding other useful manifested and latent variables to provide further insights into the AEC performance and use different contract types and alternative analysis techniques, such as compilation of SEM and Analytical Hierarchy Process (AHP) with Fuzzy Neural Networks (FNNs).

Future studies should concentrate on assessing how the suggested risk management enhancements within AEC projects are implemented in practice. Although statistical modeling and questionnaire data were used in this study, real-world construction projects should test the usefulness of the main conclusions in reducing risks and enhancing consultant performance. Longitudinal research on these metrics will shed more light on the treatments’ long-term effects on AEC consultant performance and project success.