Performance Expectancy and Support Drive AI Adoption in Indian Construction

In an empirical investigation into digital transformation across the Indian building sector, researchers have identified that perceived operational gains, workplace social dynamics, and general awareness serve as the primary drivers behind professionals’ intentions to embrace artificial intelligence. The quantitative study, published in Discover Civil Engineering by investigators Nuvotalu Ringa, Himmi Gupta, and Harsh Vardhan Samalia, examined how practitioners appraise emerging computational tools within an expanding infrastructure economy. By adapting the Unified Theory of Acceptance and Use of Technology framework to incorporate sector-specific dimensions, the authors evaluated whether conventional barriers such as software costs, procedural difficulty, and perceived risks systematically impede digital adoption. Their analytical results demonstrated that anticipated performance benefits exert the strongest positive influence on adoption intentions, while actual implementation on active construction projects is governed primarily by organizational facilitating conditions and technical infrastructure rather than personal reservations regarding financial expense or complexity.

The context of the investigation reflects broader macroeconomic dynamics occurring throughout India, where massive infrastructure development and urban expansion coincide with persistent operational challenges. The authors documented that despite India maintaining one of the largest construction markets globally, projects across the nation routinely encounter cost overruns, severe project delays, inefficient resource distribution, and coordination deficits among distributed project stakeholders. Digital systems and machine learning applications have been positioned as viable technical solutions capable of mitigating these recurrent difficulties through automated progress tracking, predictive project analytics, algorithmic risk evaluation, safety monitoring, and enhanced cost forecasting. However, the researchers emphasized that the practical deployment of artificial intelligence tools across architectural and engineering firms remains uneven, indicating that technological capability alone does not ensure widespread operational uptake. This observed disconnect between theoretical utility and practical implementation established the core motivation for investigating the organizational, psychological, and behavioral determinants governing professional acceptance.

To rigorously assess these adoption patterns, the research team adapted the Unified Theory of Acceptance and Use of Technology, an established theoretical baseline originally formulated by Viswanath Venkatesh and colleagues by synthesizing components from eight earlier technology acceptance frameworks. The classical UTAUT architecture posits that four core constructs—performance expectancy, effort expectancy, social influence, and facilitating conditions—determine individual behavioral intentions and ultimate usage behavior, frequently modulated by demographic traits such as age, gender, and experience. Recognizing that cutting-edge data-driven technologies introduce distinct challenges within emerging economies, the researchers systematically extended the conventional framework by introducing four supplementary predictor variables: awareness, trust, perceived risk, and implementation cost. Awareness captures the extent of practitioner familiarity with AI capabilities; trust reflects confidence in system reliability, algorithmic accuracy, and data security; perceived risk addresses anxieties regarding performance failure, cybersecurity vulnerabilities, and prospective employment displacement; and cost accounts for expenditures on software licenses, technological infrastructure overhauls, personnel training, and ongoing system maintenance.

The methodological framework relied on an online survey administered to 450 construction professionals distributed throughout multiple administrative territories in India, including states such as Rajasthan, Nagaland, Chandigarh, Haryana, Telangana, Bihar, Uttar Pradesh, Himachal Pradesh, Maharashtra, Delhi, and Uttarakhand. Using a convenience sampling strategy aimed at individuals actively participating in technical and project management decisions, the authors assembled 121 valid and fully completed responses, reflecting a final response rate of approximately 26.9 percent. The measurement instrument comprised 34 individual items evaluated on a five-point Likert scale ranging from strongly disagree to strongly agree. Demographically, the resulting cohort skewed heavily toward male respondents, who represented 81.0 percent of the sample, with 62.8 percent falling within the 26 to 35 age bracket. Furthermore, 57.9 percent possessed postgraduate academic qualifications, and 53.7 percent identified professionally as engineers, complemented by representation from project managers, architects, and academicians across varying tenures of professional experience.

Data processing conducted in IBM SPSS Statistics followed a structured statistical sequence encompassing frequency distributions, descriptive metrics, reliability estimations, normality diagnostics, bivariate correlation analyses, and multiple linear regression modeling. Multi-item scale reliability assessments yielded Cronbach alpha values spanning from 0.715 to 0.924, confirming robust internal consistency across the primary composite constructs. Facilitating conditions exhibited a reliability coefficient of 0.912, actual use behavior registered at 0.924, and behavioral intention reached 0.901, while the awareness scale demonstrated an acceptable exploratory reliability of 0.715. Trust and perceived risk were evaluated via single-item indicators to curtail survey fatigue, precluding Cronbach alpha calculation for those specific parameters. Subsequent normality testing through the Shapiro-Wilk procedure produced statistically significant p-values below the 0.05 threshold for all variables, demonstrating departures from normal distributional curves. Nevertheless, the authors retained parametric linear regression methods based on methodological precedents establishing the robustness of regression techniques against moderate distributional violations in empirical datasets exceeding 100 observations.

Initial Pearson correlation assessments revealed that behavioral intention shared statistically significant positive relationships with performance expectancy (r = 0.571, p < 0.001), effort expectancy (r = 0.502, p < 0.001), social influence (r = 0.529, p = 0.001), cost (r = 0.428, p < 0.001), awareness (r = 0.398, p < 0.001), and trust (r = 0.207, p < 0.05). Conversely, perceived risk demonstrated an insignificant bivariate correlation with adoption intention (r = 0.142, p > 0.05). Among the predictor variables, notable correlations occurred between performance expectancy and effort expectancy (r = 0.680, p < 0.001) as well as between effort expectancy and social influence (r = 0.604, p < 0.001). To ensure these intercorrelations did not compromise the analytical integrity of the subsequent predictive equations, multicollinearity diagnostics were computed. All variance inflation factor values remained substantially below the conventional diagnostic threshold of 5, ranging from 1.14 for perceived risk up to 2.84 for effort expectancy, with corresponding tolerance values between 0.35 and 0.88, thereby confirming that collinearity did not distort the estimated regression parameters.

The two-stage multiple linear regression analysis yielded revealing distinctions between variables that shape adoption intentions versus those governing actual practical usage. In the first regression model evaluating predictors of behavioral intention, the overall equation proved statistically significant (F = 13.671, p < 0.001), accounting for 45.9 percent of the observed variance (R² = 0.459, adjusted R² = 0.425). Performance expectancy emerged as the primary determinant of intention (standardized beta = 0.403, p < 0.001), followed by social influence (beta = 0.280, p = 0.003) and general awareness (beta = 0.170, p = 0.036). In contrast, effort expectancy exerted no statistically meaningful effect within the multivariate model (beta = -0.013, p = 0.912). Similarly, cost (beta = 0.111, p = 0.198), perceived risk (beta = -0.048, p = 0.515), and system trust (beta = -0.102, p = 0.220) failed to achieve statistical significance. Consequently, formal hypotheses linking performance expectancy, social influence, and awareness to behavioral intention were supported, whereas proposed negative or positive dependencies on effort, cost, trust, and risk were not substantiated by the data.

The secondary regression model, structured to evaluate the drivers of actual use behavior, likewise demonstrated robust explanatory power (F = 61.343, p < 0.001), accounting for 51.0 percent of the total variance (R² = 0.510, adjusted R² = 0.501). Within this specification, facilitating conditions functioned as the dominant empirical predictor of genuine artificial intelligence utilization across construction operations (beta = 0.553, p < 0.001). Individual behavioral intention also maintained a statistically significant positive influence on actual system deployment (beta = 0.291, p < 0.001). These empirical findings verified both proposed operational hypotheses, demonstrating that while psychological intention provides the necessary motivational orientation, the presence of institutional facilitating conditions—including accessible computing hardware, dedicated technical assistance, reliable internet access, and formal enterprise training—constitutes the decisive operational bridge that translates internal willingness into executed professional practice.

In interpreting the absence of statistically significant effects for effort expectancy, cost, trust, and perceived risk, the authors highlighted multiple contextual and methodological explanations. The non-significant beta for effort expectancy, despite its notable bivariate correlation with intention, stems partly from its shared variance with performance expectancy, suggesting that practitioners who anticipate substantial operational enhancements inherently expect to navigate technical learning curves without viewing usability as a primary deterrent. Furthermore, the demographic composition of the sample, which was predominantly composed of younger and postgraduate-educated technical personnel, likely contributed to diminished sensitivity toward software interface difficulty. The researchers also noted that respondents appeared to evaluate AI platforms as strategic corporate investments where anticipated operational returns in scheduling accuracy and defect mitigation eclipse immediate software license expenses or apprehensions regarding technological uncertainties at the preliminary stage of intention formation.

Based on these findings, the authors outlined targeted practical recommendations for construction enterprises, professional engineering associations, and public infrastructure agencies seeking to accelerate digital modernization. Because performance expectancy constitutes the primary driver of adoption intent, construction firms should avoid generic promotional claims and instead execute bounded, project-embedded pilot demonstrations—such as deploying automated defect recognition or predictive scheduling algorithms on active job sites—to document tangible productivity and rework metrics. To leverage the pronounced impact of social influence, professional engineering bodies can establish peer-learning forums, site visits, and formal tool endorsements that showcase peer success. Finally, addressing the decisive role of facilitating conditions requires government and industry alliances to establish shared regional digital infrastructure, high-speed remote site connectivity, and subsidized training initiatives tailored specifically for small and mid-sized contractors who lack internal capital to develop proprietary technical support ecosystems independently.

The investigators explicitly outlined several methodological limitations that bound the generalizability of their conclusions. The achieved sample size of 121 respondents, while statistically suitable for linear regression modeling, restricts broad extrapolation across India’s highly fragmented construction landscape, which ideally warrants cohorts of 150 to 200 participants or more. Additionally, the sample drew heavily from convenience networks and exhibited geographic clustering in specific states like Rajasthan and Nagaland rather than mirroring the proportional economic distribution of national construction volume. The reliance on single-item metrics for complex constructs such as trust and perceived risk represents another acknowledged measurement constraint, preventing the capture of multidimensional risk facets such as financial liability, algorithmic opacity, and data privacy. Future research avenues recommended by the authors include deploying comprehensive multi-item measurement scales, engaging broader cohorts that incorporate senior executive decision-makers, and testing potential moderating interactions across diverse developing economies.

Subject of Research: Determinants of artificial intelligence adoption in the Indian construction sector using an extended UTAUT model

Article Title: Determinants of artificial intelligence adoption in the Indian construction industry using the UTAUT model

Article References: Ringa, N., Gupta, H., & Samalia, H. V. (2026). Determinants of artificial intelligence adoption in the Indian construction industry using the UTAUT model. Discover Civil Engineering, 3(1), Article 239. https://doi.org/10.1007/s44290-026-00635-3

Image Credits: AI Generated

DOI: 10.1007/s44290-026-00635-3

Keywords: artificial intelligence, Indian construction industry, UTAUT model, technology acceptance, performance expectancy, effort expectancy, social influence, facilitating conditions, behavioral intention, use behavior, civil engineering, construction management, digital transformation, regression analysis, predictive analytics, project management