Artificial intelligence in the construction industry: A review of present status, opportunities and future challenges

The paper addresses the persistent productivity and digitization challenges in the construction industry—cost/time overruns, safety incidents, low productivity, and labor shortages—highlighting that construction remains one of the least digitized sectors. It posits Artificial Intelligence (AI) as a transformative technology, already successful in other industries, whose subfields (machine learning, computer vision, robotics, knowledge-based systems, optimization, NLP, automated planning/scheduling) can help tackle construction-specific problems. The study frames three research questions: (1) What are the areas of AI application in the construction industry? (2) What are the future opportunities for AI application? (3) What challenges affect AI adoption? The objectives are to critically review AI applications and techniques used in construction, identify opportunities for increased application, and delineate adoption challenges. The work’s importance lies in synthesizing six decades of research to guide stakeholders on pathways to AI implementation and to address barriers limiting sector-wide benefits.

The review covers foundational AI concepts (types: ANI, AGI, ASI; components: learning, knowledge representation, perception, planning, action, communication) and major subfields applied to construction: machine learning (supervised, unsupervised, reinforcement learning, deep learning), computer vision, robotics, knowledge-based systems (expert systems, case-based reasoning, intelligent tutoring, intelligent interfaces/linked systems), natural language processing, optimization (including evolutionary algorithms), and automated planning/scheduling. It summarizes prior applications across construction domains such as health and safety monitoring, cost estimation, supply chain/logistics, risk prediction, site monitoring and performance evaluation, offsite assembly, materials/plant management, tender evaluation, conflict resolution, sustainability assessment, and waste management. Trend analysis indicates optimization has historically dominated AI-in-construction research, with machine learning surpassing knowledge-based systems in the last decade; robotics is rising (3D printing, exoskeletons, UAVs); NLP remains least studied. The review also integrates related emerging technologies—IoT, blockchain, cybersecurity, quantum computing—and their intersections with AI and BIM in construction.

An extant literature review was conducted using SCOPUS as the primary source, validated against IEEE, ACM, and ScienceDirect. The timeframe spanned 1960–2020 to capture six decades since early AI research in the 1950s. Searches used 29 free-text keywords covering AI subfields and techniques (e.g., robotics, computer vision, machine learning, expert/knowledge-based systems, optimization types, NLP, deep learning variants, evolutionary algorithms, automated planning/scheduling) combined with "Construction Industry" using advanced search. Inclusion criteria: English-language publications describing or evaluating an AI subfield/technique delivering a practical construction application, determined via abstract/title or full-text when unclear. For result sets exceeding 100 items in mature domains (e.g., optimization, KBS), conference papers were excluded on the rationale that many were later journalized. Data extracted per article included application area, techniques used, and findings. Of approximately 1800 assessed publications, 1272 were deemed relevant and included for further investigation.

- Over 60% of AI-in-construction research occurred in the last decade. Optimization has been the most persistent focus area; machine learning surpassed knowledge-based systems recently; robotics interest is growing (e.g., 3D printing, UAVs, exoskeletons); NLP is the least explored subfield in construction. - Mapped AI subfields to construction applications (e.g., ML for safety and scheduling; computer vision for inspection and cost estimation; planning/scheduling for logistics and control; robotics for offsite assembly and material handling; KBS for tendering, risk, waste; NLP for interfaces; optimization for planning, design, risk, energy, sustainability). - Synthesized advantages and limitations by subfield: common benefits include cost/time savings, productivity and safety gains, and accuracy; common constraints include incomplete data, high deployment costs, knowledge acquisition/validation issues, scalability/computing power needs, and integration complexity. - Identified 14 subdomains for AI applications with state-of-the-art and opportunities: resource and waste optimization; value-driven services (estimation/scheduling, site analytics, job creation); AI/BIM with IoT, smart cities, AR, blockchain, QC; supply chain management; health and safety analytics; contract analytics; voice user interfaces; AI-driven audit systems for construction financials. - Proposed a maturity perspective: computer vision, robotics, and NLP as emerging; ML and automated planning/scheduling as ripe; KBS and optimization as mature within construction research. - Highlighted integration opportunities with IoT, BIM, blockchain, cybersecurity, and quantum computing to enhance transparency, trust, real-time analytics, and computational efficiency.

The review directly addresses the research questions by cataloging where AI is being applied in construction, outlining future opportunities, and detailing adoption challenges. It shows that while certain subfields (optimization, KBS) are mature within construction, more recent advances (deep learning, modern computer vision, NLP) are underutilized relative to their potential. Findings underscore the importance of integrating AI with BIM and IoT for real-time site analytics, safety prediction, and waste minimization, and with blockchain to improve trust and transparency in supply chains and contracts. The discussion emphasizes that sector-specific constraints—unique sites, fragmented processes, cultural resistance—necessitate explainable, secure, and adaptable AI systems. It recommends explainable AI to build trust, adversarial ML and cybersecurity-by-design for robustness, and workforce development to address talent shortages. Overall, the synthesis provides a roadmap for targeted AI adoption in high-impact areas (safety, SCM, estimation/scheduling, waste/resource optimization, contract analytics, financial auditing) and for leveraging allied technologies to overcome traditional barriers.

The study consolidates six decades of AI applications in construction, explaining AI concepts/subfields and mapping them to construction challenges, with a synthesis of benefits and limitations. It reveals rapid growth in the past decade, a shift toward machine learning, and emerging interest in robotics, while NLP remains under-explored. Contributions include: (1) a comprehensive state-of-the-art review across AI subfields; (2) identification of application domains and future opportunities (e.g., AI-driven waste analytics, deep learning for cost/time estimation, holistic site analytics, AI-enabled SCM, contract comprehension, VUIs, financial audit systems); (3) articulation of key challenges (culture/trust and XAI, security/adversarial ML, talent, high initial costs, ethics/governance, computing power/connectivity). Future research directions include constructing domain-specific datasets (e.g., contracts, voice commands for sites), developing holistic, cloud-powered analytics platforms integrating BIM/IoT/AI, robust and explainable models for safety and finance, secure AI-blockchain solutions for SCM and contracts, and exploring quantum computing to optimize AI solutions.

- Language and source constraints: limited to English-language publications; SCOPUS used as the primary database with validation via IEEE, ACM, and ScienceDirect—potential omission of relevant non-English or non-indexed works. - Selection constraints: for large result sets (e.g., optimization, KBS), conference papers were excluded, potentially omitting cutting-edge findings not yet journalized. - Timeframe: coverage restricted to 1960–2020; developments post-2020 are not captured. - Methodological scope: qualitative synthesis without meta-analytic effect size estimation; relies on reported findings which may vary in methodological rigor. - Domain variability: heterogeneous construction contexts and unique site conditions limit generalizability of some AI applications across projects.