Digital transformation in Central European construction: AI adoption, performance, barriers

Artificial intelligence has been billed as the next great leap for the construction industry, promising everything from self-optimizing schedules to generative design and autonomous safety monitoring. But a new study from Central Europe suggests that the road to AI on the building site runs directly through a technology that most firms already have: building information modelling, or BIM. The research, published in Mobile Networks and Applications, finds that the firms best positioned to adopt artificial intelligence are not the largest or richest ones, but those with the deepest digital maturity—and that the biggest obstacles to AI adoption are people and skills, not machines.

The study was carried out by Tomáš Mandičák, Katarína Krajníková and Peter Mésároš of the Faculty of Civil Engineering at the Technical University of Košice in Slovakia, together with Ivan Marović of the Faculty of Civil Engineering at the University of Rijeka in Croatia. It addresses a long-standing blind spot in the literature on construction digitization. While wealthy markets in Western Europe and North America have been studied extensively, comparatively little is known about how firms in smaller Central European economies—and especially small and medium-sized enterprises, which dominate the regional construction sector—are progressing along the digital pipeline from computer-aided drafting, through BIM, toward AI.

The researchers posed three questions. First, how widely is BIM used across the different stages of construction projects? Second, does that use actually show up in measurable performance outcomes such as cost, waste and sustainability? And third, what stands in the way of firms moving from BIM to artificial intelligence? To answer them, the team ran two surveys grounded in official statistical records. One covered construction firms in Slovakia, Croatia and Slovenia, three neighboring markets at visibly different stages of digital development. The other focused specifically on Slovak firms’ experience with AI and the barriers they encounter.

The picture that emerges on BIM maturity is strikingly uneven. On a five-point scale of BIM maturity, the mean score was 1.33 in Slovakia, 2.49 in Croatia and 4.33 in Slovenia—a spread so large that the differences are statistically unambiguous. The researchers tested the country differences with the Kruskal-Wallis test, a non-parametric method suited to comparing more than two groups when the underlying distributions cannot be assumed to be normal. The result, a test statistic of H = 86.77 with a p-value below 0.001, confirms that the gaps between the three national samples are not a product of sampling noise. Slovenia’s firms, in other words, operate at a level of BIM integration that their Slovak counterparts have yet to approach, with Croatia occupying a middle ground.

That unevenness matters because BIM maturity turned out to be strongly associated with real-world performance. Using Spearman rank correlation analysis, the team found that firms reporting heavier and more sophisticated BIM use also reported better outcomes across cost, material waste and sustainability indicators. The correlations are remarkably strong by the standards of survey research: up to 0.93 for material cost performance and 0.92 for emission reduction. A Spearman coefficient close to 1 would indicate a nearly perfect monotonic relationship, so values in the low 0.9s suggest that, within this sample, the link between BIM depth and reported performance is about as tight as such links ever get. Firms that model buildings as coordinated, data-rich digital objects appear to waste less material, spend less and cut emissions more effectively than firms that do not.

The sustainability finding was subjected to an additional check. The researchers cross-validated the BIM–sustainability relationship against a convergent model, a technique used to confirm that the correlation observed in the survey data is not an artifact of a single analytical specification. That the relationship survived this test strengthens the case that BIM is not merely correlated with green performance by coincidence—that the structured, information-rich workflows that BIM enforces genuinely feed into how efficiently materials are used and how much carbon a project embodies. This matters because construction remains one of the world’s largest sources of waste and emissions, and tools that measurably shrink a project’s material and carbon footprint carry enormous aggregate value.

Against this backdrop, the state of AI adoption is sobering. Genuine, operational use of artificial intelligence among the surveyed firms remains rare. This is not for lack of conceivable applications; the research literature describes machine-learning systems that generate construction schedules from BIM models, computer-vision platforms that monitor site safety in real time, digital twins that fuse sensor data with building models to optimize facility operations, and explainable AI methods designed to make algorithmic decisions legible to engineers and clients. The gap, it appears, lies between what the technology can do and what the typical Central European firm is equipped to absorb.

The most consequential finding of the study concerns what separates AI-ready firms from the rest. The researchers examined which firm characteristics correlated with AI readiness and found that digital maturity—not company size—was the decisive factor. The correlation between AI readiness and the use of digital tools was 0.49, and between AI readiness and broader digitalization 0.37, both meaningful associations. Company size, by contrast, showed a correlation of just 0.01, which is statistically indistinguishable from zero. In plain terms, a small firm with a deep digital foundation is a far more plausible AI adopter than a large firm still working with disconnected documents and 2D drawings. The implication for policy and industry strategy is that simply scaling up or subsidizing large players will not accelerate AI diffusion; building the digital base across the many small and medium-sized enterprises will.

When firms were asked what actually blocks AI adoption, the answers clustered overwhelmingly around human and organizational factors rather than technical ones. Shortages of qualified staff, the cost and difficulty of training an existing workforce, uncertainty about the return on investment, and managerial unfamiliarity with what AI can realistically deliver were cited far more often than problems with the technology itself. This pattern echoes a long line of technology-adoption research, from the classic Technology Acceptance Model onward, which has repeatedly shown that perceived usefulness and perceived ease of use—both fundamentally human judgments—govern whether organizations embrace new tools. It also aligns with the Technology–Organization–Environment framework, in which the technological context is only one of three forces shaping adoption, alongside organizational readiness and the external environment.

The authors frame their results as a dose of realism for an industry often swept up in AI hype. The findings, they write, point to a solid digital and BIM base as AI’s realistic starting point. Firms hoping to deploy machine learning for schedule optimization or cost prediction first need the structured digital data that BIM produces; without it, there is nothing for the algorithms to learn from. The study’s regional focus also serves as a reminder that digital transformation is not uniform even within a single corner of Europe. A construction firm in Ljubljana and one in Bratislava may face the same global technology frontier, but they start from very different places on the maturity curve.

For Central European policymakers, the study arrives with concrete signals. The strong BIM–performance link suggests that continuing investment in BIM adoption pays off not only in productivity but in the sustainability outcomes that EU decarbonization policy increasingly demands. And the finding that barriers are concentrated in people and skills implies that the highest-leverage interventions may be education and workforce development—curricula that combine civil engineering with data science, targeted training programs for SMEs, and support for managers to understand what AI can and cannot do—rather than hardware subsidies or software licenses. The project was supported by the Slovak Research and Development Agency under contracts APVV-22-0576 and APVV-17-0549, both aimed at researching digital technologies for sustainable construction.

The wider lesson extends beyond the three countries studied. As the construction industry worldwide grapples with stagnant productivity, chronic waste and mounting pressure to decarbonize, the temptation to leapfrog straight to AI is understandable. But this study suggests that there are no shortcuts: the firms that will benefit from artificial intelligence are the ones that have already done the slower, less glamorous work of digitizing their models, their data and their workflows. In construction, it seems, the future arrives one maturity level at a time.