Applying circular economy principles to the construction sector demands structured data access, system interoperability, and predictive capabilities. An experimental project launched in Austria is seeking to overcome these barriers by leveraging an AI-driven infrastructure for managing construction site data.

Within the broader context of the ecological transition, construction remains one of the most environmentally impactful sectors. According to the United Nations Environment Programme (UNEP) – the UN’s coordinating authority for global environmental issues, which supports sustainable development through science, policy, and cooperation – buildings and infrastructure are responsible for approximately 37% of global energy-related CO₂ emissions (Global Status Report for Buildings and Construction, 2022).

That’s not the full picture. Construction and demolition waste (CDW) also accounts for nearly one third of all waste produced in Europe, with material recovery rates still varying significantly among member states (Eurostat, 2023).

Efforts to implement circular economy practices – such as reuse, selective disassembly, and design for deconstruction – are often hindered by a lack of systematic access to detailed information about the material composition of existing buildings.

A number of reports, including the Ellen MacArthur Foundation’s Reimagining Our Buildings and Spaces for a Circular Economy (2022) and Interreg Europe’s Sustainable and Circular Construction (March 2024), stress that for the built environment to truly embrace circularity, “it is essential to develop systems capable of tracking, analysing, and connecting materials throughout the entire lifecycle of a construction project”.

---

Takeaways

---

Regulatory Framework and European Objectives

European policy has long recognised the value of digitalisation in improving construction waste management and promoting material recovery. As early as the Waste Framework Directive (2008/98/EC), the EU established a minimum target of 70% recovery for non-hazardous construction and demolition waste – a goal reiterated and reinforced in the 2018 Circular Economy Package, which introduced stricter measures for waste prevention and reuse readiness.

In 2020, the EU Strategy for a Sustainable Built Environment identified digital material passports as a top priority. These tools are designed to log and share technical, environmental, and commercial information about building components – making them traceable and reusable. Central to this vision is the Digital Product Passport (DPP), a flagship initiative now in development under the Ecodesign for Sustainable Products Regulation (ESPR).

Operationally, the 2020 Circular Economy Action Plan highlighted construction as a priority sector, encouraging the development of digital, interoperable, and scalable systems to improve the efficiency of material and information flows across the built environment.

Implications for the Italian Context

Italy has progressively aligned with European directives, but implementation remains patchy. The National Strategy for the Circular Economy (SNEC), adopted in 2022, explicitly designates construction as a high-impact sector for transformation. Nonetheless, access to interoperable data is still limited.

The National Waste Management Plan (PNGR 2022–2028) notes that only a small portion of construction and demolition waste is reused as secondary raw material. Although the National System for Environmental Protection (SNPA) promotes the use of digital tools for tracking and reporting material flows, no binding regulatory framework currently mandates pre-demolition inventories – unlike in the Netherlands or Germany.

Italy’s adoption of Building Information Modelling (BIM), mandated for public works over €15 million since 2023 under the Public Procurement Code, is a significant milestone. Yet much of the data produced on construction sites still lies outside the BIM scope – dispersed across non-standardised, semantically unstructured formats – limiting its potential to support circular practices.

The Challenge of Unstructured Data and Revitalyze’s Experimental Proposal

This is the landscape in which Revitalyze, a Tyrolean startup based in Austria, is operating. The company has launched an applied research initiative to tackle a critical bottleneck: the dominance of unstructured data within construction site information flows. Contracts, floor plans, bills of quantities, meeting notes, and construction schedules are frequently stored in incompatible formats – PDFs, images, audio files – making systematic analysis virtually impossible when assessing a building’s circular potential.

As David Plaseller, co-founder and CEO of Revitalyze, explains: “Around 90% of the data generated in construction projects is either unstructured or not readily accessible in machine-readable form”. Revitalyze’s strategy centres on integrating a centralised data lake with an NLP (Natural Language Processing) pipeline that leverages existing large language models (LLMs) to automatically extract entities and relationships from construction documents.

The ultimate aim is to generate a contextual knowledge graph that dynamically represents both the logical and material structure of a building – enabling the evaluation of reuse potential or selective deconstruction opportunities during the demolition phase.

From Data to Action: Process Automation and Current Limitations

The project moves beyond data extraction, exploring the development of specialised AI agents – autonomous software tools designed to perform specific tasks based on project data. These include:

• Automatic generation of material inventories
• Assistance with preparing bids for public tenders
• Semantic analysis of contracts and technical documents
• Automated transcription and semantic indexing of site meetings

The platform has been tested in six pilot projects in collaboration with Austrian construction firms, under semi-experimental conditions. The user interface is designed to interact via natural language, but the model’s responses are strictly limited to the documentation corpus of the individual project – it does not access external datasets. This structure allows, in principle, for greater control over source validity and the traceability of reasoning.

As Plaseller notes, “AI agents are a first step toward semantic project management, but they remain dependent on the accuracy and completeness of the initial document corpus”. For now, the system enables only “weak” automation – focused on extraction, search, and reassembly of data. Yet the team is working on agents capable of executing operational decisions aligned with user-defined objectives, drawing inspiration from the emerging class of task-based agents in advanced Robotic Process Automation (RPA).

Open Reflections: Towards a Computable Understanding of the Built Environment?

The Revitalyze experiment demonstrates that circular construction challenges are not just about design – they are fundamentally informational. Efficient reuse of materials depends on structured, queryable data detailing presence, quantity, composition, and condition – information still largely absent from traditional construction workflows.

The fusion of general-purpose language models with domain-specific ontologies – an approach still in its early stages in the construction field – could unlock a new generation of tools for semantically modelling existing structures. Yet key challenges remain unresolved, including:

• The quality and reliability of source documents
• Governance of AI models
• Interoperability with BIM tools and enterprise document systems
• Regulatory and environmental assessments of recovered materials

In the absence of standardised digital representations of the built environment, initiatives like Revitalyze’s are mapping out largely uncharted territory. Here, AI does not replace technical expertise – but can serve as a powerful enabler, particularly during the most opaque and fragmented phases of the construction lifecycle.

Glimpses of Futures

Per comprendere To envision how artificial intelligence might drive circularity in construction, we can apply the STEPS framework – an analytical model exploring impacts across five dimensions: Social, Technological, Economic, Political, and Sustainable.

S – SOCIAL
Integrating intelligent systems on-site will redefine human roles – minimising repetitive tasks while requiring new skills in data management and interpretation. AI-driven circularity demands multidisciplinary training that blends construction expertise, environmental knowledge, and digital fluency.

T – TECHNOLOGICAL
As language models evolve and integrate with semantic infrastructures like knowledge graphs and ontologies, new conversational systems will emerge – able not only to interpret but to orchestrate complex operations. The challenge lies in ensuring interoperability, transparency, and accountability.

E – ECONOMIC
Tracking reusable materials and automating documentation can unlock significant efficiencies – lowering costs and accelerating workflows. However, broad scalability depends on standardising data formats and ensuring widespread access to high-quality, structured information.

P – POLITICAL
While EU policy aims to establish a digital construction ecosystem, regulation remains fragmented – caught between sustainability goals and weak enforcement. True impact will require alignment between environmental rules, procurement frameworks, and interoperable digital standards.

S – SUSTAINABILITY
A computable understanding of the existing building stock is essential for sustainability. AI can accelerate this shift – but only when supported by open data ecosystems, validated tools, and transparent AI governance.

---