Circularity and sustainability assessments aligned with the UN SDGs, Circular Economy (CE), and ESG metrics are fast-paced real estate due diligence requirements. Enhancing sustainability benchmarks involves extending building component lifespans to reduce CO2 emissions and waste. Integrating components into a closed loop requires efficient reuse potential evaluation amid dynamic market trends, policy shifts, and technological advancements. Traditional reuse assessments rely on scarce, slowly trained experts and lack scalable digital means to extend their expertise.

The main research question is whether it is possible to develop an AI-driven Reuse Viability Index (RVI) to measure and predict the sustainable viability of building component reuse, classifying components by their reusability, drawing upon resource assessments, open data, market history, and cost-effectiveness without continuous expert interaction.

The design brief for the RVI is to:

• Assess value and classify building components in existing and new buildings by reuse viability, guiding strategic deconstruction decisions for real estate owners, developers, and contractors.
• Be transparent, easy to implement, system-agnostic, and updatable to industry changes.

The RVI extends the conventional profit-margin by integrating multiple dimensions of key indicators - economic, environmental, technical, and social - into a single weighted composite score.

The methodology employs an AI-centred workflow:

1. To support the RVI, a Reuse Ontology, Universal Classification Adapter, and Classification Translator with a graph database have been developed to unify heterogeneous building component data from multi-country reference systems.
2. Mini-LLMs (e.g., all-MiniLM-L12-v2) with an OpenAI API fallback are used to aggregate data (prices, demand, technical details) from reuse marketplaces, CO2 databases, case studies and surveys. The same models define the RVI indicators and determine their weights using equal, dynamic and entropy-based (information theory) schemes.
3. ML models (CatBoost, XGBoost, NN, and LightGBM) learn from marketplace histories to establish and validate stable RVI score predictions, and accordingly classify building components.

The RVI outputs a ready-to-use percentage score of reuse viability that repurposes profit into a sustainability benchmark by integrating SDG and CE frameworks, addressing reuse barriers within market dynamics, and providing updatable guidance for circular real estate decisions.

An analysis of pan-European reuse marketplaces showed that windows, doors, and certain structural elements score high on the RVI; components with high deconstruction costs hover around 50%, while bulk items require adequate transport volumes to ensure environmental benefits.

Expert surveys confirm that component-specific RVI scores match industry assessments, and ML models deliver stable predictions for dynamic RVI indicators. If modelled carefully, basic component data - dimensions, quantity, condition, location, and price - predict reuse potential within a ±10-20% margin, with additional inputs further improving accuracy.

By integrating open data with semantic, graph-based, and ML methods, this thesis presents an AI-driven benchmarking and classification tool - the RVI - that quantifies and predicts building component reuse potential, unlocking added reuse value for real estate assets and informing EU-wide sustainability policy toward a climate-neutral 2050.