Swedish municipalities and authorities need reliable and up-to-date information about the cultural heritage values of their building stock, but manual inventories are time-consuming and existing documentation is uneven. These values must be identified in municipal comprehensive plans, yet few municipalities have carried out the inventories this requires. This project develops and quality-assures AI-assisted methods for describing buildings and supporting cultural heritage assessment using street-level imagery, building registers and existing heritage records. The work forms part of the Vinnova-funded project "Inventory of Cultural Heritage Values Using AI - for Enhanced Preparedness and Resilience" (2026-00968, June 2026-June 2029), coordinated by RISE in collaboration with Gothenburg City Museum and the Swedish National Heritage Board. Gothenburg provides the first case study and expert reference assessments. The National Heritage Board supports scaling up the method to the rest of Sweden. We will evaluate open-weight vision-language models for extracting structured building attributes, such as facade materials, architectural features and signs of alteration (e.g. replaced windows or later renovations). Building on this, we will investigate how these observations can support expert assessment of cultural heritage values. The research will evaluate models, prompting strategies and inference settings with expert-reviewed reference data and existing model-based workflows. The analyze will address attribute-level accuracy, consistency, uncertainty, structured-output validity and computational efficiency, including variation across building types and image quality. AI-generated descriptions will support human review rather than replace expert heritage judgements. The initial NAISS allocation will support reproducible model evaluation and inference runs on selected datasets, establishing the quality and resource requirements for larger-scale studies. The wider project also explores the structuring and georeferencing of older heritage documentation. Expected outputs include evaluated methods, reusable tools and documented workflows, disseminated through scientific publications and project partners, with code and data shared where licensing and data protection permit. The long-term aim is to improve municipal planning, careful renovation and preparedness through more accessible digital knowledge of the built heritage.