AI-Assisted Analysis
AI-assisted inspection with inspectors in control.
AI-assisted inspection, with inspectors in control. PFI-ai uses built-in models to detect common façade defects and privacy-sensitive content. Configurable inference runs on scalable infrastructure, giving users control over model selection and processing settings. Before findings are included in reports, inspectors review confidence levels, correct classifications and remove false positives. Reviewed annotations become labelled data for quality assurance and continuous model improvement. Every detection and annotation remains fully traceable, recording whether it was created by a user or model together with its associated metadata.

As of October 2026, available ML models detect people, vehicle licence plates, cracks, corrosion and spalling. Models are continuously improved, with new detection types and versions released quarterly. Users can select the models and versions deployed for each inspection.