Presentation
An elevator maintenance company receives inspection reports from nearly twenty independent inspection bodies, each with its own PDF layout and some only as scans, and its staff keyed every observation in by hand before assigning the work. SEADEV built a tool that recognises which body issued a report and extracts its general details and its list of observations, using a template for each layout. When a layout is not yet automated or the report is a scan, the user outlines the relevant areas on the page and the text is read by OCR. Each report is linked to the right site by its address, and each observation is compared with similar past ones to suggest whether the customer or the maintainer must act, which team should handle it and whether a quote is needed. The observations then reach the company's maintenance dashboard as structured data instead of retyped text.
- Report extraction (2021 – 2022) recognises which of 19 inspection bodies issued a PDF report, extracts its details and observations with one template per layout and an OCR-assisted mode for scans, and suggests who acts on each observation.
- Report intake (from 2024) reads incoming e-mail and files the PDF attachments by recognising a report's layout from its first page.
- AI extraction (from 2024) extracts the report data with a large language model.
On the Platform This software Its client Its service
