INSTRU imports, maintains and calibrates laboratory instrumentation for sectors such as petrochemicals, and had built up years of technical service documentation — work sheets, calibration reports, protocols and regulations — scattered and hard to consult. Before investing in a full tool, it wanted to validate whether AI could classify that document archive and answer questions about it with sufficient reliability.
We designed and ran a scoped proof of concept: automatic document classification by type (work sheet, report, protocol, certificate) with metadata extraction using language models, and a natural-language query interface over a representative sample of the documentation.
The proof of concept, validated with the client, made it possible to assess on their own documentation what was feasible and with what degree of reliability, before committing to a larger investment.
For confidentiality reasons, some details in this case are presented in anonymised or aggregated form. Images are illustrative and may not show real product screenshots.
We use cookies
We use cookies to analyse site usage (Google Analytics, Microsoft Clarity) and improve your experience. Essential cookies do not require consent.
