A construction company implemented an automated system within Bitrix24 to reconcile supplier invoices against project estimates. The goal is to detect cost overruns, such as price hikes or excess material usage, before payments are processed.
The system processes invoices via OCR and language models to extract data. It validates calculations internally before comparing items against the estimate. To handle discrepancies in units of measurement, the system uses a coefficient directory to convert values into a unified standard.
The software normalizes product names to match items in the estimate, using AI to suggest links. It learns from user confirmations to improve accuracy over time. The system tracks cumulative volumes and flags deviations based on predefined business thresholds.
When an invoice exceeds established limits, it is routed to a manager for approval. The tool does not automate payments but provides data to support decision-making, aiming to prevent the accumulation of small, unnoticed overruns.
