Automated Document Verification for the Chamber of Commerce
Scenario
The Italian Chamber of Commerce system processes an enormous volume of administrative filings and documentary data on a daily basis. Ensuring the high quality and accuracy of registered information is a strategic priority for the entire system. While some checks are automated, many verifications still require meticulous manual intervention by operators. In this context, Deix collaborated with the chamber system to implement an innovative algorithm aimed at analysing filings related to changes in corporate officers. This type of filing, which accounts for approximately 11% of the total, is among the most common and requires significant cross-verification work, making it an ideal area for an optimisation project.
The challenge
The main complexity lay in the need to verify the consistency of information across heterogeneous data sources for each individual filing. The goal was to develop a system capable of simultaneously analysing and comparing: the request submitted by the user, unstructured documentary attachments (e.g. the resolution minutes), and the company register record already present in the system. The challenge was not only to extract key information, but above all to automatically identify any discrepancies, missing data, or inconsistencies, flagging them precisely for the operator to carry out a targeted review.
The solution
Drawing on our expertise in Generative Artificial Intelligence, we built a multi-agent solution leveraging our JINN development platform. By appropriately combining data extraction modules, semantic analysis of textual language, and customised comparison tools, the solution operates in two main phases:
Intelligent Information Extraction
An agent analyses the content of attached documents, such as minutes, to extract the fundamental information.
Consistency Analysis and Reconciliation
The data extracted from the documents is compared with what is declared in the filing and what is already recorded in the chamber register.
Detailed Attention Report
The final output is not a simple approval or rejection, but a detailed report highlighting situations requiring attention, allowing the operator to focus only on cases that need further investigation.
Results achieved
The solution introduced a strategic change in the verification process, generating value on multiple fronts: automation of the preliminary check drastically reduces filing management times, allowing operators to focus on higher-value activities; systematic and automated analysis minimises the risk of human error, ensuring a superior level of consistency and reliability of information in the business register; and the algorithm ensures that every filing is analysed according to objective and uniform criteria, strengthening the robustness of the entire verification process.
Future developments
The algorithm's architecture is modular and can be adapted to analyse other types of filings and documents, becoming a strategic asset for the digitisation of further processes.