Top Financial Institution managing thousands of credit facilities across LATAM
A critical operation for our customer is the routine audit of credit facilities. This historically involved manually verifying loan files (e.g. contracts, invoices, insurance documents) against loan tapes to ensure compliance with credit covenants. This manual process, requiring more than 20 auditors, was costly, time-consuming, and lacked traceability.
Auditing 100 files used to take an auditor ~3 days, creating a growth bottleneck for our customer’s operations.
In 10 weeks, our customer automated 80% of its audit operation by leveraging Arkham’s unstructured data management, process orchestration, and Generative AI capabilities. With our platform, 100 files can now be audited in 3 hours, effectively increasing the capacity of the audit team threefold.
As illustrated in the figure above, the process of automating an audit operation with Arkham comprises three steps: 1) Integrating the documents to be audited into our data platform and extracting their text with Textract/OCR, 2) Formulating specific prompts for a Large Language Model to identify and extract key fields, and 3) Implementing a distance-based model to perform the auditing process.
As illustrated in the figure above, the process of automating an audit operation with Arkham comprises three steps: 1) Integrating the documents to be audited into our data platform and extracting their text with Textract/OCR, 2) Formulating specific prompts for a Large Language Model to identify and extract key fields, and 3) Implementing a distance-based model to perform the auditing process.
Let's delve into the implementation details of each step:
Our customer established a VPN connection between Arkham’s platform and the SFTP repositories containing the loan documents for auditing (contracts, invoices, and promissory notes). These loan documents were then integrated into our Data Platform, utilizing our Spark-powered Data Extraction capabilities
Our customer established a VPN connection between Arkham’s platform and the SFTP repositories containing the loan documents for auditing (contracts, invoices, and promissory notes). These loan documents were then integrated into our Data Platform, utilizing our Spark-powered Data Extraction capabilities
The Pipeline Builder, a key application in our platform, is designed to simplify complex data transformation workflows and the application of Generative AI Models. With this tool, customers can automate their recurring audit processes, streamlining their operations with precision.
After setting up the required pipelines, the customer conducted several test runs and configured the final audit output into a modular workbook (picture below). The setup has enabled our customer’s audit team to complete an audit in hours instead of days.
Following the successful automation of 80% of their audit operations, our customer is planning to deploy Arkham’s automation capabilities to several other use cases such as compliance monitoring.
