Every organisation wrestles with the tedium of reconciliation. How can automation help yours?
Reconciliation is the process of comparing and matching the transactions recorded in two or more financial records to ensure they align. This typically involves reconciling documents such as bank statements, invoices, and credit card statements, against the corresponding internal records. The goal of the reconciliation process is to identify and correct discrepancies, ensuring that the accounts are accurate and balanced.
The reconciliation process is often painful due to its time-consuming nature, especially for businesses with large transaction volumes. Manual reconciliation increases the risk of human error, as it requires cross-referencing hundreds or thousands of transactions by hand. Discrepancies between internal records and external documents are common, often arising from timing issues or missing entries. Identifying the root cause of these discrepancies can be frustrating and take considerable time.
The integration of an automated process and artificial intelligence can change the lives of your reconciliation team, helping to optimise and reduce the processing time of what is a repetitive, thankless task. In this blog post, we will present an AI-powered solution designed to automate the reconciliation of over a thousand statements each month, aligning them with internal records. We will explore how this solution boosts efficiency and accuracy, highlighting the key advantages it brings to the reconciliation process.
The design of this AI-powered solution includes three key stages:
1) Extracting internal records from the ERP or finance system;
2) Reading digital PDF statements to collect external records;
3) Reconciling the records between the internal records and the PDF statement.
To build this solution, we used UiPath to automate the process of extracting internal records from the ERP, and UiPath Document Understanding to perform the Intelligent Document Processing (IDP) steps of the process.
The Intelligent Document Processing workflow for the statement reconciliation process consists of five key phases:
- Data Ingestion;
- Document Classification;
- Data Extraction;
- Data Processing;
- Data Reconciliation.
Data Ingestion: The Data Ingestion phase involves capturing data from various content types and preparing it for processing. In this Statement Reconciliation process, documents are received via email from hundreds of vendors and organised into folders. These statements come in different layouts and have multiple pages.
Document Classification: The documents are then classified into different categories, either manually or automatically. Advanced solutions can offer category suggestions based on existing taxonomies. At this stage, the UiPath Document Understanding tool is introduced to classify documents as statements or non-statements keywords. This is important because statement layouts can resemble other documents received via email, such as invoices.
Data Extraction: After confirming that only statements are ready for processing, the automation begins the Data Extraction phase. A trained machine learning model extracts the details from the documents content and checks the confidence score for each extracted field. For this solution, it extracts the Vendor ABN, Vendor Name, and the statement transactions. If the confidence score for any field extracted is lower than the expected threshold – 85% in this case – the automation pauses and initiates the Human-in-the-Loop flow. Human-in-the-Loop refers to involving a human in decision-making and confirming details when needed. In this solution, we used UiPath Action Center to make the PDF documents available for review. The business team then reviews the statements, confirming the accuracy of the extraction or correcting any discrepancies. Any corrections are fed back into the ML model so that it improves over time.
Data Processing: The data processing phase will loop through all the statements’ transactions, make the necessary adjustments to the information extracted, and prepare the data for the reconciliation. Since the robot receives hundreds of statements in multiple layouts, each one follows a different pattern, and they need to align with the internal records’ patterns.
Data Reconciliation: This is the phase in which the reconciliation between the statement’s transactions and the internal records takes place. The robot will match various details to find the transaction in both sources and highlight the ones that were not found, allowing the business team to take action.
As a result of this project, the automation has given over 1,200 hours back to our client after one year of operation. The statement reconciliation process previously required a team of 17 members, spending an average of 2.5 hours per day for at least 10 days. Now, the robot processes each statement and completes the reconciliation in just 3 minutes, while the business team needs no more than 5 minutes to review the file. This significant time savings has allowed the finance team to focus on higher value tasks, while the automation continues to handle an increased reconciliation scope of over 90%, significantly reducing the risk of errors.
The implementation of an AI-Powered solution for the reconciliation process has provided substantial benefits for the client. By automating this process, it has increased efficiency and enabled the business to handle what was a complex and time-consuming workflow, with minimal human intervention.
Connect with our human team using the form below, and see how we can support your organisation!