Document Analysis: How We Reduces Manual Effort With AI- Systems

Document Analysis: Creditreform, with its office in Leer, relies on AI-based software for processing and archiving documents, from receivables management. With an automated document analysis, the effort involved in processing correspondence was reduced by 90 percent.

The unique solution from the Cologne-based software provider Evy Solutions is its AI-supported, text-based approach, with which relevant information can also be read and classified from unstructured data. Therefore, the solution for document analysis does not care where in the text the critical information is located. Depending on the application scenario, customers benefit from a degree of automation of almost 100 percent and up to 90 percent cost and time savings in document processing.

Document Analysis: Searching Through Correspondence By Hand

Receivables management receives a lot of mail every day with various documents that must be classified and assigned to the respective procedure. In the past, the employees had to manually search through the entire correspondence for the individual file number, enter it and transfer the document to the archive accordingly. It was also necessary to specify the respective document analysis and classification for the claim process. On average, it took 40 minutes every day until all documents were classified and correctly archived.

Almost 90 Percent Time Savings When Capturing Correspondence

Today, thanks to the software, the whole process only takes about five minutes and looks like this: All of the daily correspondence from all parties involved in receivables management (lawyers, creditors, debtors, bailiffs, etc.) by post, fax, and email When it arrives at the company, it is entirely loaded into the main folder – directly by email or scanned. And from there, it is transferred to the software via an interface. The cloud software recognizes the file number of the respective claim procedure in each document and classifies the document. The solution then adopts both information in the file name. The paper is automatically stored in the appropriate folder based on the file name.

Quickly Ready-To-Use Document Analysis Thanks To Self-Learning AI

In the present application scenario, the solution for document analysis is constantly being further developed, which is why the team exchanges ideas with the manufacturer every one to two months and considers how specific issues can be mapped. So far, the manufacturer has been able to find a solution for every individual application request and update the software accordingly. In addition to the strong support, the self-learning solution generally scores with a high learning speed and can also be flexibly set up from individual modules depending on the application: From the classification, sorting, and storage of the documents to the content separation of batch documents, reading out information or checking the extracted content and

Document Analysis Enables Automatic Document Separation

The following milestones for the application scenario are currently being worked on: With the help of the AI, the software is to learn not only to read the file number and classification from the documents automatically but also, depending on the type of document, a lot of other information that is important for the further processing of the papers. The enforcement notices contain, for example, various data on creditors and debtors, such as the respective addresses, the type of company of the creditor or the different file numbers of the local courts, as well as all amounts receivable such as reminder fees and other dates (from when is which claim).

It is planned that the software will automatically read out all relevant data, And the automatic document separation, thanks to document analysis, is also an essential factor that is to be mapped via the solution. Previously, when scanning documents, employees always had to manually insert a separator sheet to mark where a new record begins – this is to be avoided in the future by the solution automatically separating the documents correctly. But where does a copy start, and where does it end? In addition to the file number, the software needs to be trained for other references, such as information on the letterhead. Because the respective file number is not always on the first page of a new document, or there are only handwritten file numbers, which are sometimes difficult for the software to recognize. 

Automatically Trigger Internal Follow-Up Processes

The most significant potential of the solution lies in the triggering of internal follow-up processes, which the software automatically triggers through the AI-supported reading of certain information. To achieve this level of automation, a separate user project group was set up, which analyzes the various types of incoming documents and considers which follow-up processes can be triggered and which information needs to be readout. The team has already been able to identify some of the papers that come into question. For example, a letter from the residents’ registration office with new address details for a debtor could automatically trigger a resending of the notice via the software.

It would also be possible to react to attachment decisions or notifications of the decree with an automatic adjustment of the deadline or an immediate resubmission. And with incoming messages about the deliverability of an email, the software could read out and classify the various reasons for this. The reduction in workload that automation achieves is already greatly appreciated: the manual effort that employees have to deal with has fallen significantly. And there is still a lot of potential in the automatically triggered follow-up processes, which will be successively explored.

ALSO READ: Do Telegrams Still Have A Future?