AI-powered contract processing

Realizing that manual scanning and preparing piles of documents isn’t effective in the long run, IMPULS-LEASING wanted to minimize time spent on digitalizing collected documentation while maintaining high accuracy and data security.

The must-haves

The most important requirements were that the solution had more than 90% accuracy and that the scanned files, or any information on them, never left the local network as the data contained personal information.

Before uploading to the DMS, the final validation step must be executed manually. Considering the rigorous scrutiny of the industry, a detailed audit log had to be tracked for the scanning and finalizing of the documentation.

IMPULS-LEASING is an international company present on the market for over thirteen years offering financial services such as financial leasing, operational leasing and operational leasing with residual value for cars, trucks and trailers, construction machinery, agricultural mechanization, machinery and equipment.

AI-powered solution

Completely on-prem scanning process that scans the entire documentation at once, uses OCR and ML model for document type classification and an overview website for finalizing DMS export

Read more

Here is a more detailed overview of this case study. Find out how does the model work and what benefits it enables.

How did we help?

We created a solution that helps employees accelerate the process of scanning and preparing documentation. Instead of scanning each distinct document at a time, all the pages are scanned at once and later separated using an AI model, which speeds up the scanning process tenfold.

The entire documentation is scanned at once from any workstation. All the workstations are connected to a central processing station using shared folders.

The documentation is separated into multiple single-page documents from which the text, image and optimized PDF are extracted with ABBYY OCR software to be used by AI model, verification web site and DMS export respectively.

A locally hosted text-based AI model is used to automatically separate the large PDF into the original documents with 95% accuracy. Some internal forms and documents are outfitted with barcodes for easier classification.

To visualize and adjust the final output each worker can use a specialized verification website where they have complete control over the documentation. Access to the verification website is granted only to workers that belong to the appropriate security group.

Only after they verify that the documentation is ready, is it uploaded into the main DMS. For this step, the optimized searchable PDFs are used.

*Considering data sensitivity and specific security measures referring to personal information management, the solution includes only such tools and software that guarantee no outgoing traffic.

Before any development started, we familiarized ourselves with not only the process but the whole documentation pipeline. With this wide overview of the department’s workings, we could come up with the best suitable solution and gain ideas for possible later upgrades.

Due to an agile approach, we were quick to adapt to any new requirements and build the solution following their growing wishes for which our client showed great appreciation.

bonsai.tech has not only shown great dedication during development but also took a considerable amount of time to fully understand what we could benefit from.

- Board member, IMPULS-LEASING