The foundation of a high-performing Invoice Management Solution starts with automated invoice processing. From collecting the right documents (structured or unstructured) to capturing line-item data, the science of processing invoices can be complicated, complex, and costly without having the right partner and technology on your side, and Cass makes integrations into your current systems fast and easy through our APIs.
We have the ability to receive various structured or unstructured document types in either electronic or paper formats:
Import assorted data files from multiple documents sources via OCR into a single data warehouse:
Through our machine learning and artificial intelligence technology, we automatically understand unstructured data that comes into your organization - from invoices and statements to weight slips and contracts.
We organize your data to be easily understood and readable in a standardized format.
We convert data and other information captured into a digital format for consistent and easy search and retrieval.
The platform allows the training of custom models for each document type. When training the platform to understand a new document type, we annotate several sample documents to show the platform where the data is shown. This is repeated for several variations of the same document type allowing the model to build an understanding of how to read any variation of the same document type.
No, the platform does not use any templates for any functionality.
Yes, our models are very tolerant of changes in design and layout, even if the model hasn't seen that design of the document before, it should still be able to find the info. If the model cannot find the information required, we have fail-safes in place to pass the document to a human. This human-corrected data can be used in a future version of the model.
After the data is extracted, the data is cleaned up and normalized and then sent through validation checks. Here the platform is doing logical and confidence to checks to ensure the data is valid. For example, we'll check that all line items in an invoice are equal to the total billed. If validation checks fail, the document is passed to a human for a second opinion.
The combination of custom-trained models, validation rules and targeted second opinions from humans results in a level of accuracy that is greater than humans alone extracting the data whilst benefiting from drastically reduced processing times that the model provides.
Yes, if the handwriting is sufficiently legible, our character recognition should be able to read it. We also support a lot of languages and different character sets.
Electronic data interchange (EDI) invoices are electronic versions of a paper invoice.
Our system will take the unstructured data from the invoice and convert it to structured, useable data in the form of a JSON file. Clients can also have Cass deliver the data via visual reports based on the data pulled from the invoices.
Yes, the system will automatically reject pages that contain no information. Things like cover pages, images, and blank pages are rejected right at the start.