Edit

See fully automated process

Remember our problem

The issue relates to the reality that every bank has its specific documentation. Each receives an overwhelming amount of applications every day. The overall workflow is about 3,500 documents per day. As a result, this bank has 278 employees that process claims daily.

However, it is not the number of papers that makes this work so difficult. The amount is only the first problem. All other issues are even more dramatic and demanding: There are too many types of documents; from various partners, in different languages, and on different schedules; processing is usually time-consuming and prone to errors.

There were five particular problems for this bank:

Variety	There are 11 document types in total: 8 document types are processed in daily review; and 3 more document types are processed monthly.
Priorities	Of these documents, types #1, 5 and 6 make up 90% of the workflow.
Connections	Document types 1 and 8 are interconnected — it's necessary to validate data extracted from document 8 against document 1.
Different clients	There are 15 top vendors supplying invoices monthly.
Languages	Invoices come from 10 different countries in 5 languages (English, German, Japanese, Chinese, Spanish).

All these problems mean that we need to create a complex business process. Let's consider this Use Case in terms of possibilities the WorkFusion platform offers.

See the fully automated solution

The collaboration between the Data Analyst and Subject Matter Expert helped us streamline the process of operations with documents. Here is how the fully automated solution can look.

Stage	Description	Problem solved
Input Storage (multiple types of collateral)	In the first step, there is a scheduled procedure that picks up all the new files in a shared folder or any other storage, sorted by data of arrival or document status or anything else.
ToD	Document is passed over the business process for conversion from PDF into HTML/XML. Electronic documents go directly to the preprocessing step to define its language. All other document types (emails, attachments) go to the end of this workflow since they will be handled in another business process.	Variety: After the tagging and extraction of all information from documents, what's left is unformatted text.
Classification (language)	After OCR, documents are routed to the classification step: According to the current workflow, documents come in English, Spanish, and Japanese languages. We'll create three workflows, one for each language, since we'll need separate models.	Language: The machine can process various languages simultaneously without compromises in quality.
Preprocessing (type definition)	Next, define the document's type by its name: In Spanish, we can receive 5 document types; in English, 2 types. All of them go to separate models to be processed. In Japanese, we receive invoices only, so that's one model.	Variety: This helps to sort documents by type.
ML 1 / 2 / 3 language top clients	For model training, we define (if possible) top vendors per document type (or if there is a variety in layouts or something else). The data set is collected per each model, and models are trained separately in the development phase. In production, the trained models will extract values from the incoming documents.	Different clients: Helps to determine the relationships between top vendors and document specifics.
Postprocessing	Raw extracted data may need conversion before being recorded into the database. For example, a value, which is a text string in the document, is stored as an alphanumeric code. Records that meet the requirements, for example, with all fields extracted, skip manual review and are pushed directly to the data store. That's straight-through processing (STP).	Variety: Helps to unify values.
Manual review	Records that do not pass STP are reviewed before being pushed into the database. That won't necessarily imply a review of the whole document, but often it's some small part of it that the model didn't process.	This part does not solve any particular problem but facilitates the resolution of every problem type above.
Push to Data Store	The decision is pushed through the system to the final client.

The section on Tagging over Document ends here. You can proceed further with the Automation Guide materials.