Do You Know the Way to RPA?
PHOTO: Andrey Larin
With all the hype surrounding robotic process automation (RPA), it can be helpful to remember that the vast majority of the data going through RPA automations comes from or ends with a document. In fact, I would estimate that roughly 80% of RPA automations fall into this category.
Because so many RPA automations involve documents, we can identify common document processing patterns and use cases. Here are the five most common ways businesses use RPA to improve document processing:
1. Transactional data entry or direct processing
Transactional data entry is the mother of all unattended bot automation.
Unattended bots do not run on a user’s personal workstation. Rather, they run on one or more physical or virtual machines that are designed to perform more batch and longer-running transactions. Unattended bots are tailor-made to process documents entering the organization on their own schedule, where timely and highly accurate data entry into downstream business systems is critical.
We’re talking about the most important operational documents of an organization like invoices, insurance claims, electronic health records, shipping documents, etc. This is the lifeblood of data that flows through the veins of any organization. Regardless of the document, the processing pattern is often the same:
- Documents are captured and recorded.
- OCR and machine learning automatically categorize documents and extract transactional data.
- Extracted data is validated using internal data sources.
- Validated data is automatically entered into downstream business units through the application’s user interface (just like the user does).
- Real-time prompts are generated and displayed to skilled knowledge workers to handle exceptions.
- Transactions are recorded and logged for auditing and compliance purposes.
RPA realizes the elusive, brigadoon-like promise of direct processing. While the entry of transaction data is usually done by unattended bots, there are also many visited bot applications. The bang for your buck depends on the volume of documents, the length of the transaction, and the verification requirements.
2. Document capture
Document capture is not always an easy task. While most document processing platforms can retrieve documents from scanners, cameras, surveillance folders, and email accounts, this is not an exhaustive list of the channels through which documents enter the company. Many organizations are also forced to acquire documents from locations such as websites, FTP sites, EDI translators, and so on.
Consider this use case for a defense company providing public procurement information to its subscribers. Prior to implementing RPA bots, employees searched multiple government sourcing sites throughout the day looking for new and changed requests. This is time consuming and tedious work.
After implementing RPA, staff were able to perform more customer-facing functions while the bots were able to comb the websites, determine which requests are new (versus updates), and update their internal request tracking system accordingly. The result is faster, more frequent, and more accurate processing at a fraction of the manual cost.
Another company receives most of its orders from a large customer who publishes orders on its various departmental websites and also sends orders through EDI. Before RPA was implemented, this organization had dedicated staff to monitor both the websites and the EDI account for incoming orders. After implementing RPA, these document sources are now monitored and processed by bots, which in turn send the orders for fulfillment and billing. The company reports that the cost of order fulfillment has decreased by 25% and the customer’s order volume has increased by 12% due to increased customer satisfaction.
RPA can be an incredibly powerful multi-channel document capture tool.
3. A link (ie “Jumpto”) to required documents
The main reason why companies do not centralize their document storage or duplicate documents across different business systems is the need to search for documents from different contexts. That means accounting may need to pull an invoice from a supplier record in an ERP system, while an HR professional may need to view some form of life event change in an EHRS.
Typically, to meet these requirements, a custom connection to the repository must be made from every application screen in every business unit. Not only is this a lot of work, but in most cases it is impossible.
However, through a bot visited by RPA, a “jumpo” button can simply be placed on any screen with no changes to the application. When this button is pressed, an automation starts that extracts screen data to provide context and then uses that data to “skip” the associated document.
The visited bot acts as a universal adapter that can determine the application and data context for a particular user and present the correct documents to the user. The same goes for contextual requests flowing in the opposite direction or something called a “reverse jumpo”. For example, when a user is viewing an image (e.g. an invoice), they can click the Jumpto button to extract the metadata needed to set up the context and navigate through the ERP system to the associated invoice.
The possibilities of creating a powerful, contextual hyperlinking experience between data, documents and system records are endless.
Related article: BPA vs. RPA: How Are They Similar, How Are They Different?
Despite the power of machine learning and its ability to perform automatic classification and data extraction of documents, many documents are still manually indexed or augmented with metadata from various business systems.
“Quickcopy” is the process by which a user opens a record in a business unit and, using the same button described above, starts an automation that extracts selected data from one or more application screens and then quickly copies it to the metadata fields of the open document. As with jumpto, the user can reverse the quick copy by transferring data from the metadata fields included with an open document to fields within the selected record in the business unit.
As a side note, the more you adopt the Jumpto pattern, the more Jumpto allows you to interact with documents in a central repository from any application screen on the desktop.
5. Document assembly and form extraction
Compiling documents and extracting forms has always been around. Leveraging the power of RPA, any application screen can easily be used as a source for compiling a document or as a destination for data extracted from a form.
Document composition is the process of extracting data from application screens and combining it into a document or email. Think of this as a universal mail merge feature that can be added to any application that gathers data or inserts it into any document. A good example is creating a contract in DocuSign that collects input from one or more selected records in an application.
Form extraction is really just the other side of the same coin. In this case, data is extracted from a form and pasted into the fields of an open application record. Collecting documents and extracting forms are used by both visited and unattended bots alike.
Do you know the way to RPA?
RPA and document processing work together to increase productivity like peanut butter and jelly. Borrowing from these five patterns can make the productivity-reducing tasks more efficient that could slow your business down.
Joe Labbe is Vice President of Business Development at KnowledgeLake. His primary responsibility is to work with KnowledgeLake’s systems integration and OEM partners to integrate the KnowledgeLake intelligent automation platform into business solutions for their customers.