Many enterprises are investing heavily to implement RPA software. This technology promises to reduce or eliminate manual tasks, accelerate workflows, and enhance customer experience, all of which adds up to a significant return on investment—or so it would seem.
In conversations with dozens of executives, it’s clear that the first act in the “robotics evolution” has not been a slam dunk for many, especially when companies try to scale the localized proofs of concept.1
There are a few RPA use cases in banking that point to early success:
- BNY Mellon’s trade clearance project netted an 88 percent decrease in processing times.
- Danske Bank’s automation of customer account onboarding led to efficiency savings of 27,300 cases per year2.
However, these are niche projects that illustrate RPA success on a small scale. Extrapolating from a handful of success stories can lead to overblown expectations that enterprises simply can’t bank on (pun intended).
Keep reading to rethink RPA and re-evaluate the operational agility and end-to-end automation you were promised.
Unattended RPA vs. Attended RPA
Unattended RPA—the holy grail of process automation—promises full automation and hands-off processes that run 24/7 with no human intervention. Unfortunately, many banking processes don’t naturally lend themselves to unattended RPA, and manual intervention is often required as soon as a document is involved.
In areas like account opening/closing, mortgages, and loans, RPA tends to break down because it must cope with the ingestion and processing of complex documents that are not searchable. In each of these workflows, identification documents (like driver’s licenses and passports) will be scanned, income statements may be required, and proof of employment letters included.
On its own, RPA software cannot extract the necessary information/values to proceed to the next step in the workflow, and human intervention is needed. What started as unattended RPA quickly becomes something else entirely: humans working alongside RPA bots (i.e. attended RPA). Consequently, processes slow, staffing costs rise, the risk of human error increases, and morale suffers when staff is forced to perform menial manual tasks.
Why RPA Isn’t Delivering ROI
RPA software can underperform against banks’ expectations in several ways. In some cases, banks start their RPA initiatives with pilot projects to prove the concept. RPA may work well in these limited projects because workflows are automated in very specific areas, like trade settlements, invoice processing, or payment operations. Unfortunately, RPA often fails to scale to larger, more complex processes, and the expected ROI that would come from wide-scale implementation across many bank functions is lost.
Whether it’s the customer ID needed to process a loan or doctors’ notes to support an insurance app, workflow automation is broken when RPA encounters unstructured data that it can’t process or extract key information from. It’s a “garbage in, garbage out” scenario.
For example, an institution may have already implemented RPA in its loan application process. They may have seen an initial reduction in the resources required and improvement in speed because they were able to automate 30 percent of the workflow with standard RPA.
However, whenever the RPA software gets to a step in the process that involves unstructured data—like a scanned document or letter of employment—the automation breaks. Manual intervention is required to extract the necessary information (like salary or length of employment) and move it to the next step in the chain.
How to Reinvigorate RPA ROI
Efficient access to clean, usable data is critical to RPA success. To amplify RPA ROI, enterprises must adopt a robust data extraction solution that transforms semi-structured or unstructured documents into relevant data that the RPA engine can use.
Look for an automated data extraction software that discovers, cleanses, and enriches unstructured documents. The software should then automatically organize and cluster the documents, intelligently extracting additional attributes. Once these steps have been taken, the newly transformed data assets can now be processed by RPA bots that will use this data to become smarter. By funnelling clean, structured data into the RPA engine, enterprises extend end-to-end automation, derive keener insights, and achieve greater operational agility. As a result, knowledge workers are better utilized, customer experience is enhanced, and the desired RPA ROI is achieved.
Most banks that have implemented RPA projects have seen increases in staff morale and data accuracy, as well as some of the hoped-for declines in processing time, costs, and operational risk. But many have yet to realize their initial ROI expectations because of RPA’s inability to handle unstructured data.
Deploying enterprise-grade data extraction software—an “RPA enabler” at its core—allows organizations to automatically extract key elements from complex documents and feed that data back into the RPA engine. By reducing manual intervention and automating the critical processes that used to fail due to lack of quality data, banks drive better results from their RPA projects—and achieve the ROI they expected when they began.
1McKinsey Digital: Burned by the Bots: Why Robotic Automation is Stumbling
2THE LAB Knowledge Work Factory: Robotics in Banking with 4 RPA Use Case Examples