In this podcast, hear from Eric Kujala, Product Marketing Director, ICE Mortgage Technology, and Nolan Johnson, Sales Engineer at HPA, as they discuss how the mortgage industry is harnessing the power of robotic process automation (RPA) and Artificial Intelligence (AI) today and the key things to consider when evaluating RPA for your business.
Ellie Mae Open House Podcast
Get peace of mind in the unknown with RPA and AI.
More of a reader than a listener? Check out a summary of the podcast below, minus all of Nolan's antics.
What is RPA?
RPA is simply training robots on how to perform your business processes so your human employees can focus on work that provides greater value to the business.
What should lenders do to effectively apply, shop, deploy RPA?
To apply RPA, HPA first helps the lender understand and define their goal behind automating. Is it to speed up time to close? Are they looking to keep staff low in an effort to seamlessly scale? When a lender is solely looking for a "lift and shift", we recommend prioritizing processes that require the most manpower. From there, we target processes that 1) could benefit from 24 hour processing 2) are menial yet carry a higher overhead 3) require rework if done incorrectly.
When lenders are shopping for an RPA solution, it is important to understand the resource requirements of each option and what makes most sense for your business. Options include:
- Licensed RPA software, which can be slow and expensive, but gives you the most control. This route is slower because the learning curve associated with RPA is steep. It’s more expensive because of the cost of licensing, support personnel, and maintenance. RPA is fragile, any changes to the process or underlying system will break the automation, especially when best practices aren’t baked into the architecture of the program. As robots break they require constant maintenance, which eats into any ROI you will realize once the licenses and personnel are paid for.
- Managed Service Providers, or MSPs, to implement the RPA software for you. You still incur the licensing cost, but the risk of the build and the change management is passed on to the service provider. When going this route it is important to understand what fees will be incurred for every part of the project, from consulting, to design, build, management, and monitoring. You can expect to pay 30%-60% on top of the typical cost of licensed RPA software, according to HFS Research.
- RPA-as-a-Service. Full disclosure, I work for an RPA-as-a-Service provider, but will remain objective because whatever you choose ultimately has to work for your business. With RPA-as-a-Service, you’re purchasing access to the latest RPA software plus the automation expertise to ensure quality and consistency in your RPA program. Your team will work with the automation team for knowledge transfer of the process being automated, the rest is handled by the provider. While this option gives you less control because you don’t technically own the software, you also don’t own the issues when the robots break.
How is Artificial Intelligence utilized in RPA?
AI, or Artificial Intelligence, is primarily applied in RPA to enhance robot capabilities beyond rote process execution. In the mortgage industry, the best use for AI is in the consumption of documents, which are often a source of unstructured data. To give you an idea of what unstructured data is think of a scanned-in paper form. The data within it is valuable but a robot can’t just look at the paper and inherently know what is there. It must be trained on what each bit of data in the form represents and how it will be used in automation. Another primary use of AI is exception handling. When a robot encounters a situation that was not included in its coded instructions, it can be trained on how to handle the situation when it encounters it again in the future.
What is a common misconception between AI and RPA?
AI is a broad term that encompasses a variety of capabilities, from machine learning to natural language processing to optical character recognition. It is often regarded as some form of magic that fixes everything. In reality, you’re simply using these capabilities to train robots on how to handle a wider variety of scenarios that don’t fit neatly into the RPA box (stable, repetitive process, using structured data, etc.) Don’t get me wrong, there are plenty of processes in origination that do fit neatly into the RPA box. For the processes that don’t, these capabilities can make robots more intelligent. Of course, intelligent robots are not without their faults. It takes repeated, consistent training, and a detailed and built-in understanding of regulations before they’re ready to fly solo. With that being said, RPA engines such as HPA’s, do a good job of sending signals of data, issues, improvements back to the automation operator. By doing this, you are allowing a stop-gap between the signal and the decision, and allowing sign-off and development to account for whether or not that signal is positive, negative, or just white noise.
What are your industry predictions heading into 2021?
This year was unique with rates being at all-time lows and most forecasts coming out of Fannie Mae and Freddie Mac show that rates will remain flat going into 2021 as long as the Fed keeps rates unchanged. 2021 origination volume is forecasted to be lower than 2020 which could present some challenges to lenders with regards to staffing decisions as they evaluate their short-term versus long-term growth plans. Lastly, the remote employee is here to stay. The continued trend of working from the home office will effect lenders in their needs assessments with regards to people, technology, and overall competitive strategy.