Ensuring the loan origination process goes smoothly is a continuous challenge. Mortgage lenders process hundreds of applications daily, but with so many documents involved, the process requires significant manual effort, and is prone to error and delay. Now imagine if you could intelligently automate the process, ensuring loan applications are complete and documents are automatically uploaded into your loan origination system (LOS). Imagine automating specific functions, like tasking, conditioning, and decisioning, to digitally transform your entire process.
With robotic process automation (RPA), you can automate interactions between critical applications, data entry and extraction, and reporting. Whether or not you use the latest, most sophisticated POS, LOS, and core banking systems, RPA can be layered on top of your current systems to eliminate and streamline business processes. And because it builds upon your existing technology, RPA can provide immediate relief for limited staff and remove the need for temporary workers as your workload expands. Intelligent document processing further accelerates and supercharges the mortgage origination process. By combining RPA with machine learning and artificial intelligence, costly steps like document classification, manual data entry, and document analysis are eliminated.
Watch this on-demand webinar to learn how you can take your mortgage processing to the next level and discover:
- How to integrate older, closed, non-API based systems with your modern applications
- How to identify tasks loan officers are doing today that could be performed by a bot
- How bots can aid in underwriting with reviewing documentation
- How Fortra solutions provide a higher-quality mortgage lending experience through a live demonstration
Welcome, everyone, to today's webinar on Taking Mortgage Processing to the Next Level with Intelligent Automation. My name is Erik Fisher, and I'll be your presenter for this session. At HelpSystems, we understand that building a better IT is not a destination, it is an ongoing commitment. IT complexity is the top risk for managing an IT group. On average, corporations have over 900 applications being supported by IT. We are building a product portfolio that can help IT organizations get the critical functionality they need, while reducing the complexity they face.
We focused on two broad categories; security and automation, to bring value to our customers. On the security side, we provide a broad portfolio of solutions that include data security such as our GoAnywhere Secure File Transfer software and our Clearswift Data Loss Prevention product line that is in the Gartner Magic Quadrant. Our Core Security IGA, or Identity Governance and Administration product, offers an integrated identity and access management suite, including tools to manage [drought 00:01:09] profiles.
Finally, we have our infrastructure protection products that secure networks, such as pen testing and vulnerability assessment services. On the automation side, we also have a broad suite including infrastructure automation, such as VCM or Vital Capacity Management, workload automation, like our JAMS product line for back-office or batch automation, and of course, HelpSystem's Automate RPA on which we're going to spend our time and focus today.
Let's take a look at what we'll be covering in our time today. First off, we'll take a look at some of the challenges in mortgage processing. Then we'll examine some of the ways financial institutions are using RPA in banking and mortgage processing. Next, we'll show how to leverage intelligent automation for the mortgage origination process, and finally, we'll show a little bit of Automate and Automate Intelligent Capture in action.
What are some of the challenges in mortgage processing? First, we'll take a quick glimpse at the current mortgage market. The average time to close a loan is roughly 47 days. The 2020 origination volume was around $3.9 trillion, $777,000 of which coming from new loans, which was a 14% increase year over year, and $2.4 trillion coming from refinancing. This giant refi wave has put tremendous stress on loan originators and mortgage banks, highlighting even further some of the pain points of the origination process that are ripe for automation.
Ensuring the mortgage origination process goes smoothly is a significant challenge. Mortgage lending companies require more documents than other organizations in virtually any other industry. The time and effort required to process these forms is a significant strain on resources. Mortgage lenders process hundreds of applications daily, but with so many documents involved, the process requires significant manual effort and is prone to error and delay.
Ensuring data privacy, accuracy, and regulatory compliance is essential in this multi-step, heavily regulated process. Many financial institutions rely on older banking systems, so the importance of digital transformation and reinventing a process that was intended to be an in-person process to include new technologies like E-signing is an ongoing consideration. Add to that the COVID pandemic, the forbearance that comes out of that, potential changes in the 2021 market, like increases of purchases and the inability to hire quality loan officers and underwriters leads to more opportunities to incorporate RPA to streamline your processes.
All right, we've made it to our first polling question. How many loan applications are you receiving each month? We'll give a minute for those responses to come in. It looks like most people are unsure and then we have a range from anywhere from less than 100 to a fair amount with 1,000 plus. So, quite a range of applications.
Now that we've received and reviewed some of the challenges to loan origination and processing today, we'll take a look at how financial institutions use RPA in banking and mortgage processing. Some of the common areas we see at our bank and credit union clients using RPA to enhance their processes are loan underwriting, which we'll take a little look at today, debit card fraud processing, regulatory compliance adherence, creating an audit history of every step in the process, Regulation D violation, letter processing, account closures, so closing inactive credit and debit cards, especially during the [CSHMD 00:05:46] process in an error free fashion, loan bundling, eliminating manually preparing each loan individually by allowing RPA to group and prepare loans, priority management for incoming customer inquiries, the know your customer process management by enforcing anti-money laundering policies, identity, and risk verification, ACH stop payment processing, daily and overnight processing, including pulling data or running or creating reports overnight so they are ready for the start of business, automated payroll information management, copying members' existing payroll departments into a new one, cleaning and mining payroll data to ensure accuracy.
A few areas that aren't on the slide that we've seen RPA used, as well, is checking old overdraft authorization docs for signatures, logging appraisal info into loan origination systems or core systems, CIF review, and website and client portal testing. So, creating a dummy account and continuously logging on to the online banking system and navigating around and doing transactions to make sure that all of it's working so that they're able to find out if there are any issues before the clients notice them and hopefully resolve them.
What is the role of RPA in intelligent mortgage processing? Robotic process automation can increase productivity, reduce costs, and eliminate errors by handing manual repetitive tasks over to software robots. Some of the ways we see this done is getting data from older systems into modern applications and vice versa. A great example of this that we've done a bunch of is taking the application data and documents from a newer point of sale system, and then inputting and uploading those documents and applications into an older, more closed loan origination system, like a Mortgage Cadence, LLC or a mortgage bot system where the only interface is through a web portal.
We've also seen it work very well with taking low-level, manual tasks from loan officers and giving them to software robots. Requesting services like titles or MI quotes, emailing applicants about missing documents, being able to loop through a folder and make sure that all the required documents are there and notifying people if there's missing documents, or missing data, and doing some of those approval letters and rejection letters, and then letting go of tedious document review process, especially during underwriting. Using our Automate Intelligent Capture solution, we're able to review these documents such as bank statements, credit reports, consent forms, et cetera, to make sure that they meet the qualifications necessary to approve the loan.
Here we can see a quick example of how Automate can take data and enter it into a system via the user interface. In this instance, the client acquired another bank and had to close some branches and therefore needed to migrate accounts into their core system, which in this case is Fiserv DNA. I like it because it shows how automation can run through a task script, which you see in the middle, and take data from an Excel file, which is on the far left, and then log into the core system on the right in the same manner that a human would, and then enter all that data in. It does a nice job of showing how Automate can process these mundane tasks without any breaks or meetings or driving employees crazy with mind numbing work.
How do we leverage intelligent automation for the mortgage origination process? Just imagine if you could intelligently automate the process, ensuring loan applications are complete and documents are automatically uploaded into your loan origination system. Just think if you could automate tasking, conditioning, and decisioning to digitally transform your entire process. We'll now examine some of the ways we've been able to automate and enhance the mortgage origination process using our Automate RPA and the Automate Intelligent Capture software solutions.
Here are some of the ways we've used RPA and Intelligent Capture to supercharge the mortgage origination process reviewing file logs and feeding data back into retail POS systems. A great example of this is we have a client who uses Blend and then encompasses their LOS. There's already integrations between them via APIs, but we're using Automate to review the log file that comes out of there and check for errors and then notify appropriate loan officers of those errors, as well as being able to feed relevant data back to that retail POS as necessary.
Automating credit prequalification process. We've done a little bit of work in this in a few ways. First off is doing a FICO matrix triage. Based on the customer's FICO scores, loans can be sent in different directions. The best scores can get sent directly to a loan officer and fast tracked. The worst scores can, basically, get a rejection letter sent out automatically and then that middle area can be sent to a credit department to determine if the scores can be improved or not. We've also used this in pre-flight loan processing. Taking advantage of bots to do all the things your loan officer needs to do to really review the application: initial disclosures, MI rate quotes, et cetera.
For loan applications that pass initial scrubbing, Automate can automate the routine tasks such as clicking the right buttons to do a rate disclosure or other procedures. Automate can also pull key data from the LOS and move it to your choice of provider, not just the ones that are integrated with your LOS system. So credit checks, background checks, flood certifications, that type of stuff, if you want to use a provider that's outside of, or not natively integrated within your LOS system, Automate can take that data and format it and pass it along to that company and then take their response and plug it back into your LOS.
Creating files and sending files securely. Automate can also take the data and create files or reports, move those files or reports to different network shares or share points, or upload to SFTP or FTP sites, and then notify the stakeholders that the files are available for download.
Send conditional approval and rejection letters, pulling key data from the LOS. We mentioned the low FICO score as triggering a rejection letter. Also, after a loan officer approves a loan, Automate can generate the conditional approval letters to the customer, pulling in the conditions and requirements from the loan origination system. We can also update a retail POS with new data status, too, to keep those systems in line, and also, review loan applications for completing this using our Automate Intelligent Capture solution, we can ensure that bank statements are for consecutive months and that the ending balance allows for payments of closing costs. We can make sure that consent forms are signed. We can review credit reports to make sure there's no fraud, that type of stuff and also review signed loan packages and break those signed packages into individual docs, review to ensure all documents are present and have the critical information necessary, then rename them and save them to an ECM or core system or wherever they need to go.
Working together: RPA and Intelligent Document Processing extend mortgage automation capabilities, and enable organizations to maintain control of critical information with speed and accuracy. With robotic process automation, you can automate interactions between critical applications, data entry and extraction, and reporting. Whether or not you use the latest, most sophisticated POS Los and core banking systems, RPA can be layered on top of your current systems to eliminate and streamline business processes and because it builds upon your existing technology, RPA can provide immediate relief for limited staff and remove the need for temporary workers as your workload expands and contracts. Intelligent Document Processing further accelerates and supercharges the mortgage origination process. By combining RPA with machine learning and artificial intelligence, costly steps like document classification, manual data entry, and document analysis are eliminated.
Delivering on the promise of digital transformation with an intelligent automation platform requires the ability to harness both structured and unstructured data. But not all data is created the same. Structured data is organized, clearly defined, and searchable, but represents only about 20% of the data in each company. Robotic process automation excels at structured, or quantitative, data automation.
Unfortunately, more than 80% of all data within organizations is unstructured. Unstructured data is typically qualitative. It is difficult to interpret and analyze because it has no predefined structure.
The key to unleashing the true power of RPA is through unassisted machine learning that can harness both structured or unstructured data. Automate Intelligent Capture automatically learns data fields on forms by using custom machine learning algorithms without endless hours of building templates. This allows for fast setup of previously unseen forms because the system is learning which document information to capture, getting smarter and better over time.
Out of the box, organizations see 70% automation and within weeks of learning automation is that more than 95%. Once that data has been accurately captured and turned into structured data, Automate can do its thing and feed it into any necessary system or compare it against existing data within a system or load the process documents into a doc management system. Really, whatever next steps the process entails, Automate can handle from there.
Now we're to our second polling question. What loan origination system do you use? We'll give a little bit more time to get those responses in. All right, looks like a whole lot of Encompass users and then a fair mix of the other ones.
So now it's time for our live demonstration of Automate RPA and Automate Intelligent Capture. To do this, I've created a couple workflows that allow us to loop through data that would come from an LOS to mimic an underwriting decisioning process. We're able to look at FICO scores, LTVs, product type, et cetera, and all the various fields that go into the underwriting decisioning process, and what we're doing in this is kind of cherry-picking the creme de la creme of the application. So the ones where every box is checked and we can automatically approve those loan applications.
Let's kick off this process and then I'll go through what we're running a little bit here, too. We'll start this and then we can see that our Automate process will get started here, momentarily, and what we're doing is walking through and basically, like I said, pulling data from an LOS, looping through that data, making sure that all the values are meeting criteria. FICO's are above a certain range. LTVs are below a certain range. Its primary residence, et cetera, and then from there, we're going and we're actually, if all those boxes are checked, we're going to go and loop through a document folder for each of the loans and make sure that all of the relevant documents that are needed are there.
Once that's done, we're going to pull the credit report and push that into our Automate Intelligent Capture system so that we can look to make sure that there's no consumer statements, there's no bureau remarks, and there is no fraud alerts, and if all the boxes are checked there, then we have another workflow that takes the output and it's triggering off of a file being dropped out of AIC into a folder, and it's going to do that kind of final credit check of those values that it captured from the document, and then if everything is all good, it's going to send a conditional approval letter. If there's any issues that's going to mark that there was issues and at which step. If we go to our process, we'll see that the steps can continue to roll through here and eventually what you're looking at on the screen now is our Automate Intelligent Capture main screen, as well.
If we look at the reports, we've actually dumped in our initial model ... actually, we need to kick this off again because I left a file open, so ... let's close that and restart. Now we'll see our process go through and we can walk through some of the steps within the tasks.
As we see, our processes running. It's doing the decisioning portion of it. You'll see an Excel file pop-up, which is where the error was because I had it open already, previously, and it's going through and on the left-hand [person 00:20:47] that we couldn't see quickly, checking all of those, those different categories to make sure that FICOs are okay, et cetera, and then it's going to move on to the next step where it's doing the document check. It's actually looping through a folder system and checking that the documents are there, which it's doing right here, and then letting us know if all the documents are there, and finally, it's moving onto the next step where it's going to be pushing the credit reports into our AIC system, and then that process can run.
While that process is running, we'll take a look at some of the actual tasks themselves and how they're built. So the task builder within Automate is a drag and drop step-based system to where we have 70-plus actions and probably over 700 sub actions. We can interact with all different types of systems. Creating new tasks and scripts is as simple as basically dragging an action into the task builder and then filling out the forms. Within this, you can see we're using various conditional logic. If/then statements. We're looping through data sets. We're writing to Excel files. Really, any of that additional data. We can access webpages. Pull data from that, enter data, click buttons, do everything that's necessary through all of these native actions. That being said, if you already have, Visual Basic scripts, PowerShell scripts, Python scripts, batch files, et cetera, that you want to incorporate in some of these workflows and processes, Automate can handle those and kick those off as well, too.
If we hop back to AIC here for a minute, we'll see if we refresh the reports that we've got some data getting pushed through. I had sent one earlier, so we can review it later to this class verification process. Right now, we see that our first document is already in the capture process and the various steps within the Automate Intelligent Capture process are image processing, breaking that file out into tables, easily read data, PNG files, and then going through an OCR process which allows us to, in the areas that it's not sure of, use two different OCR engines to capture those results and then from there, it goes into a classification section. So based off of where it puts it into various document types that are specified. Once it knows what document it is, it moves into the capture section, where it's actually going in and pulling that data, or the data elements that are required, for each document type, and finally it goes to a data verification.If the data doesn't meet certain criteria so that a user can actually go in and point and click around and say, "All right, this field actually needs to come from here," and so on, and then it gets exported in a variety of manners, either as a CSV, XML, a SQL database, whatever's necessary there to export the data into other systems or into a file format that Automate can then take and continue processing.
That last little part that popped up was actually the Automate checking to make sure that all of the fields from this capture were already there. As we can see there, our files have move through this. The final piece of this was an email that was sent for the initial approval. We can see that one of these loans met approval, so I put together just a real basic email.
You can create HTML-formatted emails. You can use different variables so you can pull the borrower name, email address, and everything from within your systems and have it automatically populate emails and send those out, as well as adding attachments or any necessary supporting documents, as well.Just to dig a little deeper into the Automate Intelligent Capture, we can actually take a look. I've preloaded this one earlier so that we can take a look at the data classification and data verification processes within this. I actually want to go to the administration panel here and just turn on, I had it set up so that as we're running the automation, that the files just automatically flow through if they meet the criteria. But now we actually want to take a look at those different steps.
We'll turn that on and we'll go to our class verification and load our batch there so you can see how the document type classification works within the system. We'll load our batch and you can see it will pull up a document and in this case, we're just having this be one credit document. If we really needed to, we could break this out into the main credit document and then the supplemental ones, as well, too.
Once we're confident that this is the type of document it is, we can submit this and it'll move on into the process. We can see now, it's at this capture point and then it's going to stop at this data verification point so we can ensure that the data we're pulling and this is also where we'd set it up to train the data, the model, to know which fields to pull, where to pull the fields from and the beautiful thing with Automate Intelligent Capture is that it does use machine learning so that you're able to ... the more documents you pass through, and the more verification you do, the more accurate it gets to where it gets to a point where it's very low touch.
We'll let this get to the data verification point here. Now we can load our data verification batch, and again, this is an example of one of the documents that we pass through in our initial run. All right. So we can see where it's pulling data from. In this case, if we want to point where to pull this consumer statement or these bureau remarks, we can actually have this highlighted green, go to the correct document page that we need to pull this from, and just point to show, all right, we need to pull this from this particular field right here.
You can just draw the area you need to pull and it will populate that and then as you, the more you do this as you're training the model, it will know that, "All right, we're pulling this no consumer statements portion from underneath consumer statements and we're pulling our bureau remarks from underneath the bureau remarks," and it dynamically learns this and continues to improve accuracy going forward. The other piece that we pulled here are actually more of a line item table detail. It's pulling elements from our fraud alert table and populating those.
Then we're taking the export from this. I'm not going to submit this just because we've already processed it, and then making sure and going through and making sure that those various fields that we're pulling from there are what we need them to be. It needs to say no consumer statement found, no bureau marks and not have different fraud alerts in there. We're able to automate that underwriting decisioning process for perfect loans and send a conditional approval letter.
Now, we're at the time for our final survey question. Are you open to exploring how Automate and Automate Intelligent Capture can further streamline your origination process? Please answer that when you get a chance and then we'll move on to any questions and answers that you may have.
There was questions about this deck and the recording. We are recording the webinar. You'll be able to download a version of that. That'll get sent out to you and I believe we'll be able to share this deck as well. Any other questions you have? There's a questions segment where you can enter any questions. You can also use the chat if needed, and obviously, if you think of anything after the fact, feel free to contact us and we'd love to get back to you and talk to you about any of those additional questions or to show you more of what Automate or Automate Intelligent Capture can do.
I'm not seeing any new questions, so at this time, I'd like to thank everyone for joining us today. We really appreciate your time and learning about HelpSystems and Automate RPA and Automate Intelligent Capture, and how we're able to help streamline some of the mortgage processing and origination processing. For more information, feel free to look us up at www.helpsystems.com or email at [email protected] Thanks again for joining us and yeah, we'll share some contact information, as well, too. That was the last question I got. So, everyone have a wonderful day. We're glad you were able to join us and we will hopefully talk to you in the near future. Thank you.