AWS for Industries

How Amazon Machine Learning can Help Lenders Automate the New Round of Payment Protection Program Loans

Last week, the Small Business Administration (SBA) released new guidance for Paycheck Protection Program (PPP) and placed emphasis on helping smaller lenders and borrowers. According to the SBA, “small businesses, as defined by the US Census, employ 59.9 million people across the country.” A big push for the new guidance is to help “ensure increased access to PPP for minority, underserved, veteran, and women-owned business concerns,” the agency said.

When COVID-19 started near the beginning of 2020, the CARES Act was approved and provided relief via the PPP to millions of US small businesses. The first few rounds of the PPP program provided $525 billion in forgivable loans to small businesses between April to August, and was estimated to save over 51 million jobs. This program has recently been reinstated and will continue to provide relief to more small businesses by keeping their employees paid.

At the start of the PPP program, many financial institutions struggled to administer these small business loans quickly to help small businesses across the United States. On top of the sheer number of requests that were insurmountable, the first iteration of PPP was riddled with funding issues, persistent problems with the agency’s E-Tran loan management system, and repeated mid-course rule changes. Many banks were overwhelmed with submissions as well as paperwork. Depending on the type of business, each loan application request is roughly 3-5 pages each and the supporting documents quickly add up to over 30 pages. Multiply that by the number of people that applied, that’s a ton of paper (30 million plus pages). The need for financial institutions to process these loans quickly became very clear and across the board, lenders dealt with numerous technical glitches and overwhelming demand.

One company that was able to overcome these challenges was Kabbage, a data and technology company that provides loans to small businesses, as it quickly became the second largest PPP lender in the nation by application volume, serving over 297,000 small businesses. By integrating Amazon Textract into its PPP loan origination pipelines, Kabbage processed documents such as tax filings and proof of business documentation that resulted in nearly 300,000 approved loans, totaling $7B.

Amazon Textract helped support 80% of Kabbage’s PPP applicants to receive a fully automated lending experience and reduced approval times from multiple days to a median speed of four hours.” – Anthony Sabelli, Kabbage

The lessons learned by implementing machine learning (ML) to help speed up the loan application process in the first round last spring and summer helped financial institutions improve in areas of technology, processes, and infrastructure. However, the latest PPP round, known as PPP2, is not without new challenges, particularly when it comes to determining eligibility and underwriting standards for a second PPP loan.

For example, PPP2 allows existing borrowers with fewer than 300 employees to apply for a second loan of up to $2 million as long as they show a revenue decline of at least 25%. But the likelihood of a small business having audited quarterly financial statements is slim, and lenders would have to find the middle ground and quickly find creative ways to determine loan eligibility. In the same way that ML tools helped improve the loan origination process for PPP1, ML tools can do the same for PPP2.

Over the years, AWS has been building ML and Artificial Intelligence (AI) services to help overcome document processing similar to processing the massive amount of PPP loans at warp speed. Intelligent Document Processing (IDP) can be “broadly defined as any application that can capture data from documents — email, text, scanned images — and extract relevant data for further processing using AI technologies like computer vision, OCR, or NLP”.  Although IDP is relatively new, some of our AWS customers like Kabbage, BlueVine, Baker Tilly, and Biz2Credit turned to AWS during the pandemic to help speed up the PPP loan processing and keep small businesses running. Using pre-trained AI services and ML tools, our customers used a number of services like Amazon Textract, Amazon SageMaker, Amazon Comprehend, and/or Amazon Augmented AI to quickly process these loans.

AWS offers the broadest and deepest set of AI and ML services, and similar to other AWS solutions, you only pay for what you use. This flexibility allowed companies like Kabbage and BlueVine to quickly evaluate and implement a solution within days, which in turn improved its customers’ loan experience. BlueVine automated the loan processing of parsing and analyzing PPP forms to eliminate significant bottlenecks in the process and to ensure the documents were verified in accordance with the law, within a few days. This saved its customers days of waiting for loan approval and provided a streamlined approval in hours.

Customers in the financial sector have benefitted from using the following AI and ML services with no ML experience required:

  • Amazon Textract: Extract printed text, handwriting, forms, and tables from virtually any document using OCR to eliminate manual processes. This allows you to process PPP forms, in English and Spanish, in an automated way which enables you to process loan applications quicker.
  • Amazon SageMaker: Helps you build, train, and deploy ML models for virtually any use case including building custom models for text extraction and analysis so you can quickly process print, handwritten, and electronic documents, such as PPP loans.
  • Amazon Comprehend: Extracts relevant details, insights, relationships from text and categorizes documents using natural language processing (NLP). You can isolate the information needed and classify supporting documents like financial statements submitted with your PPP loan application.
  • Amazon Augmented AI (Amazon A2I): Makes it easy to build the workflows required for a human to review the output from any ML or AI service, providing you the human oversight needed for precision. In a time when you want to make sure you’re in compliance with the PPP and finance laws, having a human checking these forms and applications can add human oversight to ensure compliance.

Using one or a combination of the preceding services can help you process more applications and financial data from both structured (like mortgage applications) and unstructured sources (such as social media) quicker, while giving you the option to add a layer of human review. If you are interested in learning more on how to get started using Amazon Machine Learning, we have a few resources available for you:

Learn More or Get Started

Learn more about Amazon AI and ML services, contact our sales team if you have more questions or download our eGuide on automating PPP loans.

Ready to get started?  You can use a number of resources like code samples and GitHub examples for all the services mentioned previously. Amazon Textract has a video demo on automating PPP loan applications by Kashif Imran, Principal Solutions Architect at AWS, that can help you get started now or you can use the GitHub sample code specific to PPP loan applications.

Partner Resources

We have a wide range of AWS Partners that are helping our customers get started with these services, which are available via our Amazon Partner Network (APN).

Andrea Morton-Youmans

Andrea Morton-Youmans

Andrea Morton-Youmans is a Product Marketing Manager on the AI Services team at AWS. Over the past 10 years she has worked in the technology and telecommunications industries, focused on developer storytelling and marketing campaigns. In her spare time, she enjoys heading to the lake with her husband and Aussie dog Oakley, tasting wine and enjoying a movie from time to time.

John Kain

John Kain

John leads Worldwide Business & Market Development for Banking and Capital Markets at Amazon Web Services (AWS). He works with customers to help them transform their existing businesses and bring new, innovative solutions to market by leveraging AWS services. John has more than 20 years of experience developing solutions for financial institutions. Prior to joining AWS, he led key programs for J.P. Morgan, Nasdaq, and two venture-backed financial technology companies. John holds a Bachelor of Engineering in Electrical Engineering from the Cooper Union for the Advancement of Science and Art.