Back to Industry Trends & Insights

AI vs. Outsourcing Takes Center Stage at Industry Conference

Ai Blog 8 28

AI Foundry’s Alex Levi Presents to a Full House at Lenders One Conference in Seattle

Alex Levi, VP Sales
August 28, 2019 

Earlier this month, I presented at the Lenders One Conference in Seattle. Lenders One is a Co-Op of about 150 mid-market lenders, and their events are a great forum for learning and exchanging ideas about industry best practices, market conditions, industry trends and technology. At their last conference, held six months ago in Austin, I was on stage for a discussion on the role of automation and AI in the loan manufacturing process.

This time around, the conference topic was AI vs. BPO – a hot topic in the industry that is getting more attention all the time. I shared the stage with the President of Trellix, an Altisource mortgage BPO company, and we spoke to a full room of engaged attendees as we tackled this challenging topic. In summary, my narrative came down to five key points:

  1. Using AI in lending operations results in significant automation opportunities.
  2. AI technology and machine learning algorithms enable faster, more accurate automation than OCR-based methods.
  3. AI is capable of making decisions and taking actions based on data collected from documents and LOS, and as a result can perform tasks that are currently performed by humans in minutes vs. hours.
  4. AI enables game-changing outcomes, including lower loan manufacturing costs and much faster loan review.
  5. Learning how to use AI is a critical skill and is highly important to the survival of a lender. Those that master it are going to thrive, while others that do not are going to be replaced.
  6. BPO organizations traditionally deployed automation technologies to lower their own delivery costs and ensure a better business outcome for their clients.

After my presentation was finished, I answered several questions from the audience. Here is a summary of some of the top questions.

How Do You Get Started with AI?

The most important thing to do is to get started, one way or another. This technology is new, its impact on processes and people is significant, and every company is going to implement and adopt practices that best suit their culture and skills. Examples of where to get started could include some (relatively) small and easy to implement tasks, such as

  • Taking an incoming group of documents and separating and indexing it in minutes vs. hours with no need to check and triple check outcome.
  • Running closing checklists so that the closing function can be significantly automated.
  • Automating income and assets document analysis.

Why invest in AI that deals with documents if more and more data is available digitally, i.e. appraisals, income and assets documents?

Most of the data that is used in loan manufacturing comes from documents that enter the process as PDF images. Even if documents are delivered electronically, they are images of documents. It is true that providers like Blend and similar can deliver borrowers income and assets data as data not documents. Users are required to provide their passwords to online financial services to enable this capability. There are many documents that are not available via online services. Many documents that start off as digital data get converted into PDF’s and images because a large portion of the downstream workflows in settlement and closing are still paper-based.

Market data suggests that at best, there is going to be around a 50-50 split between documents and digital data for a long time to come. Irrespective of how the data is acquired i.e. digitally or through image recognition, AI investments are more about ongoing learning and adjustment of business workflows based on these learnings. Our rules engine implementation is driven by both – data extracted from documents as well as data delivered electronically.

Where in the lending process is AI around document recognition ready for production?

While there may be some natural skepticism about using AI for document recognition in production environments, we are looking for early adopters and fast followers who understand that in the beginning, document classification and data extraction accuracy rates might be lower than 90%, but post implementation it is going to improve significantly. Over time, as more and more loans are processed, it is going to reach better than human-level accuracy in the 93-98% range. We want to work with organizations that have a partner mentality and are willing to team up with us to close the accuracy gap.

In all, there is significant opportunity for lenders who embrace AI. From faster processing, to reduced costs, to higher volumes, better accuracy rates and more, AI is going to be a key competitive differentiator that will not only give an organization an advantage in the market, but may be essential for its survival as the industry evolves over time.


Connect with the author: 
Alex Levi
Vice President of Sales. AI Foundry