MBA Annual 23: Decoding the AI playbook for mortgage CXOs ?
The industry narrative is shifting from "What is Generative AI" to "How do we use Generative AI" - what does this mean for mortgage CIOs ?
The upcoming MBA Annual 23 is drumming up significant fan-fare among industry leaders not only because it’s the last major mortgage conference of the year but also because of the range of topics it touches upon.
A lot has happened over these last nine months in the context of overall market dynamics & macroeconomics - an underperforming purchase market, rising inventory shortage, consecutive Fed rate hike and a looming recession.
On the technology side, things haven’t been completely uneventful either.
With Generative AI becoming season’s hot favorite - literally overnight, there’s a lot of “figuring out” that’s still going on.
For mortgage CXOs, the question arises - how to blend Generative AI into their existing people, process & tech stack in a way that’s win-win, given the current high interest rate environment ?
This is where MBA Annual’s tech tracks, assumes relevance.
It touches on three key areas -
AI in lending,
Maximizing ROI of existing tech stack
Evaluating latest digital solutions
A small note before moving ahead ..
If you are headed to MBA Annual 2023, then let’s meet for coffee.
Now back to our deep dive -
AI in lending
The tech track on “AI in lending” attempts to deconstruct the pros & cons of implementing the latest Generative AI capabilities in loan production & servicing.
This couldn’t have been timed better.
With Generative AI, the range of impact areas across the lending lifecycle, has increased significantly.
In our understanding, there are four potential areas where Generative AI can be particularly helpful:
Improving loan application experience for homeowners: One of the biggest advantages of LLM is its ability to understand & generate natural language text. This is particularly useful - in guiding customers through loan applications, helping them fill out the forms correctly and addressing any query they may have during the process. Further, this could drastically reduce the cognitive load for applicants while cutting down the back & forth with loan officers.
Assisted credit analysis: A fine-tuned custom LLM application can help credit analysts process information such as credit score & financial history - to assess risk levels of a loan applicant.
This could involve prompting the application with simple queries in natural language to mine insights from credit reports, income statements, tax returns, and other financial data. It could also provide decision support with detailed explanations to brisk up risk assessment for credit analysts. At a later stage in the loan lifecycle, Generative AI could be deployed to monitor loan performance and flag delinquencies & default risks much ahead of time than traditional predictive analytics.Loan Underwriting: Just as in credit analysis, underwriting is another area where Generative AI can provide actionable & real time decision support. Typically, income verification & financial statement analysis can be automated using an IDP solution. However, by adding a layer of custom LLM, underwriters can automate workflows such as - risk scoring, loan pricing and loan product recommendations.
Loan Servicing: Using Generative AI along with intelligent document processing (IDP), can help loan servicing teams automate a large part of their servicing portfolio operations. For example - using an IDP solution, bulk loan files can be accurately boarded into the servicer’s system of records.
Once boarded, Generative AI can be used in downstream workflows to provide personalized recommendations for loan repayment & assist borrowers on issues such as payment reminders, billing inquiries, and account management
The above areas of applications are just tip of the iceberg.
Each of these stages of the lending lifecycle, could have multiple use cases depending on how Generative AI sits on top of an enterprise’s existing data sources & applications.
One such use case could be mining instant answers from bulky documents such as Freddie Mac Seller’s Guide.
At Vaultedge, we created a Generative AI chat application to demonstrate how you can ask questions off complex mortgage documents and extract not just data, but also detailed explanations.
Here’s a quick demo of how Vaultedge Generative AI works for mortgage documents.
Going back to MBA Annual - the dedicated session on “AI in lending” brings all key parts of Generative AI under one thematic umbrella, so that mortgage CXOs have an unfragmented view of what could work in short to medium term.
Maximizing ROI of existing tech stack
This session hosted by Blend on “Making the most of your tech stack” - intends to explain different methods of evaluating the performance of your existing technology stack against current business processes & outcomes.
The focus here is to offer a framework that could help CIOs unearth value maximizing opportunities while improving adoption of existing technologies, rather than onboarding new ones.
This is especially true in the current market conditions, where leadership teams are looking at cost optimization opportunities while making strategic investments in projects that aligns with their medium to long term roadmap.
Evaluating latest digital solutions
A CIO’s tool-set should not be limited to just existing tech stack or Generative AI. It should scan the broader solution ecosystem to spot what’s upcoming & relevant to current business realities.
This is where Gartner’s technology hype-cycle has been a leading indicator to help CIOs look in the right direction.
Very interestingly, both Generative AI & Intelligent Document Processing (IDP) - land on the rising curve of Gartner’s latest Hype Cycle Report.
Traditionally, IDP has been used for data extraction in loan pre-processing & some extent in mortgage underwriting.
When paired with Generative AI, IDP solutions can gain interpretive abilities.
This means not only recognize documents & extract data, but also interpret the linkages between individual files such as loan applications, W-2, 1003, paystubs etc - as a connected whole.
Such fine-tuned LLM powered IDP can perform much complex & goal seeking tasks such as:
Summarize loan documents,
Interpret agency guidelines & clauses specific to a borrower’s loan file,
Provide decision support at various steps - income verification, generating loan estimates & spotting potential frauds.
With ongoing usage, such an IDP system can become a “decisioning co-pilot” to the loan processing & underwriting teams.
Given that both Generative AI & IDP are early in their lifecycle - MBA Annual’s session on “The Latest in Digital Solutions” aims to spotlight not just areas where emergent tech can play a role but also explore a range of digitization opportunities from eNotes through automated verifications to digital wallets.
Under this single umbrella theme, the session will cover new opportunities to reduce cost of lending & improve overall borrower experience.
Wrap up
We may be a little biased towards these three sessions at MBA Annual - as they resonate closely with the work we do at Vaultedge.
Over the past few years, we have worked with industry leaders like Ocwen PHH & BSI Financial in automating manual processes around document indexing & data extraction.
In the process we have been able to deliver not just significant time & cost savings but also build a lifelong trust - something we’re super proud of.
Before we wrap up, here’s to all patrons in Philly - we’ll be at the MBA Annual 23 between Oct 15th to Oct 18th.
If you’re at the event, let’s meet for coffee & see if there are areas that we could jointly work on.
See you in Philly, then !