Renters and homebuyers face many challenges this year, especially when it comes to affordability. But technology—particularly using generative artificial intelligence—can help solve these challenges in new ways. At least, that was the premise of last week’s FHFA 2024 TechSprint, a conference focused on innovative housing use cases for a new generation of artificial intelligence.
The sprint event culminated with a demo day on Thursday, where 12 teams presented their solutions to a panel of technical, regulatory and housing experts. These teams have achieved some impressive results, providing a roadmap for businesses looking to harness the next generation of artificial intelligence.
The use cases are organized around four focus areas:
- Consumption experience
- Assess creditworthiness
- Operations and Risk Management
- obey.
Last year, FHFA conducted its first technology sprint, focused on the front end of mortgage lending and increasing technology adoption. This year, FHFA expanded the housing issues they want to consider to include multifamily housing and services, with a special focus on areas where artificial intelligence can be deployed.
Every team that proposes a use case must address risk and compliance issues. Hallucinations, false positives, and inaccuracies are just some of the risks that have been identified, and artificial intelligence is less of a “set it and forget it” than other types of technology. Artificial intelligence is always learning, which means it must be fed and trained with accurate, up-to-date data. These updates must be built into the process from the beginning.
Below is a summary of all 12th generation AI proposals. The first four listed are considered by FHFA to be the most promising use cases for their specific focus.
domino multifamily rentals
- Domino will provide tenants with personalized readiness ratings, providing an Uber-like experience for today’s 30 million tenants and property managers.
- The app pulls credit reports, runs background checks, and pre-approves tenants at specific prices so tenants know how much they can afford each month.
- Tenants receive pre-qualification results that meet their criteria, and the app connects them with landlords/property managers.
- Compliance Note: There are many opportunities for fraud (especially with other AI), so guardrails are needed.
- Awarded the most promising artificial intelligence user experience use case.
Homeowner Assistance Model (HAM) single family services
- The focus is on early identification and intervention for distressed borrowers, starting as late as 32 days in order to keep them in their homes.
- HAM flags loans, conducts reviews, develops intervention plans and contacts borrowers to review options. HITL validates the plan.
- Serves as a relationship-building tool for service personnel.
- Compliance Note: Personally identifiable information must be removed before artificial intelligence is involved, and human verification mitigation plans are required.
- Awarded the most promising use case for artificial intelligence credit assessment.
Fraud detection tools Multifamily
- Leveraging artificial intelligence to combat AI-related fraud, including deepfakes, in multifamily operations.
- Replace manual, people-based processes to increase the speed and accuracy of fraud detection.
- The best use case is for large multifamily operators as it requires extensive training material to complete.
- Compliance Note: Synthetic data can eliminate the need for large data sets.
- Won the Most Promising Artificial Intelligence Operation Application Award.
ADA compliant Multifamily
- A web application that uses image assessment to ensure multifamily units are ADA compliant and thereby meet GSE fair housing goals.
- ADA Comply uses photos to flag which parts of a unit are out of compliance and produces new photos showing how to fix them, such as a before-and-after comparison of the property.
- Cost estimates for repairs will also be provided.
- Compliance instructions: Adopt “no sticks, only carrots” incentives.
- Awarded for Most Promising Risk and Compliance Use of Artificial Intelligence.
loan friend Single parent family origins
- Loan Buddy is dedicated to helping limited English proficient (LEP) buyers through the mortgage loan process by converting documents, explaining terms, providing an overview and answering questions in the language of the user’s choice.
- Loan Buddy is designed as a collaborative robot that will ask a question to someone on the loan team if it is not confident about its answer. Conversations with lending partners will be recorded and provided to the lending team.
- Many LEP borrowers are mobile-first, so Loan Buddy will be optimized for mobile devices.
- Compliance Note: Data will be deleted after 7 days to address privacy concerns.
Tools for bringing new capital into the multifamily space multifamily investing
- This tool will simplify multifamily investing to attract new players and/or make it easier for existing investors to purchase or renovate properties.
- This tool provides information on all multifamily opportunities within a given area that match the user’s financial and experience levels. Basically an AI mentor who holds investors’ hands throughout the process and provides opportunities and resources.
- The application identifies available aid programs, summarizes forms and builds a business case for investment.
- Compliance Note: Drawing information from vetted sources should reduce risk.
DPA Navigator Single parent family origins
- Streamline and automate the down payment assistance process for DPA providers and lenders through the DPA Search App, which matches borrower and loan characteristics to currently available DPA resources.
- Leverage your lender’s existing workflows to access your DPA more easily. Obtaining data for AUS via large language models.
- With down payment assistance, 5.8 million renters could afford their mortgages.
- Compliance Note: Data integrity is critical and AI requires the most up-to-date information as DPA funds have been used and are no longer available.
AI 4 Fairness Single Family Home Refinance
- The goal is to increase refinancing opportunities for Black and Latino borrowers by improving their creditworthiness and ensuring fair valuation of their properties.
- Increase refinance education for borrowers who traditionally don’t even try to obtain refinance.
- Guided by artificial intelligence, homeowners measure and take photos of their homes. The tool checks data accuracy against other sources and filters out sensitive private information and images to reduce bias in the process.
- Compliance Note: There is a risk that property data cannot be accurately analyzed and this risk can be mitigated by verifying the data from other sources.
Mortgage Thinking Artificial Intelligence The Origins and Services of Single-Family Homes
- Consolidate and synthesize regulations to give lenders and servicers the answers they need in one place because it covers all agencies and regulators.
- Unlike current solutions that only superficially treat legal texts, Mortgage Mind will provide guidance on institutions and circumstances that may affect interpretation.
- Conceived as an industrial utility.
- Compliance Note: Inconsistencies and overlaps in guidance between different federal agencies and between state and federal agencies make this difficult.
Gen AI Buyback Solution Single parent family origins
- By controlling the number of repurchases through artificial intelligence, artificial intelligence can “liberate” data from the GSE’s information-rich repository to better understand risks and identify issues such as bias.
- This can be achieved by using synthetic data to train underwriters and predictive models to create a “digital twin” of the mortgage market.
- Synthetic data is artificially generated data through simple rules, statistical modeling and simulation, and is used by SWIFT and other financial organizations.
- Compliance Note: Ethics controls must be in place.
Requiem Single parent family origins
- Provide immediate compliance and quality control feedback based on existing regulations and company policies to reduce origination costs.
- Reversal the QC process so it happens from the beginning.
- Collaborative robots listen to customer conversations and provide immediate feedback. For example, it can detect that a loan officer is not extending a VA loan to a qualified borrower, so it prompts this.
- Compliance Note: A mortgage-specific LL.M. utilizing open source is required.
Precision Artificial Intelligence Compliance Engine (PACE) single family loan
- Industry-sponsored solutions led by the FHFA Alliance will lower the barriers to entry for artificial intelligence adoption.
- PACE will serve as a trusted third party to verify transactions and create artifacts for future audits.
- The Situation-Aware LLM will continually assess policy compliance, instant loan quality assessments, and look at every application, not just a sample.
- Compliance Note: Consumers provide consent multiple times over the course of a long session.
