Loan origination is ripe for automation using artificial intelligence. Manual processes for validating applicant identity, income, assets, and eligibility are inefficient, inconsistent, and costly for both lenders and applicants. AI offers a path to radically transform loan underwriting by improving speed, accuracy, and the applicant experience.
This article explores how generative AI and predictive modeling are modernizing the mortgage industry through enhanced decision-making. Key topics include:
- The role of decision AI in loan underwriting
- A case study on Direct Mortgage's transformation
- Benefits including faster approvals, enhanced compliance, and applicant conversions
- Future directions, like automated appraisals and personalized recommendations
Generative AI Gives Mortgage Lenders a Competitive Edge
Let’s start off by looking at a recent case study.
Direct Mortgage is pioneering the future of mortgage lending. Despite nearly 30 years in business, their loan application workflow was still predominantly manual and paper-based as of early 2021.
CEO Jim Beech had unsuccessfully tried automating it for years before partnering with us. In just 30 days, we customized an AI agent that could accurately classify and extract data from paystubs; the most complex document type in an application.
This initial automation improved applicant approvals from weeks to days. With increased confidence in AI, we expanded to over 200 document types including bank statements, tax forms, and insurance documents. Today, AI agents handle the entire application process - classifying documents, extracting and validating applicant data, and underwriting entire loans with minimal human involvement.
According to Jim Beech, AI agents have reduced their approval time from weeks to minutes and lowered document processing costs by 80%. Just as importantly, faster and more accurate loan decisions have increased applicant satisfaction and conversions.
Decision AI Levels the Playing Field for Mid-Market Lenders
Generative AI refers to models that can generate content, text, code, images or other outputs given custom inputs and prompts. Generative models have fueled innovations like chatbots.
In the context of mortgage underwriting, generative AI ingests applicant data from forms and documents to determine eligibility and risk. But unlike rigid traditional software, it dynamically adjusts its evaluation methodology based on changes to lending guidelines, applicant profiles, and loan products. This adaptability, hyper-personalization, and automation is fueled by predictive modeling and machine learning algorithms underneath the hood.
Predictive models utilize historical training data to determine the likelihood of various outcomes, such as whether an applicant will default or prepay their mortgage. When paired with generative AI, predictive models enable complex decision making that optimizes lender objectives like balancing profitability, defaults, and conversion rates across applicant segments.
And implementation is rapidly becoming turnkey: Freddie Mac's Loan Product Advisor recently added access to third-party verification services and fraud tools as well as integrated assets, income, employment and identity verification functions. Loan Product Advisor distills insights from its analytics models into simple recommendations, allowing originators of any size to compete with large banks.
The Future: Ubiquitous Decision Intelligence
Mortgage lending is just the tip of the iceberg. Across finance, insurance, and healthcare, manual document review and subjective human decision-making result in high costs and inconsistent applicant experiences. AI-based automation will inevitably penetrate most facets of document-intensive workflows including:
- Claims processing - evaluating coverage and determining approvals or denials
- Policy underwriting - establishing risk levels, appropriate coverages and premiums
- Compliance - identifying documents, data, or processes that violate laws or regulations
As AI adoption accelerates, expect to see AI-driven recommendations and automated approvals become standard across the finance, insurance and healthcare sectors, leading to faster and fairer decisions. Key trends to watch include:
- Compound improvements from model iteration - With more training data, predictive accuracy continually improves
- Stacked models for advanced capabilities - Combining multiple algorithms provides more representative outputs
- Workflow augmentation versus pure automation - AI's highest value propositions enhance human decision making rather than replacing it
Across verticals, achieving AI's full potential requires pragmatism - focusing investment in areas where technology gaps allow for competitive differentiation and measurable value creation. Lenders must be willing to embrace usability and conversion as equal priorities alongside compliance and risk mitigation.
As pioneers like Direct Mortgage demonstrate, however, generative AI and decision intelligence are set to transform applicant journeys by eliminating friction through automation. Competitors who fail to effectively leverage data and AI risk obsolescence within the next 3 to 5 years. The opportunity to reshape market positions by deploying AI, therefore, is urgent and substantial.