Is AI in the mortgage industry?

What is AI in mortgage lending?

Artificial intelligence in mortgage lending refers to advanced computer systems that can learn from data, recognize patterns and make decisions with minimal human intervention. Unlike traditional automated underwriting systems that follow rigid, preprogrammed rules, AI uses machine learning algorithms that continuously improve their performance by analyzing vast amounts of historical loan data and outcomes.

At its core, AI in the mortgage industry leverages several key technologies:

• Machine learning enables systems to identify complex patterns in borrower behavior and creditworthiness that humans might miss.
• Natural language processing (NLP) allows AI to read and interpret documents like pay stubs, tax returns and bank statements, extracting relevant information automatically.
• Computer vision technology can verify document authenticity and detect potential fraud by analyzing visual elements of submitted paperwork.

AI vs. traditional automation

What sets AI apart from conventional mortgage automation is its ability to adapt and refine its decision-making over time. Traditional automated systems can only execute the specific tasks they're programmed to perform, while AI systems learn from each new application they process, becoming more accurate and efficient with experience.

This distinction matters because it means AI can handle nuanced situations and provide more personalized assessments rather than simply checking boxes against predetermined criteria.

For mortgage lenders, this represents a fundamental shift from rule-based automation to intelligent, adaptive technology that can enhance both efficiency and accuracy throughout the lending process.

» LEARN: How to qualify for a mortgage

Key applications of AI in mortgage lending

The potential impact of AI reaches far into aspects of the mortgage process for both the borrower and the lender, from the beginning of the exploration phase to closing and beyond.

Document processing and verification

AI can assist with the documentation review portion of the loan application process, including automating the review of paystubs, bank statements, tax returns, W-2s and other documents through optical character recognition (OCR) technology.

Instead of manually reviewing each document, AI systems can automatically extract and verify information in a matter of seconds. This can make quicker work of the submission process, which the borrower may appreciate, and can potentially reduce the number of human errors on the lender's side.

AI can also cross-reference information across multiple documents to ensure consistency, flag missing information and even detect alterations or fraudulent documents, improving both accuracy and speed throughout the documentation review process.

Automated underwriting and risk assessment

The use of automated underwriting is not a completely new concept, but more lenders may choose to adopt this approach as AI gains in popularity. For example, Fannie Mae uses a proprietary program called Desktop Underwriter for its underwriting process, which provides faster data validation, which in turn can make the underwriting process more efficient.

Using AI to create a well-rounded applicant profile can potentially benefit borrowers as well, since it may increase approval odds based on other factors outside the traditional approval parameters.

AI can use algorithms to assess a borrower’s creditworthiness in more advanced ways compared to the traditional underwriting process of relying on a credit score. Not only can AI analyze credit history, income and debt-to-income ratios, it can drill down even further. AI can analyze a borrower's spending patterns, social media behaviors and bill payment history to form a more robust picture of a borrower's financial habits.

AI also has the capability to recommend individual loan options for potential borrowers. It not only factors in personal financial information, such as income and spending habits, but it also can incorporate risk tolerance so a borrower receives a much more personalized loan recommendation. This can offer a helpful solution during the application process because it may suggest loan options that someone might be unaware of.

Customer service and communication

Chatbots are hardly a new concept, but they are a good reminder of the use of AI with customer service inquiries. Lenders can now use AI-driven chatbots and virtual assistants to provide more personalized assistance to borrowers by answering specific questions, offering guidance through the application process or facilitating communication with lenders.

Fraud detection and prevention

Whether a borrower is taking out a new mortgage or refinancing, AI can assist lenders in looking for application anomalies or red flags that might indicate fraudulent activity. AI can do this by analyzing any inconsistencies in applications, such as with employment verification, instances of identity theft or inflated income statements. AI excels at this because it can analyze patterns across thousands of applications simultaneously, identifying issues that human reviewers might miss.

AI fraud detection systems work in real time, flagging suspicious applications immediately rather than discovering issues weeks into the process. These systems can detect subtle red flags like mismatched addresses across documents, unusual spending patterns that don’t align with stated income or synthetic identities created by combining real and fake information. By catching fraud early, AI protects both lenders from financial losses and legitimate borrowers from identity theft.

Additionally, the mortgage industry is subject to a long list of regulations, which can make it harder for lenders to stay compliant. AI can help streamline the reporting process lenders face by automating much of the reporting.

» MORE: 9 mortgage questions to ask your lender

Pros and cons of AI in mortgage lending

While AI offers numerous benefits around efficiency and customer service, it has its downsides, too.

Regulatory compliance and ethical concerns

As AI becomes more prevalent in mortgage lending, regulatory oversight has intensified to ensure fair lending practices. The Consumer Financial Protection Bureau (CFPB), Department of Housing and Urban Development (HUD) and Federal Deposit Insurance Corporation (FDIC) all maintain strict requirements for AI in mortgage lending, particularly around fair lending laws like the Equal Credit Opportunity Act and Fair Housing Act.

AI bias in lending

One of the most pressing concerns with AI in mortgage lending is the potential for algorithmic bias. Because AI systems learn from historical data, they can inadvertently perpetuate past discrimination if that data reflects biased lending practices.

For example, if an AI model trains on loan data from periods when certain communities faced systematic lending discrimination, it might continue those patterns even without explicitly considering protected characteristics like race or gender.

Compliance requirements for lenders

To address these fair lending AI concerns, regulators now require lenders to take several protective measures:

• Regular model auditing and validation: Lenders must test their AI systems for discriminatory outcomes across demographic groups.
• Decision transparency: Systems must document and explain how algorithms make lending decisions.
• Disparate impact testing: Lenders must prove their models don't disproportionately deny loans to protected classes.
• Human oversight: Final lending decisions should include human review, not solely algorithmic determinations.

Lenders must also stay current with evolving regulatory guidance, as agencies continue refining their expectations for responsible AI use in mortgage lending.

Future outlook and market trends

The mortgage industry stands at a pivotal moment in AI adoption, with lenders weighing the technology’s potential benefits against implementation challenges and regulatory uncertainties. Understanding current adoption patterns and emerging technologies can help borrowers and industry professionals anticipate how AI will shape mortgage lending in the coming years.

Current adoption rates and industry surveys

Based on a recent Fannie Mae survey among lenders considering adopting AI, the appetite is strong for using AI in numerous applications, both in interactions with customers and behind the scenes. Lenders are particularly interested in increasing the use of AI in complex categories such as compliance, underwriting and property valuation.

“Lenders are grappling with record-high costs associated with mortgage origination, placing additional strain on already stretched consumers and borrowers,” explained Mark Hinshaw, the former CEO of Candor Technology, which uses AI to approve or decline mortgage applications for lenders. “Enter artificial intelligence: poised as the most significant catalyst for cost reduction and accuracy improvement in decades.”

However, the same Fannie Mae survey also showed that lenders need more confidence with AI, as it does not have a long track record of results with lenders and there can be potential bias in the program. The technology is still in its early adoption phase across the industry, with many lenders taking a cautious approach as they evaluate implementation costs, integration challenges and regulatory implications.

Emerging technologies and generative AI

The latest development in AI mortgage lending involves generative AI platforms similar to ChatGPT, which can create content, answer complex questions and assist with document generation. These tools show promise for streamlining loan officer workflows, generating personalized borrower communications and creating detailed property reports. However, these emerging technologies also present new challenges.

“While AI has made strides in mortgage origination, the complexities and regulations surrounding activities like underwriting remain formidable challenges. Generative AI platforms, in particular, face hurdles in overcoming biases and inaccuracies inherent in their training data,” Hinshaw said.

The industry is closely watching how generative AI evolves, particularly around issues of accuracy, explainability and regulatory compliance before widespread adoption occurs.