From cures for rare cancers and fully autonomous driving to shortened workweeks and increased leisure time, the disciples of artificial intelligence preach its life-changing potential for every corner of business and society.

Yet since ChatGPT was launched in 2023, kicking the AI revolution into overdrive, life in the mortgage industry has largely remained business as usual.

Findings released Monday from a first-of-its-kind survey examining AI adoption in the mortgage industry show regulatory uncertainty and meager productivity gains continue to hinder broader adoption. That has left most industry stakeholders unconvinced that the technology — at least so far — is living up to the hype.

Advertisement

“Mortgage lenders trail other financial institutions (e.g., insurance, banking) in AI development,” read a report jointly published by the Mortgage Bankers Association (MBA), Boston Consulting Group and the American Association of Residential Mortgage Regulators (AARMR), a national organization representing state mortgage regulators.

The three groups said in a press release that the wide-reaching survey represented “the first collaborative effort by MBA, lenders, servicers and state regulators to examine the same set of AI adoption and governance questions.” Participants in the survey account for about 40% of annual origination volumes.

The survey examined AI deployment across 38 use cases over the mortgage lifecycle, in addition to prevailing governance and oversight norms. Use cases fell into five broad categories: sales and marketing; origination; servicing; secondary and capital markets; and corporate functions and operations.

AI investments to accelerate

Nearly 95% of respondents said they have at least one use case in which they are actively deploying AI, but only 80% said they have at least one use case fully scaled.

The average respondent had about 10 of the 38 surveyed use cases in production, but 100% of lenders said they expect to increase their AI spending over the next year. To that end, more than half of large organizations — defined as those with more than $5 billion in annual revenue — said they expect to invest more than $50 million in artificial intelligence in 2026.

Even as industry adoption of AI has lagged other financial services sectors, consumers report high expectations that AI will be used during the mortgage process. Confidence that AI will lead to positive outcomes during that process deteriorated in 2026, however, compared to how borrowers felt about AI the year before, some surveys show.

That disconnect crosses into the industry setting as well. Some experts have described the enterprise AI entering the mortgage space so far as a massive letdown. Others have told Scotsman Guide that the cyber-related risks AI poses to companies and borrowers are accelerating faster than industry safeguards.

The recent survey — which was fielded online from April through July — is an attempt to identify and begin bridging some of those disconnects.

“These findings give the industry a clear picture of the opportunities and challenges ahead, from regulatory uncertainty and governance to turning early adoption into measurable results,” said Rick Hill, vice president of industry technology at the MBA, in Monday’s press release.

The MBA, for its part, published a white paper in June that called for a “unified, principles-based risk management framework” in the absence of clear regulatory guidance for industry participants thus far.

Constraints to AI adoption

In addition to drafting a new framework for novel risks, current mortgage laws should reflect AI’s impact on existing compliance standards, the MBA believes.

But regulatory uncertainty still emerged as the “binding constraint” to faster AI adoption, cited as the largest barrier by 59% of respondents, followed by 45% of respondents expressing uncertainty around return on investment. No other constraints were cited by more than 25% of respondents.

Governance continues to lag adoption, the survey also revealed, even as Fannie Mae and Freddie Mac have rolled out mandatory governance and reporting frameworks. Underlining risks related to cyberattacks and data privacy, more than one-quarter of lenders reported that “shadow AI” use is occurring outside of approved tools and environments.

Nearly all organizations have “basic AI controls,” according to the survey, but less than 60% of respondents said they continuously monitor models once they are live.

AI adoption has been largely concentrated in workflows related to productivity and document processing, the survey results show, such as corporate functions and origination tasks. Mortgage capital markets and default servicing are “largely untapped.”

But it is still early days concerning AI adoption in the mortgage industry, the report emphasized, underscoring opportunities for cooperation between industry stakeholders and regulators as the technology matures.

“This survey gives regulators our clearest picture yet of the AI landscape within the mortgage industry,” said AARMR board member Cliff Charland in Monday’s press release.