Risk and benefits of using AI in mortgage marketing

prasit2512 - stock.adobe.com

Generative artificial intelligence holds a wealth of potential — and risk — for the mortgage industry, but despite the challenges, the developing technology is finding its place within company workflows. 

Some of the greatest potential for adoption lies in marketing uses. 

"Imagine if you Googled a topic, and then clicked through several links, and then summarized what you found in those links. Imagine if a machine could do that for you in 30 seconds," said Adam O'Daniel, chief marketing officer at Guild Mortgage.

"It's not delivering me any data that I couldn't have probably found through Google search. It just saved me the time and in sorting through it and compiling the data."

Across business segments, AI is demonstrating value as a tool that drives efficiency and even fuels inspiration among marketing professionals, even though widespread apprehension remains. Although mortgage and real estate companies have the same concerns around risks as others, their marketing teams and loan officers are testing the waters to varying degrees and learning to tailor AI for their specific needs.

"It's a starting point for many, and it has been helpful if you're, for instance, having a creative block," said Whitney Blessington, chief marketing officer at Churchill Mortgage. "We call it like someone to brainstorm with, even though it's not a person."

Generative AI benefits also result from its ability to conduct quick research. "It can help you come up with good topics," O'Daniel said.

A useful but still-developing technology opportunity
Mortgage companies, more than other types of companies, appear open to exploring how artificial intelligence might help their marketing efforts. 

While some forms of AI are already used in the underwriting context, especially for tasks related to data extraction and processing, concerns about enforcement of possible noncompliance leave some lenders wary about applying the technology in a customer-facing capacity. Marketing tasks, though, offer the opportunity to see how AI can improve efficiency within the appropriate guardrails. 

In 2024 research released by Arizent, 64% of mortgage industry professionals said they would be open to utilizing artificial intelligence for a majority of their marketing and promotional tasks in a hypothetical scenario where regulations did not exist. Interest in the mortgage industry far exceeded the percentage of similar responses in six other financial sectors, none of which surpassed 50%. 

At the same time, 55% within home lending said they would use it for most tasks associated with research and fact checking. 

Its use in advertising, though, still presents some risk of bias in outreach, according to recent guidelines issued by the U.S. Department of Housing and Urban Development. 

But despite the industry's enthusiasm, the "A" in AI doesn't stand for accuracy, and human marketing professionals will need to remain a fixture, mortgage leaders say. Even when used for research purposes, users have found themselves running into factually incorrect responses.

"You can't count on it blindly," Blessington said. "You still have to do your homework." 

"I think the biggest thing is, today, it really helps someone streamline their workflows," she added, comparing it to an intern who might conduct low-level administrative work, such as writing metadata descriptions or alternative text for images. 

"It helps you go from ideation to planning to actual content," said O'Daniel. However, when generative AI "writes" any of its own content itself, it fails to perform to the standards the industry might want, he said.   

"It may use terminology that is more appropriate for a bank and not an independent mortgage lender, and so you have to adjust the terminology. Some of the more finer nuances of the business — it doesn't fully deliver."

Current use scenarios and risks
Use of artificial intelligence, particularly generative AI like ChatGPT or Microsoft Copilot, is still in its nascent stage within the mortgage industry; but with expectations of rapid expansion, it stands to change how future work can be done.

Entering AI waters may seem daunting, but the technology also offers customization that can facilitate ease of use, according to Ginger Bell, who regularly conducts seminars on artificial intelligence for real estate professionals. Bell is a co-host of the podcast AI Clubhouse and founder of housing industry video platform Edumarketing.com.

A loan officer or lender can customize their generative AI to home in on situations or guidelines it commonly addresses. "You can actually just type a scenario, and it reads the guidelines," Bell said, while cautioning verification is still necessary.  

"You can also ask it to cite exactly where it's pulling that information from, and a lot of it is just training it to be able to ask the questions correctly, telling it what you want in terms of the response and then how you want that response to look." 

Bell commonly sees ChatGPT being used to assist in composing emails and social media posts, and some mortgage professionals also employ it to write video marketing scripts. Users can  tailor a gen AI tool by feeding it their previously written transcripts, articles or other work, eventually training it to sound more like their own voice, she said. 

But oversight and enhancements need to remain top of mind as well, said Jason Perkins, co-founder and president of Bonzo, a provider of communication engagement software and a mortgage customer-relationship management system. 

"I look at AI-generated content as a frame of your business, not the be-all," he said. "Personalization is what drives conversations."

Generative AI can also quickly build marketing campaigns through a series of prompts —  a set of instructions or steps to create messages with given parameters that might address a specific topic or target a borrowing segment. The prompts can also ensure that necessary disclosures and licensing information are included.

"A lot of companies need to realize this is a big compliance opportunity to make sure that your loan officers are providing their information in a compliant way," Bell said.

However, while businesses have the capability to personalize their prompts and content via an open source generative AI platform, a number of companies are instead turning to enterprise versions that protect proprietary information and maintain compliance. Certain accounting firms go as far as requiring employees to use personalized generative AI under enterprise editions that remain closed sources, according to Bell. 

"There's a lot of folks who use what's available to consumers on ChatGPT and other platforms like that, and certainly, it's a great tool, but we're trying to be very thoughtful about how to use those platforms," O'Daniel said.

"You use a public platform — the data that I upload to the model stays with that model to fuel future learnings, which is amazing; but we might want to share information from a product guide or some other company program that we don't want to be out of our control," Guild's marketing leader added. 

When using a public platform "be aware as far as not putting any nonpublic information in there because it is open source," Bell advised. In addition to potential noncompliance, it opens up businesses to cybersecurity risk. 

Reliance on public artificial intelligence platforms without proper vetting of the content they produce also carries risk of potential copyright infringement, according to Perkins. 

"They're just aggregating data off of the internet," he said. "Businesses and companies are going to put fences around their data," meaning companies need to be aware of how loan officers and staff use AI-generated content in social posts or advertising. 

Future potential and customer trust 
While marketing content crafted from AI has primarily appeared in written form, artificial intelligence is taking hold in other creative outlets. "Now there's so many new technologies that are being built around this," Bell said.

Advanced generative AI tools that alter photographs already exist, alongside emerging businesses that produce original imagery and videos based on an individual's likeness and voice from a single recording.

However, while AI-generated imagery video represents one of the next growth phases for automation, it also brings with it a potential for misuse by fraudsters and a conundrum for businesses of all types who want to use technology to their advantage without eroding relationships with clients. 

"I think there's a number of questions around how that affects your brand," O'Daniel said. 

"It can go both ways. There are people who would appreciate more frequent informational updates from their lender and from their loan officer. So if the technology can help us deliver more frequent helpful information, that can build trust; but if the customer feels as if they've been misled and that this avatar is not really their loan officer, that can destroy trust. So I think we have to be very cautious."