Artificial intelligence is already reshaping parts of the mortgage process — and it’s moving faster than some in the industry may realize.

At a recent lender panel, several executives shared how they’re integrating AI into everything from pre-approvals to document scanning.

But while automation is accelerating, the consensus was clear: underwriters still have a vital role to play, especially as deals grow more complex.

“This is a people business. The underwriters aren’t going anywhere,” said Andrew Gilmour, Senior Vice President, Residential at CMLS Financial. Gilmour described how CMLS has already built an end-to-end AI-driven approval process and is now testing full automation for certain deals.

“The goal is not to replace humans — it’s to eliminate repetitive, low-value tasks so we can redeploy our people to where they’re needed most: product development, training, and complex deal structuring,” he said.

Gilmour framed the adoption of AI as a game-changing advance for the industry:

“In two to three years, [in AI] we’ll be going from the horse and buggy to cars, and it’s something that I think has got to be embraced.” –Andrew Gilmour, CMLS

Devon Ajram, Vice-President and National Director of TD’s Broker Services, noted that TD has been investing in AI for years, including through its acquisition of Toronto-based AI innovator Layer 6.

He said those investments have positioned TD at the forefront of AI integration.

Much of TD’s AI deployment so far has focused on colleague- and customer-facing tools, aimed at improving the advice conversation and enhancing customer solutions. Ajram emphasized that the bank’s focus is primarily on internal systems rather than fully automating adjudication.

“We’ve done some piloting around AI decisioning for pre-approvals,” he said, adding that TD also uses AI in forecasting and modelling to manage adjudication capacity on its proprietary side. Looking ahead, the bank is developing a segmentation scoring system that would allow customers with complex credit needs to be routed more efficiently to the appropriate retail risk team.

Ajram was clear that the intention isn’t to replace underwriters, but to support them.

“We’re not going to be closing underwriting departments tomorrow, and I doubt that’s going to be in our future,” he said. “This is still very much a collaborative tool — not something meant to replace the human element.”

AI gains traction in prime lending—but complex files still need a human touch

First National is focusing its AI efforts on alternative lending, where complex documentation and non-traditional income sources can present unique challenges.

Elena Robinson, Vice President of Residential Sales, said the lender has been testing tools to streamline bank statement reviews and scan income documents like pay stubs and letters of employment.

“There’s a place for AI,” Robinson said, noting that while the technology can help reduce turnaround times and assist with fraud detection, it’s not yet ready to replace experienced underwriters, particularly given the growing complexity of both prime and alternative deals.

“There are still so many factors you have to look into,” she said. So yes, AI may help in terms of documentation, but when it comes to the underwriting itself, you still need that human perspective.”

First National is also looking into auto pre-approvals — a more straightforward use case for automation — but Robinson stressed that broader adoption will take time. “It’s still in the beginning stages,” she said.

Nick Kyprianou, President and CEO of RiverRock Mortgage Investment Corporation, said his firm is using AI behind the scenes — not for adjudication, but to support analytics, reporting, and marketing efforts.

“If you put enough data into it, you can start doing an analysis on your clients, where they’re coming from, which ones are working best—it builds a lot of reporting,” he said. “So, the better you know your your clients, you’ll be more efficient in doing your business.”

Lenders expect massive gains in underwriting efficiency — but not at the expense of advice

Gilmour expanded on CMLS’s AI capabilities, noting that the lender has been testing fully automated pre-approvals using algorithms aligned with internal credit policy. If a file doesn’t meet the standard rules or finds inconsistency, it’s kicked out to an underwriter for review.

Currently, about 10% of CMLS’s loans are fully committed using rules-based algorithms, he noted. “We’re here now. We can auto-approve full files all the way through with AI,” Gilmour said.

“All we’re trying to do with this technology is augment the service levels, allow all of us to be more efficient and I think the reality is there’s going to be 100x improvements in terms of underwriter efficiency within two to three years,” he added. “And that’s not like just saying it, we’re seeing it already.”

Still, Gilmour said the end-consumer likely won’t notice much of the change. And that’s fine, because the human element — especially when it comes to providing guidance — isn’t going anywhere.

“They still need advice. This is still the biggest decision that they’re presumably going to make in their life as it relates to assets and liabilities,” he said. “And so we really want to get rid of the noise that’s associated with checking and reviewing basic stuff and get back into the business of training our staff on solutioning, working on product development and so forth. Our underwriters aren’t going anywhere.”

adjudication AI AI in mortgages Andrew Gilmour cmls Devon Ajram Editor's pick lender panel mortgage underwriting td technology underwriters

Last modified: April 11, 2025