Mortgage AI Won't Fix Broken Operations, Industry Leader Warns

Operational foundation matters more than AI adoption speed

Mortgage lenders racing to implement artificial intelligence may be asking the wrong question. Rather than focusing on how quickly to adopt AI, companies should first examine whether their operational infrastructure can actually benefit from the technology, according to Emmanuel St. Germain, CEO and founder of Choice Mortgage Group.

The core insight: AI amplifies existing systems. When those systems are fragmented or unstable, automation simply scales dysfunction faster.

Why it matters

Many mortgage companies gutted their operations teams during the 2022 market contraction, creating patchwork workflows across multiple vendors. These firms now face a critical choice—rebuild operational stability before deploying AI, or risk automating chaos that damages the referral relationships and pull-through rates that determine long-term survival.

The hidden cost of lean operations

When the mortgage market tightened in 2022, conventional wisdom pushed lenders to cut operations staff and outsource processing. While this reduced overhead temporarily, St. Germain argues it created companies now struggling to deliver consistent client experiences through systems never designed for pressure.

The lenders best positioned for AI adoption maintained operational stability throughout the downturn. They protected the processes and people connected to client experience and referral relationships—two assets nearly impossible to rebuild once damaged.

Pull-through rate emerges as the critical metric. When operations fragment, pull-through suffers, directly impacting what referral partners care about most, even when they don't articulate it explicitly.

Where AI actually delivers value today

St. Germain reports genuine operational advantages from AI in specific back-office applications. His company's COO and CFO use AI to transform reporting and business tracking, compressing work that previously required days into a fraction of the time.

However, AI underwriting remains incomplete. While accuracy rates look impressive, the technology still requires human review of every decision. "You haven't removed the human from the process," St. Germain notes. "You've added a step."

The technology cannot yet replicate the value of a five-minute borrower phone call that reveals information no intake form captures—details that can completely restructure a loan's product or timeline.

Strategic advice for smaller lenders

Large lenders hold structural advantages in AI implementation that smaller companies cannot match. Rather than competing on automation, smaller firms should focus on capabilities AI genuinely cannot replicate: local market knowledge, long-term relationships, availability when deals face trouble, and judgment to deliver difficult but necessary advice.

St. Germain recommends automating tedious, error-prone back-office work while keeping client-facing interactions human. He also cautions against long-term vendor contracts in a space moving so rapidly that today's solutions may become obsolete before implementation completes.

The relationship business reality

The mortgage industry remains fundamentally a relationship business that uses technology, not a technology business that happens to process mortgages. Loan officers thriving five years from now will be those who understood what AI does well, continued doing what AI cannot replicate, and built businesses resilient enough to absorb technology shifts without losing core competitive advantages.

These details were first reported by HousingWire in an opinion column by St. Germain.