We use essential cookies to make our site work. With your consent, we may also use non-essential cookies to improve user experience and analyze website traffic. By clicking “Accept,” you agree to our website's cookie use as described in our Cookie Policy. You can change your cookie settings at any time by clicking “Preferences.”
    Ownify
    Sign inGet started

    By Frank Rohde • June 3, 2026

    Where AI Actually Wins in Mortgage Lending: The Trust Opportunity Hidden in the Numbers

    80% of millennials don't trust their loan officer. Here's why agentic AI mortgage lending finally breaks the commission-versus-advice conflict for IMBs.

    A commission-based mortgage loan officer compared to free AI based advice

    For the last couple of years at Ownify, we've been building agentic AI for mortgage lending. Today we run ten+ subagents handling work we would have hired humans for three years ago: marketing, borrower education, customer qualification, onboarding customers, underwriting and pricing homes, managing transactions, property management. Building for ourselves taught us what most of the industry is still catching up to.

    Software has always replaced the rote stuff. Document checks, data verification, ordering third-party reports, flagging discrepancies, routing files, preparing disclosures. The work of loan processors, underwriting assistants, and clerks has been progressively absorbed by code for decades, and the big LOS platforms (Encompass, nCino, Empower) were built to industrialize exactly that.

    What is agentic AI in mortgage lending?

    Agentic AI is the next layer up. Instead of a rules engine running a fixed sequence of steps, an AI copilot ingests customer questions and data, reasons about what's missing, runs the right verification, and reaches back to the borrower in plain language. Agentic AI vs RPA in mortgage operations is not a close comparison: RPA breaks when a PDF rotates ninety degrees, while an agent re-reads it and keeps moving.

    I believe the real leverage for AI is in work that requires intelligence /expertise but has historically been driven by sales people with misaligned incentives. Let me explain what I mean.

    Understanding a consumer’s financial picture and motivations. Picking the right loan product. Evaluating different financing pathways. Building the pricing strategy on a home. Writing a winning offer. These are tasks that a) require real expertise and the ability to reason, and b) historically have been conducted by high-earning, commissioned sales people. The customer is looking for an advisor to answer questions that go beyond their expertise but encounter sales people whose motivations introduce a natural bias (to close the deal). It’s this combination of technical expertise and sales bias that creates a unique opportunity for AI: provide genuinely good advice at low cost.

    The trust-versus-commission paradox

    The opening here is trust.

    The mortgage industry is engineered around a sales funnel, and the loan officer at the top of that funnel is commission-based. They are not independent advisors, however much they position themselves that way. They are salespeople, partly by design and partly because MLO compensation rules require it. When average compensation is three to five thousand dollars per closed loan, advisory work is a loss leader. A consumer asking their loan officer what's truly best for them is asking someone whose paycheck depends on the answer being "buy a loan from me."

    Younger buyers can feel it, and they will say so out loud when asked. Guild Mortgage and YouGov's 2026 Gen Z survey is a useful x-ray of why the next generation distrusts loan officers:

    • Loan officers trail real estate agents by a full twenty points on trust. Agents show up on YouTube, Instagram, and TikTok months before a buyer is ready; LOs only appear at qualification, in the part of the journey that feels most opaque.
    • Twenty-two percent cite hidden fees as a top concern, despite federally mandated Loan Estimate and Closing Disclosure forms. The disclosures exist. The understanding does not. This is a communication gap, not a compliance gap.
    • Fifty-three percent want explicit help understanding closing costs, not just a stack of documents in plain language.
    • In open-ended responses, buyers repeatedly said they want to know how their lender is compensated. That is the borrower naming the conflict out loud.
    • Trust must be earned early through education, not persuasion. Gen Z grew up with economic instability, student debt, and an unrelenting feed of stories about hidden fees and financial scams. Skepticism is self-protection.

    Eighty percent of millennial buyers do not trust their loan officer. Trust in the system is broken, and whoever rebuilds it wins.

    Now, AI has its own trust issue - a 2026 Cotality survey shows 68 percent of buyers would manually verify a significant amount of anything AI tells them, and 44 percent would pay extra for a human expert to verify an AI housing decision. There is a trust cliff: 70 percent lose trust after a significant AI error, versus 60 percent for the same human mistake. However, AI has two massive advantages here: 1) it is getting better fast. Hallucination and inference errors are decreasing by the week, and 2) it is getting cheaper fast. When we first built Owen, our mortgage AI agent, the cost per customer evaluation was $5-8, now it is cents.

    It is the cost advantage that makes AI structurally different. An AI mortgage agent does not need to earn three to five thousand dollars from you. Inference costs cents per conversation. That rewrites the compensation math and the incentive & cost to give honest advice. It is ok to give 100 customer objectively good advice and convert 1 of them into a customer if the cost to deliver advice converges to zero.

    In the current model, spending an hour each with 100 customers doesn’t make economic sense if I can only convert 1 of them.

    An AI loan officer can objectively research every loan type, every down payment assistance program, and every alternative financing path, then curate a plan for one buyer. Even when the right answer is "wait six months and pay down this card first." Strip out the commission and the conflict goes with it.

    How are IMBs using AI in 2026?

    This crystallized for me at the Housing Innovation Conference in Dallas, with Jeremy Potter, Kenon Chen, and other thought leaders on the same question: how do you reduce the cost to a first-time homebuyer? Home prices are not coming down. Rates have reverted to their forty-year average of six to seven percent and are not coming down either. The only lever left is transaction cost.

    Seven to eight percent of a purchase price goes to listing agents, buying agents, closing, and financing costs. Originating a single loan costs about thirteen thousand dollars. That is where AI for independent mortgage banks should be pointed: not at making participants more profitable, but at taking cost out for the consumer. Most of the current effort of deploying AI is focused on back-office efficiency - Stratmor has 38 percent of lenders using AI or ML, up from 15 percent, but only 7 percent using generative AI. But the opportunity on the customer engagement side are massive. Early agentic deployments report 50 percent origination volume lifts, 2.5x faster closes, and one to three hours saved in loan processing.

    At Ownify, we’re using generative AI (Owen) to build personalized homeownership plans for customers, regardless of their current revenue potential. These are well-reasoned, 8-12 page individualized financing paths over a 2-6 month horizon. They are curated based on the individual’s financial situation and capacity, and the specific property or market the customer is interested in. Owen takes about 2 minutes to evaluate thousands of loan options, down payment assistance programs, as well as alternative financing paths (shared equity, loan assumption, ground leases, seller financing, rent-to-own). Customer feedback has been great as aspiring homebuyers realize they’re getting unbiased advice rather than a sales pitch. Thus far, the engagement rate on the personalized homeownership plans has been 52%.

    Trust requires human-in-the-loop, not human-out-of-the-loop

    None of this rejects the role of a human loan officer or real estate agent. I believe the right combination here is leveraging the unbiased and cheap reasoning capacity of AI with the human-in-the-loop underwriting with audit trails, explanations, and a real person reachable when something looks off.

    The Nomis lesson, applied to a bigger problem

    When I ran Nomis Solutions, our frontline pricing engine outperformed 70 to 75 percent of loan officers. The big ROI driver was pulling the median up to roughly the top quartile. That was deterministic software replacing judgment.

    The AI wave replaces reasoning, in the part of the funnel where trust is lowest and the opportunity to rebuild it is largest. That is what we are building Owen for: an AI mortgage copilot that allows lenders to become trusted advisors at scale. An advisor to aspiring homebuyers whose only incentive is to be right.

    Because at the end of the day, people buy from people they trust.