Laying the Foundation for a Seamless Digital Mortgage Experience The STRATMOR Group’s April 2025 report, Laying the Foundation for a Seamless Digital Mortgage Experience, underscores that while the mortgage industry has made progress in digital adoption, many lenders remain at a crossroads. The study reveals that most efforts have focused on front-end digital capabilities—such as online disclosures, borrower document uploads, and dynamic applications—while significant inefficiencies persist in core back-office processes. For a truly seamless digital mortgage experience, lenders must build on these initial investments with advanced automation and AI-driven strategies. MOZAIQ couldn’t agree more. Forget RPA The report highlights the industry’s widespread use of Robotic Process Automation (RPA), with adoption rising to 48% in 2024. However, it also exposes RPA’s limitations. While RPA can automate repetitive tasks like ordering appraisals and credit scores, it is “brittle,” requiring constant maintenance and offering little scalability. STRATMOR cautions that RPA should not form the backbone of a digital strategy—modern mortgage operations demand intelligent automation powered by AI and machine learning to handle complex workflows and adaptable decision-making. MOZAIQ’s experience confirms this. We recently helped a customer disengage from their RPA vendor, saving them over $300,000 in annual software fees—fees they were unknowingly paying for minimal returns due to the lack of transparency and gross inefficiencies inherent in many RPA solutions. Digital Strategy Recommendations STRATMOR offers several recommendations for lenders as they shape their digital strategies, and while valid, MOZAIQ has a slightly different perspective: Foundational Process Automation STRATMOR recommends embracing document AI (document indexing and data extraction, known as foundational processes), assessing AI options (whether it’s from the LOS vendor or a third-party), and investing in training (but only if IT is the lender’s core competency, like it is for UWM and Rocket). MOZAIQ agrees with these principles but stresses that AI and intelligent automation must be delivered through integrated, API-based solutions that work within the LOS environment, and managed by expert partners who handle the ongoing optimization and training of the platform. Cost Should not be a Barrier to Intelligent Automation On barriers to AI and intelligent automation adoption, STRATMOR cites cost, internal IT capabilities, and other adoption challenges. MOZAIQ disagrees that cost remains a barrier—modern intelligent automation solutions eliminate upfront software licensing costs. Solutions should be delivered as a managed service, priced on a per-loan basis, and configured with low-code platforms to quickly go live. If IT is not your Core Competency, Find a Trusted Partner Similarly, internal IT capability should not be a hurdle—technology changes too rapidly for most lenders to keep pace, and it’s expensive to hire and retain great technical resources. We believe that outsourcing to a specialist partner is the more prudent, and strategic, approach. Choose MOZAIQ to Achieve a Seamless Digital Mortgage Experience For senior executives ready to move beyond outdated RPA and fully leverage AI in mortgage lending, MOZAIQ is the partner of choice. As the leader in intelligent mortgage automation, MOZAIQ delivers AI-powered solutions that integrate with your LOS, replace fragile BOTS, and enable scalability without IT headaches. By automating loan setup, underwriting, closing, and post-closing with advanced AI solutions, MOZAIQ helps lenders cut costs, accelerate cycle times, and elevate the borrower experience. For lenders serious about achieving a seamless digital mortgage experience, MOZAIQ is the partner to turn vision into execution. If you’re ready to achieve real business benefits with a winning automation strategy, follow the lead of Mortgage Lenders who choose MOZAIQ. Contact us today and discover how our Integrated-AI, End-to-End, Intelligent Mortgage Automation solutions can help you win.
The Future of Mortgage Lending is AI
The Future of Mortgage Lending Talk to any C-Suite mortgage executive, and inevitably the conversation will veer toward Artificial Intelligence—AI—and how (not if) AI creates competitive advantage, and how (not if) AI enables intelligent automation (IA) to rapidly deliver business benefits. Why is AI at the forefront of these discussions? Recent research from National Mortgage News (SOURCE: National Mortgage News, Intelligent Automation: The Fully Automated Enterprise Study 2025) provides insights as to why lenders must adopt AI-powered Intelligent Automation. Below is a concise summary of the research. If you want to harness the power of AI, read on. How AI and Intelligent Automation Are Driving Change Mortgage lenders are rapidly adopting AI-powered intelligent automation (IA) to meet critical business needs and stay competitive in a slow and tight market. According to the National Mortgage News (NMN) study, the top business objectives driving IA initiatives are the same as they were two years ago: The study further highlights the top use cases where intelligent automation in mortgages has the highest business impact. These include: Automating these high-value workflows not only speeds up the loan origination process but also improves accuracy, minimizes risk, and ensures compliance—all of which directly translate to better borrower experiences and higher margins. However, with the deployment of generative AI (GenAI) and large language models (LLMs) in intelligent mortgage automation solutions, lenders have concerns, including underwriting and decision-making accuracy, regulatory compliance and legal risk identification, customer service interactions, and data privacy and security. These worries reflect the industry’s need for responsible AI adoption to ensure consistent, transparent, and compliant outcomes. You can read a blog I wrote a year and a half ago that discusses how lenders can embrace responsible AI. Looking ahead, loan processing and underwriting are expected to undergo the most dramatic transformation through the adoption of AI over the next few years. The study predicts that these two use cases will deliver the greatest impact for lenders by driving faster approvals, enabling higher loan throughput—imagine underwriters processing ten or more loans in a day, instead of the meager two or three they do today—reducing costs, and improving decision accuracy. Read here how MOZAIQ can help you implement intelligent automation, powered by AI, in the File Review / Loan Setup process. Despite these benefits, barriers to AI and IA adoption remain. The study identifies the top obstacles as data security and privacy concerns, difficulty finding the right technology partners, IT integration issues with legacy systems, the time required to implement automation solutions, and a shortage of in-house skills to manage intelligent automation solutions powered by AI. Overcoming these challenges will require strong vendor and solution partnerships, workforce training, and a clear digital transformation roadmap. Lenders who address these hurdles head-on will be best positioned to lead the next wave of innovation in mortgage technology. How MOZAIQ Helps Lenders Achieve Their Automation Goals As the proven leader in delivering end-to-end intelligent mortgage automation solutions, MOZAIQ empowers lenders to achieve their critical business objectives by delivering pre-configured AI-powered intelligent automation solutions. Because MOZAIQ delivers end-to-end automation and not just a point solution, our partnership accelerates cycle times, minimizes errors, and ensures regulatory adherence, enabling lenders to scale effortlessly even in volatile, and slow, markets. And, with a per-loan pricing model, MOZAIQ removes the barriers to automation adoption, transforming mortgage lenders operations and helping them gain a sustainable competitive advantage. If you’re ready to achieve real business benefits with a winning automation strategy, follow the lead of Mortgage Lenders who choose MOZAIQ. Contact us today and discover how our Integrated-AI, End-to-End, Intelligent Mortgage Automation solutions can help you win.
Cut Initial UW Time
Cut Your Initial UW Time Automating Loan Submission through to Initial Underwrite Abstract The mortgage industry has long struggled with inefficiencies in the Loan Submission and Initial Underwriting process, leading to delays that extend processing times, increase costs, reduce productivity, and negatively impact customer service. Traditional responses by lenders—such as adding human resources to manage fluctuating volumes, improve quality, and shorten processing timelines—have proven inadequate and unsustainable. Recognizing these challenges, forward-thinking lenders are investing in intelligent automation to streamline workflows, reduce reliance on manual labor, and build resilient operational frameworks that perform consistently across economic cycles. Cut Your Initial Underwrite Time from 48 Hours to Under 1 Hour In our latest blog post, discover how MOZAIQ’s Integrated-AI Mortgage Automation platform addresses these challenges—enhancing accuracy, transparency, and speed in the loan origination process, all without burdening lenders with unnecessary costs. By automating document indexing, data extraction, audit, and real-time communication between stakeholders via optimized Gen AI-powered solutions, MOZAIQ significantly reduces the time to an initial underwriting decision—potentially from 48 hours to under one hour. The MOZAIQ platform’s modular features empowers loan officers with immediate feedback, supports underwriters by minimizing repeated touches, and provides borrowers with a smoother, faster experience, paving the way toward full-scale digital mortgage transformation. Continue to read the full blog below, and when you’re done, contact us to learn how MOZAIQ can streamline your mortgage fulfillment operations with Integrated-AI solutions. Blog Post: Cut Your Initial UW Time from 48 Hours to Under 1 Hour When a Loan Officer submits a loan package for an Initial Underwriting decision to a Lender (either wholesale or retail), they want that decision to be made as fast as possible, and they want their customer, the borrower, to be happy with the service. Not only does that benefit the loan officer, allowing her to increment her submitted loan count, but also the borrower, who benefits from a faster decision to close on a loan. And what about the lender? It benefits them as well to move the loan through the system as quickly as possible, enabling them to pocket the fees faster, and to minimize the use of their warehouse line. Fast is better. But there’s always a problem. The loan file is seldom complete, or it’s not accurate, the loan officer not providing all the necessary documents for the borrower type (e.g., self-employed vs. W-2, or multiple bank statements), or simply omitting one or more documents. And the loan officer doesn’t know this until the lender reaches out, sometimes hours, even days, after the initial loan submission. Not good for the borrower, the loan officer, nor the lender. And, for the lender, not receiving a complete file tends to force the underwriter, the most expensive resource on the payroll, to touch the loan file multiple times, just to issue an initial approval. Longer lead times, miscommunication, expensive resources, lower throughput. That’s inefficient. That’s cost-prohibitive. But unfortunately this is how the mortgage industry has been operating for decades. Sure, some attempts have been made to automate, introduce technology in search of efficiencies. But there never seems to be enough time, the cyclicality of the mortgage market forcing lenders to put on the brakes in down times, and to forget about long-term solutions when the times are good. Just hire more people. Pay them above-market rates, because they can’t train new ones fast enough, not in a rocketing market. And then the bottom falls out and everyone’s fired, and the lenders have negative loan origination margins, and some even shut down lines of business. Or even their doors. Automation, with Integrated-AI, are Critical Success Factors Meanwhile, the smart lenders are not waiting; they are investing in automation now. They’ve learned from their past mistakes—relying on human capital to scale, for one—and so they’re willing to invest in technology and reap the benefits later, when the market will inevitably turn for the better—they’re creating a platform and a business structure that can weather the market shocks, up and down. That’s where MOZAIQ can help. Let’s take the process from the Loan Submission (also called Loan Setup or File Review) through the Initial Underwriting Decision to illustrate the benefits of MOZAIQ’s automation solutions. Loan Submission is the entry point for a loan file into the lender’s system of record. A loan officer compiles the preliminary set of documents from the borrower, validates, to the best of their knowledge, that all the documents are included in the file and that they are accurate, and submits the loan file to the lender through a front-office point of sale (POS) system. The reality is that the submitted loan package is typically in various stages of completeness, and is seldom ready for an immediate Initial Underwriting decision. Instead, the lender, to ensure the underwriter’s time is not wasted, usually has multiple file review teams—comprised of junior resources—clean up the package, perform data-entry, essentially pre-processing the file before it gets to the underwriter. This inevitably adds time, increases loan origination cost, and slows down the entire origination process. And, once the documents have been submitted to the lender, the loan officer now is blind, with little visibility into how the loan is progressing through the lender’s system. And if the loan officer is impatient, which they usually are, they’ll pick up the phone and call the lender directly, sometimes calling an underwriter directly, someone who they have a relationship with, asking them about the status of the loan that they had just submitted and by the way can you expedite it through the process? In general, the entire Loan Submission through Initial Underwrite process for a medium-sized lender takes up to forty-eight hours. Why so long? The initial approval typically results in multiple conditions being generated. And these conditions, instead of being cleared in one session, are piece-mealed to the underwriter, causing them to review a single loan file up to four times before final approval/clear to close. But what if there was a way to automate critical portions of the process and the communication channels creating efficiencies for all parties involved, compressing timeframes? Talk to the Loan File MOZAIQ has expanded its market-leading
3 Top Ten Mortgage Lenders
3 of the Top 10 US Wholesale Lenders Choose MOZAIQ The 2024 Inside Mortgage Finance (IMF) rankings are out, and MOZAIQ couldn’t be prouder to see three of its customers in the rankings of the Top 10 US Wholesale Lenders—each with stratospheric QoQ growth rates. What an achievement! I profiled one of them last year, highlighting their exceptional 2023 growth rate. In the same article, I posed this question: Could they have achieved those growth rates without investing in automation from the start? The answer is a resounding “no.” Why? Because with intelligent automation, they can: Our pipeline has never looked better. If you’re ready to achieve real business benefits with a winning automation strategy, follow the lead of Mortgage Lenders who choose MOZAIQ. Contact us today and discover how our Integrated-AI, End-to-End, Intelligent Mortgage Automation solutions can help you win. Note: This blog post was written by a human and does not contain content generated by ChatGPT or any other Generative-AI platform.
Embracing AI in Mortgage Lending
Embracing AI in Mortgage Lending: Overcoming Fear and Driving Efficiency If you attended two of the most recent MBA conferences—the Independent Mortgage Bankers Conference and the Servicing Solutions Conference & Expo—you likely noticed a common theme: AI is here, and lenders must embrace it or risk being left behind. The sentiment of these discussions aligns perfectly with what MOZAIQ has been advocating since 2022: when it comes to AI adoption, “Don’t get left behind.” Yet, despite widespread acknowledgment of AI’s potential, few mortgage lenders have taken the leap to implement AI-enabled solutions. Why? The answer lies partially in fear—fear of adoption, fear of change management, and fear of the cultural shift that automation demands—and just plain risk management: many lenders have already rightsized for the current market, and because of the persistent anemic growth, don’t have a lot of spare capacity to invest in new initiatives. Change Management in Automation The mortgage industry remains heavily reliant on manual labor and paper documents—even in 2025. While automation can significantly improve efficiency, the thought of replacing established workflows with AI-driven solutions can be intimidating for lenders. The reason? Especially in uncertain times, people don’t like change, and lenders have been processing loans the same way for decades, with little innovation, compared to other financial services verticals, like capital markets. And because any type of AI-powered solution or automation implies change, it’s important to first communicate the “why” to teams. From a lender’s perspective, because a mortgage is a commodity product, lenders only have two levers they can play with: price, and customer service. If a process can be fully or even partially automated, freeing up processors from executing mundane, repetitive tasks, those same employees can be redeployed to higher-value work, such as enhancing customer service—for brokers, homeowners, investors—and participating in functions like file prep, pre-underwriting, and several aspects of post-closing (e.g., funding) that still require intelligent human intervention. Not to mention that automation will lower the total cost of processing a loan, giving the lender wins with both the price and service levers. Market Conditions: A Barrier or an Opportunity? Some lenders are also hesitating to invest in AI-powered automation (and other technology initiatives) due to unfavorable market conditions and the pervasive uncertainty brought on by the new administration in Washington. And at first glance, who can argue with them? Even with these market conditions, lenders must ask themselves: If the landscape isn’t improving, why not take steps now to increase operational efficiency and prepare for the future? The Smart Lenders Are Already Acting We’re working with a client that has deployed several integrated AI automation solutions, such as Initial Underwriting Submission and Appraisal Review, where configuration costs are minimal, and they only pay on a per-transaction basis. This approach allows them to “test the waters” of automation, optimize their processes to balance automation with human oversight, and position themselves for the inevitable scale they’ll need once the market improves. Essentially, they are willing to invest in solutions that allow them to minimize the risk of having to suddenly scale and grow with people, and are willing to pay for that technology now e.g., invest in a solution that only charges for funded loans, on a per loan basis—transactional and funded volume only, thereby using technology as a hedge against the market risk/uncertainty. And why not do it when volumes are low—when there’s less stress and more opportunity to refine and implement AI solutions correctly the first time? The Data Doesn’t Lie When automation increases throughput from 8 files per day to 16 for a critical underwriting process such as appraisal review, achieving a 100% gain in process efficiency, why wouldn’t lenders seize the opportunity? When an Initial Underwriting Decision timeline is reduced from 48 hours to just 4 hours—making Loan Officers, Loan Processors, and Loan Underwriters significantly more efficient—isn’t it worth taking the time to explore AI’s potential now? Reduced processing costs per loan and improved operational efficiency should be compelling enough. The question remains: Why wait? The Time to Act is Now Culture change remains the biggest barrier to AI adoption in mortgage lending. However, waiting for the market to improve before implementing efficiency-driven technology is a risky strategy. AI isn’t here to displace professionals—it’s here to empower them. Mortgage lenders who embrace AI today will not only navigate the current downturn more effectively but will also be better positioned for growth when the market rebounds. The question isn’t whether AI should be adopted—it’s whether lenders can afford to wait any longer. If you’re interested in realizing real business benefits via a winning automation strategy, contact MOZAIQ today and find out how our Integrated-AI, End-to-End, Intelligent Mortgage Automation solutions can help you win.
Solution Spotlight: Post-Close Automation
Solution Spotlight: Post-Close Automation Summary: Highlights the benefits of automating multiple processes in Post-Closing showcasing the value of an end-to-end mortgage automation solution. Last week the Mortgage Bankers Association published an article with an ominous headline: ACES [ACES Quality Management] Finds Second Consecutive Increase in Critical Defect Rate. The ACES analysis found that in Q2 of 2024, the critical defect rate—a fault in a loan file that would make it uninsurable or ineligible for sale—had risen to 1.81%, nearing the “precarious” 2% threshold. Why does this matter? As I wrote about in a blog several months ago, poor loan quality can have severe adverse impacts on a lender, including, but not limited to, being forced to buy back the loan at significant costs to the lender—sometimes incurring a penalty with a cost of up to 30% on the loan amount, as well as other penalties—financial and reputational. All bad news for a lender in today’s tight and competitive market. But it’s not just about catching and fixing the critical defects before the loan is shipped off. There’s also the issue of how long a lender has a loan on their books, and how quickly they can clear the loan off their warehouse line. Wouldn’t this be a great benefit as loan volumes increase? We think so. And so do our customers. MOZAIQ is the only software partner today that delivers integrated-AI automation solutions across the end-to-end mortgage fulfillment lifecycle. Why end-to-end? So customers don’t get stuck with costly and inflexible point solutions, and end up managing multiple, and incompatible, automation solution providers, while trying to compete effectively. Post-Close Automation The Post-Close Automation suite of products include: 1. Trailing Documents Reconciliation The automation solution confirms whether the Final Title policy, the Deed of Trust (Recorded Mortgage), and the Mortgage Note have been received by the lender on the funded loans. It involves checking the loan origination system (LOS), email inboxes, and third-party websites, such as DocProbe, for the availability of documents, and sending automated emails to the Title and Settlement agents as required, requesting the missing documents. 2. Investor Stipulation (“Stips”) Retrieval Investor Stipulation Retrieval is a critical process designed to address conditions / stipulations raised by investors to ensure smooth and timely loan sales. The process automates: Investor Conditions Review: Investors have specific conditions, or stipulations, that must be met before a loan can be sold to them. Loan Origination System (LOS) Update: Once the stipulations are identified, the relevant information is entered into the lender’s (seller’s) Loan Origination System (LOS). Compliance and Efficiency requirements are met: By addressing stipulations, the process upholds compliance with investor conditions, enhances the efficiency, and minimizes delays to ensure that loans meet all necessary conditions for a successful sale. 3. Post-Close Audit The Post-Close Audit automation solution ensures compliance and accuracy with regulatory and investor guidelines. It runs on the industry-leading Checkpoint Audit Platform enabling the underwriters to confirm that the loan meets the necessary underwriting parameters by identifying and correcting any discrepancies or errors. The platform enables the post-close auditor (human processor) to easily review and validate the loan, and fix the errors where necessary as highlighted by the automation platform, for example, for missing and expired documents, missing data, mismatched data and documents, and checks for erroneous calculations. Once the loan passes the audit, the auditor approves the loan, the results are uploaded into the LOS, and the loan is committed to an investor, ready to be delivered. A detailed write-up of the Post-Close Audit process can be found here. 4. Final CD The MOZAIQ automated Final CD solution facilitates the error-free process of generating and sending the final Closing Disclosure (CD) to all borrowers within 3 days of their loan funding, and updating the disclosures, if necessary, in the LOS. The process runs 24×7 and delivers a higher throughput of loans, enabling the lender to effortlessly scale in line with volumes, and requires less human intervention, ensuring greater accuracy and transparency for all parties. 5. Post-Close CD The automation solution records and sends the Post-Close CD to the relevant parties, with any required adjustments made to the CD, after the loan has closed. The revised document includes changes such as prorations for taxes or fees (e.g., appraisal) that were not finalized at the time of the Final CD. Any adjustments are made in the form of refunds, or cures, and sent to both the borrower and the seller to ensure accurate records are shared with both the parties. The automation ensures that the process and the loan adhere to the required compliance and regulatory guidelines, helping to maintain compliance and transparency for all the parties involved. 6. Investor Loan Delivery The Loan Delivery automation solution runs 24×7 ensuring an accurate, timely, and zero failed deliveries of the loan to the investor. It enables the mortgage lender to automatically deliver loans to GSEs—Fannie Mae, Freddie Mac, and non-GSEs—via aggregators configured in the LOS, for example, Encompass’s Investor Connect. Once a funded loan is committed for purchase to an investor, the automation updates fields in the LOS, generates the XML loan file, clears data entry validations, and submits the loan to the investor. It also generates LCA/UCD reports, post-closing forms, indexes the delivery documents, updates the ULDD details, and generates the PDF investor package via an integrated Package Delivery solution where for loans that are funded with Investors, Servicing Packages are automatically printed from the LOS, and the package is then delivered to the Servicer through the appropriate delivery methods, including Secure File Transfer (SFTP): Freddie Mac—FAST, Fannie Mae—Easy Transfer, and Encompass’s Investor Connect. One More Thing—Loan File Setup The same end-to-end platform is used to execute the Loan Setup process to prep the loan file and ensure it is as clean and complete as possible before it reaches the underwriter. The automated Loan Setup process automatically extracts data from the loan docs (we’ve trained our platform to extract data from over 200 mortgage document types), inputs the data into the LOS, triggers the required functions in the LOS, such as third-party service orders, intent to proceed, lock confirmation etc., ensures
The Loan Store Wins
55% QoQ Growth in a Down Market Since joining MOZAIQ in early 2022, I have written multiple blog posts discussing the benefits of automation and have showcased our own customers as successful adopters of automation. I first mentioned The Loan Store in a blog in October 2022, almost two years ago. I wrote about how they built their business with the objectives of always delivering exceptional customer service—to brokers and to customers—and that to win in a commoditized market (newsflash: mortgages are commodity products) they needed to streamline efficiencies so they could also compete on price. To achieve these goals, they embraced automation and outsourced some of their back-office services from day one. I then wrote about them here, in April 2023, when the MBA released their annual report on the state of the market in 2022 (lenders on average had negative loan origination margins). I explained how The Loan Store was able to absorb the assets of another mortgage lender and not break stride in their delivery of high-quality mortgage services, keeping the goals of customer service and competitive pricing intact. I mentioned them in a talk I gave at the World AI Cannes Festival in February 2024, as an example of a lender that was primed and positioned to take advantage of lower interest rates and higher loan volumes when the day arrived. That day is today. I’ll let the numbers do the talking. In the first six months of 2024, The Loan Store funded $1.9 BILLION worth of wholesale loans (Source: Inside Mortgage Finance estimates). In the first half of 2022, they had funded approximately $120 MILLION worth of loans. You do the math on their growth rate. They’re not stopping. Yes, they have a talented CEO with a talented team, and they picked the right strategy. But the key is execution. The key is efficiency. The key is scale. The key is to remain price competitive. The key is to deliver exceptional customer service to brokers and consumers. They’ve done all that. Could they have done all that without investing in automation since the beginning? You can contemplate all you want. We know that without automation they would not have been able to take advantage of the growth in demand (spearheaded by an exceptional sales strategy), they would not have been able to scale on demand, they would not have been able to maintain and surpass quality and accuracy thresholds, they would not have been able to absorb volume growth without an increase in costs, and they would not have been able to keep their costs in line with their needs to remain price competitive. If you read the blogs I mention in this post, you’ll know that we told you that the time to automate was when the mortgage market was down and in a lull. We also said, “don’t get left behind.” If you’re interested in implementing a winning automation strategy contact MOZAIQ today and find out how AI-Integrated Intelligent Mortgage Automation can help you win. Note: This blog post was written by a human and does not contain content generated by ChatGPT or any other Generative-AI platform.
Solution Spotlight – Intelligent Automation Helps Lenders Scale
How Intelligent Automation Can Help Lenders Scale The Mortgage Bankers Association’s (MBA) Annual Mortgage Bankers Performance Report for 2022 was released at the beginning of April, and the data, although not a surprise, was disconcerting. Interpreting the data As loan volumes precipitated, lenders could not adjust their capacity fast enough—the number of production employees did not decline at the same pace as origination volume declines, causing production expenses per loan to increase and already tight profit margins to evaporate and turn negative. No lender can afford to make a single mistake in this environment. And mistakes are more prevalent with lenders that have not invested in technology and automation, having to rely on employees for complex and simple tasks, from underwriting a loan to locking a rate, instead of creating a hybrid environment where technology and automation can alleviate the load of the more repetitive, mundane tasks. Add on new GSE (Fannie and Freddie) requirements and a challenging scenario unfolds: all lenders must now audit 10% of files at the Pre-Fund stage, instead of the current 5% of files; and Post Close Quality Control reviews will be required to be complete within 90 days instead of the current 120 days. Lenders have already cut their teams to the bone. How will they comply? The Loan Store The Loan Store (TLS), a national wholesale lender and one of MOZAIQ’s clients, was created with a philosophy comprised of two simple drivers: (1) be the low-cost lender with (2) superior customer service. To support these two drivers, TLS would scale by leveraging outsourcing where it made sense and invest in technology and automation. They knew that because a mortgage loan is a commodity, if they stayed focused on these two drivers, they could sustain growth and profitability, and win. Fast forward to today, and The Loan Store has automated a large portion of their backoffice: Enter Homepoint And then in early April 2023 Homepoint, the third-largest wholesale lender by origination volume in 2022, announced that it was selling its origination channel assets to The Loan Store, which will allow TLS to scale its loan origination business into a leading national wholesale and correspondent mortgage lender. Homepoint is only bringing a skeleton operations crew over as part of the sale. No technology is coming over. So how will TLS support a projected ten-fold increase in origination volume in the first three months post integration? How will they scale in the next 6, 12, 18 months? How will they comply with the new Fannie and Freddie requirements? The answer is automation. Scaling with Automation Because TLS was built from day one to be nimble and to leverage technology to enable growth, the infrastructure to scale is already there. With automation, all TLS has to do, with MOZAIQ’s help, is deploy additional Virtual Machines (VMs) on the cloud, and deploy more BOTS, or digital workers, to support the growth. In parallel, TLS is going to reduce its reliance on RPA BOTS and transition to automation via APIs, in order to reduce the inherent errors that an RPA-only implementation has, and to speed up the processing (in order to function, RPA BOTS rely on the underlying systems and screens to remain immutable, which is rare when using third-party systems like Encompass, Empower, Blue Sage or Fannie and Freddie). No changes to processes, no changes to the pipeline, no changes to the loan flow. Just a simple provisioning of incremental automation resources and the additional volume is absorbed effortlessly and seamlessly. The Outcomes The business benefits gained by automating back-office mortgage fulfillment processes include: TLS has already experienced these benefits over the past two years as it has progressively added automated processes to its loan fulfillment pipeline, and now these benefits will continue to translate directly to higher loan profitability and superior customer service, which is exactly what The Loan Store’s objectives have been from day one. Don’t get left behind—Contact us for a free automation diagnostic and see how automation can work for you and enhance your competitiveness, and your bottom line.
Gen-AI and Mortgage Lending
Gen-AI and Mortgage Lending: All Hype or All Real? This article first appeared in the July 2, 2024 edition of MBA Newslink. Gen-AI and Mortgage Lending… given the daily carnage of Gen-AI failures, one has to wonder. A META Gen-AI chatbot (now decommissioned), trained on scientific research papers, made up purported academic papers and generated content about the. . . history of bears. . . in space. A World Health Organization (WHO) Gen-AI chatbot that made up names and addresses of non-existent clinics in the Bay Area. Google’s Gen-AI engine recommended the use of “non-toxic glue” to solve the problem of cheese not sticking to pizza. The answer for this query appears to be based on a comment from a decade-old joke thread on Reddit. But the LLM was trained using this thread as source data (more on this later), so the answer… stuck. And McDonald’s just shut down its partnership with IBM whose watsonx Gen-AI technology powered McDonald’s drive-through automated order taker. How about some bacon with that ice cream? And your order of 260 chicken McNuggets is coming right up. That’s the fundamental problem with chatbots: they will always hallucinate, sometimes correctly, sometimes incorrectly—they’re engineered to make stuff up. Why? A Statistical Inference Engine By now we all know that the underlying engine of a chatbot is called a Large Language Model (LLM). The challenge is that an LLM is not a database, nor a search engine. It is instead made up of a bazillion (lots of billions) numbers that are crunched to calculate responses. The numbers in the model (think of it as a gigantic spreadsheet) are set when the model is trained, and an LLM requires infinite amounts of data to be somewhat relevant, and somewhat accurate. A chatbot creates word sequences from scratch every time a question is posed, because these bots create the output text by predicting the next word in a sequence: a Gen-AI powered chatbot is a statistical inference engine. It asks itself: what is the statistical likelihood that the next word in the sequence “I walked in the ___” is “park” or “room” or “door”. Or “tar pit”? The word with the best statistical score wins. And the score is determined by the quality of the data used to train the LLM. Biased data delivers biased answers. Toxic data propagates toxic answers. Which is why smart companies implementing Gen-AI chatbots require that an answer be traced back to original source data, and why results must be critically analyzed by humans. So, an AI chatbot based on an underlying LLM hallucinates all the time in order to concatenate words in a logical sequence. Many times it’s right. Sometimes it’s wrong. Increasing the Accuracy LLMs can become more accurate as they are continually trained with more data, and are constrained in scope. How does that work? A friend of mine at a major financial services institution is training an LLM with targeted, proprietary data. This is called Retrieval-Augmented Generation, or RAG. Specifically, the firm used a set of internal documents (“external data” in the context of the LLM) to train the LLM, with (human) subject matter experts querying the LLM and providing feedback with prompts as simple as a thumbs up or thumbs down, by typing in the correct answer (if the chatbot spat out garbage), and by going into the internal documents and modifying and optimizing the text (the source data) i.e., making the answer less open to interpretation and more exact. This is one way to create a chatbot that is “smart” enough in order to achieve a high confidence level in their answers. The external data augments the underlying LLM by being converted and stored in a vector database—even more numbers! Then, based on the query, the chatbot will return the most relevant answer, using this external data as the main source of information to formulate the answer. OK, fine, so now the chatbot is more accurate. But regardless, people will get careless as the accuracy of LLMs improves, and when errors do happen, users are more likely to miss them. And errors will continue to happen. Gen-AI and Mortgage Lending The mortgage lending fulfillment process cannot afford errors. Consumers will be negatively impacted by the inherent bias. Lenders’ already thin margins will shrink, and their reputations will be sullied. What then are safe uses of Gen-AI in mortgage lending? Summarization and Synthesis: for borrower loan files, accounting pronouncements, appraisals, financial reports, financial news impacting policies, processes, and reporting. For example, a common use case is in the call center: other than the obvious call routing and agent prompting (nothing revolutionary here, it’s done today with legacy technologies), call summaries can be auto-generated by the Gen-AI engine (trimming up to 40 seconds per call), with increased accuracy, and the ability to trigger follow up actions. Deep Retrieval: augment IDP solutions with Gen-AI to extract information from unstructured data sources. For example, Google (the same one with the cheese and the glue) has deployed a Document AI solution that has Gen-AI embedded into its platform. The platform enables data extraction from scanned documents and does not require hundreds of documents in order to be trained. With unstructured documents that have multiple variations—like bank statements—with mostly simple key value pairs, the Gen-AI-powered data extraction function requires 95% less training and quality control than what a Machine Learning-based (ML) model would need. Loan Officer and Broker Support: a specialized (RAG-ged) customer service chatbot (notice I didn’t say “underwriter support”; see hallucination notes above). A targeted chatbot allows loan officers and brokers to summarize the salient borrower information (minimizing the task of pre-compiling paperwork) and send the summary to the lender account executive to determine whether a borrower even falls within the qualifying parameters of a loan product. If trained further, the chatbot can also offer suggestions on how to structure a loan that falls within valid qualifying parameters if the borrower does not qualify for a loan product that was initially offered e.g., the chatbot analyzes the borrower’s loan file and sees that she has an
Intelligent Automation is an Ecosystem
What is Intelligent Automation? Last week MOZAIQ had the pleasure of jointly hosting a webinar with our partner Gooi Mortgage. The webinar topic was Intelligent Automation plus Expert Underwriting Resources. We showcased how intelligent automation turbocharges efficiencies and allows expert underwriting resources to be even more productive, processing twice the number of loans in the same amount of time, without sacrificing quality and accuracy. In this blog post I want to highlight one of the key themes of the webinar: what, exactly, is intelligent automation? MOZAIQ wants to set the record straight, as various definitions abound in the ether. Intelligent automation is the configuration, integration, deployment and use of automation and AI technologies to streamline functions and scale processes in support of human workers. We look at intelligent automation as an ecosystem, powered by multiple technologies, that supports expert resources. It doesn’t mean that one can’t deploy a single automation solution to solve a business problem. What’s important is to understand that each automation solution has its own challenges when it must scale across multiple, and complex, use cases. Here’s why. Of these automation solutions, the most basic is Robotic Process Automation (RPA) – it’s been around since green screens. I was using it in 1989. It’s a commodity, now. And it’s unstable, expensive (have you checked the licensing agreements from the major RPA vendors lately?), and has a low return on investment (ROI). Then there is intelligent document processing (IDP). It’s OCR technology configured for specialized uses, in mortgage lending it is typically configured to extract data from structured and unstructured content in documents. The problem with IDP is that the accuracy is finite. This is where Machine Learning (ML) comes in: Artificial Intelligence (AI) is leveraged to optimize the output from RPA and IDP processes. ML models must be trained, and require hundreds of documents, reams of data, and time; and one will still rarely achieve above 85% accuracy levels, and then only if pre- and post-processing is applied to the documents and extracted data, respectively. Finally, there’s Generative AI (Gen-AI), an emerging technology that’s prone to hallucinations, which should be avoided when processing a mortgage loan and making lending decisions. It is, however, being deployed in a limited number of mortgage use cases, for example, enhancing data extraction from documents that have high variability or where there just aren’t enough documents available to properly train the ML model. The following chart illustrates the business effectiveness vs. complexity and cost of implementation for each solution. MOZAIQ’s POV on Intelligent Automation What MOZAIQ believes is that true intelligent automation comes from combining all of these technologies and deploying them to solve for targeted use cases in support of expert human resources. In other words, solve for a legitimate business problem, and enhance the effectiveness of a lender’s most expensive resources. As an example, MOZAIQ Checkpoint Automation Platform delivers intelligent automation services across the end-to-end mortgage fulfillment lifecycle. These services include Initial Underwrite, Pre-Underwriting Audit, Appraisal Review, and Post-Close Audit to name a few. Because the fundamental framework and building blocks are present in the platform, and because Checkpoint enables best-of-breed automation components—RPA, IDP, ML, Gen-AI—to be “plugged” into the platform, our customers don’t have to worry about software licensing, configuration, installation, running or monitoring the platform. MOZAIQ does it all for them by providing Mortgage Automation as a Service. Contact MOZAIQ to find out how intelligent mortgage automation can save lenders time (by increasing processing efficiencies by 50%), and money (reducing loan origination costs by up to 50%), making the lender’s expert resources even more productive. Note: This blog post was written by a real human and does not contain content generated by ChatGPT or any other Generative-AI platform.
