AI Implementation: Where Do We Go From Here?

The previous two articles published have attempted to raise awareness of what Artificial Intelligence (AI) is and isn’t as well as the associated risks and opportunities.    This third article describes how the implementation of these technologies can change the way we operate and how, from an operational perspective, using the tools and the benefits can expand business and reduce costs. 

Fluctuating and Growing

One fact that we have to acknowledge is that AI is in flux and will most likely continue that way for some time.  For example, a recent article I read claimed that RPA (robotic process automation) is dead. It doesn’t mean that this type of AI is no longer useful, but that when the individual types of AI such as expert systems, robotic automation for individual processes as well as the disintegration of siloed processes and programs is addressed, it will result in a comprehensive integrated program that utilizes the appropriate method for the task at hand.

Featured Sponsors:


Another area that has come to the forefront of AI is what is known as IoT or the Internet of Things. In this instance, data on the internet can be incorporated into databases as needed, to complete tasks through the use of AI.  A step further in this approach is what is known as AIoT or artificial intelligence data of things.  In other words, if by using the internet to create data, that data then becomes available for others to use in their AI efforts we have effectively created an internet of AI “things”.   

Of course, this is not going to be the final product or program for the use of AI.  As we know from just listening to the news or reading articles on the technology, the future uses of AI are seen as unlimited. 


Today the mortgage industry is trailing behind other industries in the use of AI in their workflow. While there are numerous reasons for this, one of the most basic is that organizations continue to run processes exactly as they were run 40 years ago.  There have been multitudes of new technology products that can be purchased and added to make the process cleaner with fewer errors, but the fundamental changes necessary to realize the intrinsic benefits of many have not been made.  

Featured Sponsors:


Way back in 1979, loan applications were taken by Loan Officers and handed off to Processors, who put together the documentation that supported the application, including required regulatory disclosures, and handed it off to an Underwriter to evaluate for acceptability.  If approved the loan was handed off to a Closer, who made sure the title was acceptable and prepared the documents and funding.  Once the loan was closed, it was sent to Post-Closing, where it was reviewed, corrections made, documents sorted, MI filed, and the loan was then delivered to an investor. The technological support we have today was non-existent and management dealt with only people and processes. 

This linear process is still in place today although we now have multiple systems in these siloed functions to manage the data that was previously on paper. The process also has more documents and more people working on these tasks.  Despite these individual and in some cases co-joined systems, the process is still dependent on having people in place to connect the data and documentation.  Add to this the maintenance of system upkeep, the struggle to ensure the continuity of the data and process as well as the implementation of on-going updated systems and requirements and we get a view of the process  that can best be described as organized chaos and makes it near impossible to streamline the costs or effectively provide “quality” customer service.  

As a result of these Operational Risk issues (people, process & technology), the cost of producing a loan has now surged to around $8800, much of which involves manual reviews and rework.  Why?  A recent survey of the industry found that on average 62.5% or $5500 of the total $8800 is for personnel costs.  Another 11% or $1,000 covers general expenses and approximately the same amount is dedicated to secondary marketing expenses. Eight percent, or $704 is spent to support the technology used by the organization.  The remainder is spent on other various necessary expenditures.  In other words, we have not made any of the transformational transitions that have been promised by technology for years, just added more costs. This lack of vision and the resulting failure to redesign and manage effective change is definitely a roadblock to not only the use of AI, but the ability to run a profitable organization.

Featured Sponsors:


Another roadblock we face is the inability to collect, comprehend and utilize data. Collectively data is the foundation of the products produced and the driver of the processes and people we employ.  The process begins with data collection and proceeds to add additional information throughout.  Yet the systems in use today do not allow us to utilize data from disparate programs to assist in the analytic process of acceptable loans and associated profitability. As I noted in an earlier article, in addition to each individual company’s data, the industry does not share any data.  This results in the inability to utilize comprehensive industry data to develop artificial intelligence.  While Fannie Mae frequently utilizes its data to develop “tools” for lenders, in reality it is only a very small piece of the total data set that could/should be used. Having a tool biased by the use of a segregated population of loans does not provide legitimate results to the total populations.  

Organizational culture can be a roadblock to success.  The culture of an organization actually holds back any company from improving their processes as management and staff adopt a “we have always done it this way” attitude. In fact, a survey of 590 G2000 leaders by HFS Research found that 51% of the highest performing enterprises see their cultures as holding them back in their technological transformation process.  From an Operational Risk perspective, the redesign of processes and corresponding people skills significantly lags technology implementation in organizations.  The only way to address this roadblock is by radically rethinking existing process which will ultimately drive the greatest benefits to the company.  The three pillars of operational performance (people, process, technology) no matter what form the tasks take are fundamental to the ability of the company to produce what has been promised. 

Mortgage Lending Redesign

A true transition of the lending process begins with understanding what AI technology can do. Once this awareness has taken place, a strong change management team needs to be identified and concepts, no matter how “off the wall” they seem, need to be identified.  This envisioning process must include not only how the process will be designed, but the necessary skill sets as well.  

Featured Sponsors:


One of the most frequent questions asked when an AI discussion is held, is “Will I lose my job?” This fear of losing a job is one of the most pervasive concerns by the staff of all industries.  While skill sets will change, most individuals will continue to work in the same profession.  In other words, the tasks will change, but people will work. Recent studies have begun to reorganize common types of tasks performed and reorganized them into potential skill set requirements. 

One example is found in the book by Paul Daugherty and H. James Wilson entitled Human + Machine, Reimagining Work in the Age of AI, where tasks are divided into three groups. Those that involve human-only activity include leadership, creative activities and evaluative or judging skills. Those skills seen as primarily using the AI technology involve transactions, iterations, predictions and adaptations. However, there are a series of hybrid tasks that involve both human and technology efforts.  The human efforts within this hybrid set include communicating with applicants and borrowers to explain and educate them, much as we do now.  There will also be the need for individuals who train the technology.  For example, newer flexible robotic systems that work along with humans need to be trained to handle different tasks, just as we do now when machine learning is required.  There will also be those who sustain the processes and programs by incorporating the company’s risk profile, such as setting limits or allowable override decisions on profitability or legal and ethical compliance.  They do this by ensuring the quality of the data, flag errors and poor machine results, design interfaces for the AI expanded workforce.  

Within each of these groupings are a variety of tasks and corresponding skill sets, but one important factor is that the individuals fulfilling these jobs must understand the when, where and how of every action included in the lending and servicing processes.  Currently, the knowledge and skills necessary to work within the industry are siloed similar to the processes.  For example, underwriters today have a far different knowledge base than those who began working in the industry prior to automated systems.  More experienced underwriters understand why something is required, what the impact is if not and whether or not there are potential ways to address any problems.  That knowledge is quickly being lost.  In addition, because we have bifurcated the origination and servicing processes, few individuals in the production area can explain what and how servicing does and would not be able to explain why a payment was not processed.  For future work within the industry, employees must understand it all. 

Another area that is critical to lenders is regulatory issues.  In November,2018, Lael Brainard of the Board of Governors of the Federal Reserve System presented remarks on What Are We Learning about Artificial Intelligence in Financial Services.  In these remarks, he stated that “AI… is not immune from fair lending and other consumer protection risks…”  and alerted lenders to the challenges in the areas of opacity as well as the ability to explain how the system complies with these requirements.  Since this area is so critical to lenders it is important that there are individuals with the skill sets to provide this information if requested.   

The development and maintenance of credit policy is therefore critical to lenders, and in conjunction with affordable housing initiatives, artificial intelligence, through its data sets and analytics can provide more rationalization for lending parameters.  One chief executive of a financial services company recently stated that what artificial intelligence and machine learning allows is the ability to get much broader perspectives on consumers thanks to additional data, shedding light on their creditworthiness.  

Other processes in the business appear to be ripe for implementation of AI.  Two prominent ones that come to mind are quality control and post-closing.  Using expert systems and RPA, the post-closing review processes can be developed to scrub data, identify missing documents or those that need correction and notify the individual responsible.  Of course, staff will be needed to address those loans that fall outside the parameters of the review program. 

Today quality control consists of reviewing loans to identify errors in all facets of the loan process.  Using artificial intelligence tools such as expert systems, machine learning and electronic verifications ordered and controlled by the technology, QC could be done automatically throughout the origination process n 100% of the loans. Using collected data from other reviews, a more comprehensive analysis of the workflow can be conducted, and opportunities and risks identified.  Rather than the paper intensive manual process that we employ today, executives can have results as frequently as desired while significantly reducing or eliminating many of the costs included in the $8800 that is problematic today.  

The Bottom LineWhether or not the industry surges ahead with the adaption of artificial intelligence, it is coming.  For those lenders who eagerly take on the job of redesigning their people and processes in conjunction with its implementation, the opportunity to increase business while cutting excessive costs are unlimited. 

About The Author

Insellerate Adds AI Expert To Its Board Of Advisors

Insellerate, a mortgage CRM provider, has added Neil Sahota, IBM Master Inventor, United Nations Artificial Intelligence (AI) subject matter expert and noted author to its board of advisors to enhance its AI technology and strategy.

Featured Sponsors:


The integration of AI into the Insellerate platform is helping lenders manage their lead generation, prospecting, engagement, conversion, and the lifetime customer value they provide to their borrowers. Insellerate’s platform reduces the daily inefficiencies that take loan officers away from serving their clients and closing loans and helps lenders consistently connect and engage their borrowers. By optimizing critical workflows, lenders are able to perform their jobs more efficiently and borrowers receive personalized communications for their individual situations. 

Featured Sponsors:


“This technology could save hundreds of mortgage lenders from going out of business due to inefficient workflows and unprofitable, outdated marketing practices that result in high opportunity costs,” said Josh Friend, CEO and Founder of Insellerate. “Leveraging technology isn’t about eliminating loan officers. It’s about supporting them with bias-free, data driven suggestions that can optimize each unique borrower journey, extend customer lifetime value and help lenders operate at their most efficient.” 

Featured Sponsors:


“I’m happy to join Insellerate as a member of their board of advisors and look forward to helping them develop successful strategies that merge their cutting edge technology with the real time business needs of today’s mortgage lenders,” said Mr. Sahota. “Having specific and deep business expertise, knowing where the problems are and where the real opportunities exist is critical to successfully leveraging AI for stronger commercial gains and better customer experiences. And that’s precisely what Insellerate delivers.”   

Featured Sponsors:


“We’re thrilled to have Mr. Sahota advising us on AI technology and strategy,” said Mr. Friend. “It’s our goal to make sure that lenders stay in business in an aggressively changing world. As mortgage professionals we’ve experienced the antiquated technology that prevents lenders from reaching their full potential when it comes to closing loans and managing businesses. AI technology is the next big thing that will impact lenders and that’s why we’re so fortunate to have an expert like Neil joining us on our mission to help bring mortgage lenders into the future — so they’re still in business 5, 10, 20 years from now.” 

Insellerate’s CRM and Engagement Platform are standalone solutions that leverage technology to optimize intelligent engagement that uniquely serves each and every individual borrower throughout the entire borrower journey.

AI Foundry Unveils First Ever Mortgage Document Model

AI Foundry, an artificial intelligence (AI) platform company, has launched its mortgage document model, adding new functionality to its next-generation Cognitive Business Automation Platform. The document model includes an extensive set of standard mortgage document types and common variants. It incorporates the latest in AI, machine learning and machine vision to deliver a higher level of automated classification and data extraction capabilities. This document model capability will enable the mortgage industry to use AI to replace multi-week manual processes, so that mortgages can be processed from “application to underwriting” in days, not weeks.

Featured Sponsors:


For originators and banks who need the ability to automate loan reviews, the document model provides “out of the box” machine-vision-based functionality that classifies and indexes documents, dynamically identifies relevant data content within the documents, detects inconsistencies, applies rules for data validation and ultimately minimizes human interaction with the loan application material. Unlike other OCR and text-based solutions in the market, the document model uses advanced machine vision and deep learning techniques to ensure highly accurate levels of recognition and data extraction required for efficient automation. In addition, the model has standardized the process for supporting new mortgage document types or other documents that require curating and labeling of large numbers of samples and variants. 

Featured Sponsors:


“The model enables any lender to upload its loan application material and in return receive fully indexed and extracted data within seconds. The model delivers 95 percent accuracy and was trained on more than 100,000 mortgage documents, 300 document types and 2,000 data extractions to date, using both cognitive and deep neural network techniques,” said Peter Piela, Ph.D., head of solution development at AI Foundry. “The percentage of accuracy using our vision technology is comparable to human manual processes, while legacy text classification approaches fall well short of this at roughly 80 percent accuracy. The impact of using our document model means significant time savings for the lender and the replacement of expensive manual processes with far more efficient automated ones. 

Featured Sponsors:


“In addition to the document model, the platform contains a powerful rules engine that allows clients to create intelligent robotic agents to automatically monitor completeness, integrity and compliance. The rules engine enables users to make actionable inferences that trigger remedial events early in the document-processing and exception-handling phases, thereby reducing overall cycle time and the cost of remediation,” added Piela.

Featured Sponsors:


AI Foundry’s plan is to make the document model available to customers as part of the Cognitive Business Automation Platform, so they can use the existing model as well as augment the capabilities of the base model to solve specific mortgage workflow processes. The model is continuously enhanced with new variants that are deployed to the SaaS environment, making it available to all customers. The Platform and document model can be deployed to eliminate a large number of manual activities, automating document-centric, labor-intensive processes with a high degree of accuracy, freeing employees from repetitive work processes and refocusing them to more value-added activities.

AI Has Come To Mortgage Lending

Artificial Intelligence is coming!  The use of artificial intelligence is beginning to make its mark across most, if not all, industries.  A current TV ad shows a worker in the Carlsbad Beer company explaining how they develop new products using artificial intelligence.  More and more we are seeing and hearing about how it is impacting our lives.  It is foolish to deny that it will have a huge impact on the mortgage industry or that this technology will have significant impact on the overall operations of an organization, whether it is origination, servicing or secondary.  

Featured Sponsors:


The advancement of artificial intelligence technology opens new opportunities for lenders.  Most common discussions include the reduction in costs, coming primarily from the elimination of staff positions that will no longer be needed to complete repetitive tasks. Among these positions loan officers, processors and closers are most frequently mentioned.  Yet backroom operations seem to be a perfect fit for this type of technology.  For example, Narrow AI is focused on a single task reflecting work that is a comparison of documents delivered after the disbursement of funds.  Setting up loans for delivery to a servicing system and filing and payment of insurance are also single focused tasks that an AI program can easily accomplish once trained.  

Featured Sponsors:


Today’s focus on providing exemplary customer service can be aided by the use of properly trained bots. These programs will be able to handle more questions and explain more of the product features than ever before. This will give loan officers, processors and closers the opportunity to spend less time on answering these standard questions and will allow these individuals to focus on the more complex issues facing their applicants.  

Featured Sponsors:


One of the best suited for the type of activity that is best served by AI is the post close review processes and back-office reviews such as the quality control function.  Currently agency requirements include a pre-funding review and a post-closing review.   Most lenders have implemented two separate functions to meet these obligations.  In addition, because of the current emphasis on regulatory requirements, lenders have also implemented a third group of individuals who review only the documents associated with these requirements. AI technology can be trained through machine learning to conduct these reviews.  Not only does this reduce the headcount and general overhead costs, the reviews could be conducted on 100% of the loans.  While there will be variances that require human review at first, as the technology becomes adept at recognizing these variances and how they are resolved, the issues will also be addressed by the technology.  More importantly, as the volume of loans processed by this technology grows and the actions taken to resolve the issues become repetitive, AI can learn how to handle more and more of these processes. 

Featured Sponsors:


Servicing also has significant opportunities with the introduction of AI technology.  Most of the positions involved in the administration of payment processing, escrows and related activities are typically uniform and repetitive.  These are prime opportunities for this technology to be embedded.  The technology can be taught to handle these functions and when unrecognized variations occur, the issue and the related loan function can be given to a human to resolve.  

Default management will undoubtedly be one of the biggest beneficiaries of AI. This is not only due to handling of the functions associated with these activities, but with the opportunities of identifying potential defaults long before an actual default occurs. Using “deep leaning” AI, the neural networks can incorporate more and more external databases into its analysis which will increase the probability that warning signs of default will become familiar and reliable.  Imagine the system using not only information on the borrower’s credit and property value changes, but economic data, news related to job growth and company movements and historic performance of similar loans, to analyze if a borrower will continue to pay.  

Another area that holds great promise for servicing is mortgage insurance claims.  Anyone who has worked in servicing is aware of the difficulty of getting claims paid, especially FHA loans.  This technology can be taught the allowable amounts reimbursable and prepare claim forms.  Based on any rejections received, the technology will adjust going forward until very few, if any, are rejected.  Furthermore, it can adapt the payment rates for the property preservation and marketing specialists involved in the process.  

The secondary market will also be the recipient of the benefits of AI.  With the depth of information used in RMBS pools, AI will be able to provide analyses that reflect the true risk of performance rather than the quasi-accurate data used today.  It is conceivable that as AI matures, lenders will be able to “score” each individual loan and thus have the ability to provide individualized pricing. 

Despite the benefits envisioned by this technology, there are risks.  Some of these are specific to the mortgage industry while others are expected to have a profound effect on the economy as a whole.  According to many futurists and technology developers, 40 % of all companies that exist today will not survive the adaptation of artificial intelligence.  There is nothing at this point that tells us whether or not they will be large or small, local or world-wide.  Much depends on how rapidly the technology advances and which of these companies have a management team that leads the changes or instead resists this technology to the detriment of their organizations.   Ultimately as the technology matures there is the potential for massive economic and job market disruptions.  

What then are the most pervasive risks that the industry faces?  The most critical is Data.  Data is the primary fuel that powers this technology and is the most critical factor in the success or failure of AI.  The more data available to “teach” the programs, the stronger and more meaningful the results.  For example, in developing “Watson”, the trainers feed the program over one million books which contained all types of information.  With the massive number of neural networks created by this data it provided the information that allowed Watson to beat the two Jeopardy champions.  

While there are millions of mortgages with corresponding data in existence today, lenders have been notoriously hesitant to share any of the data associated with these loans, especially the servicing data.  In addition, the data found in origination systems is also critical along with the quality control data which identifies the variances found in loans and can relate these findings to loan performance.  Unfortunately, QC data has had the least structure associated with it and as a result, there are have different definitions and input across the industry.  

Without the ability to combine these massive amounts of data while ensuring its accuracy, AI results are less than reliable.  While MISMO has done a stellar job of defining data fields, they cannot control what is input into those fields in every system.  In addition, some of the most critical data is found not in the data fields, but in free form notes and comments.  While AI has the ability to review and classify unstructured data, the ability of the industry to collect and validate these extremely large amounts of data is a huge risk and will require a massive industry wide effort to have what is necessary to support AI programs.  

Another risk facing lenders is the inherent bias unintentionally built into the programs by those who are teaching it to learn.  For example, if the program trainer taught the pronoun “her” when referencing a nurse, an application for a male nurse would most likely be rejected by the program. This type of unintentional bias can exist in numerous programs and would not necessarily be discovered until it was identified by those individuals impacted by it.  

The use of AI programs would ultimately change the operations processes and the jobs associated with each piece of the operation.  While on the surface, this may appear to be a non-issue.  Human based jobs associated with the development and use of these programs focus mostly on functions that are a hybrid of technology and human knowledge.  The need for experts in all areas of the organization will be necessary to complete the programming of these programs is critical.  Yet, since the introduction of automated underwriting systems, we have seen a drastic decline in the number of specialists in each area decline. Underwriters who have spent years learning what an underwriter needs to know are rare these days.  It is more likely to have “underwriters” who are simply inputting data and taking the output from an AUS system in making a final decision, without understanding what is needed or corrected. 

Government regulations, that have put a virtual stranglehold on the industry and its operations will most likely be adapted to this new work environment.  While machines can be taught to produce disclosures and documentation, their necessity will need to be examined more thoroughly.  The risk of course is that any old regulations will be replaced with new ones, regardless of the industry.  For example, property values that are developed through on-site inspections and evaluations may be entirely replaced by deep learning programs and huge databases which are currently being developed by companies such as Zillow and Google.  

Artificial intelligence will not impact our industry, or any other industry for that matter, in isolation. The changes brought about by the advancement of this technology will impact everyone.  We have already begun to see the shift of employees away from working in offices every day, to one where many people now work from home, wherever home may be.  We see more interactive work experiences achieved through technology rather having to actually go to another office in another state or even country.  

We are also more likely to see new types of organizations or combinations of work types that have never before been considered.  Capital One has introduced the idea of “Capital One Cafes” where banking and coffee needs are serviced in one spot and I’m sure it is not the last, or only one that will occur.  AI analytics can identify where “life events” intersect and provide opportunities that will change what we need, what we expect and the timing of its receipt.  

Based on “deep learning” patterns and relationships found in lending and servicing data, it is most likely that the credit culture prevalent in the industry today will shift from a front-end focus to becoming servicing results driven.  Because of the ability to identify patterns of loan attributes, economic data, demographics and even global issues, the actual performance of loans can be translated into the ability of companies to identify the most advantageous risk profiles.  These profiles can then be incorporated into the decision tools used for evaluating applications.  

With this type of technology and the ability to predict performance, will the delivery of loans to the secondary market change?  There is a distinct possibility that lenders of all size and scope, will be able to negotiate individualized product sales, thereby eliminating the need for secondary market intermediaries.    

A study by the board of Governors of the Federal Reserve found the pace and ubiquity of AI innovation to be much greater than expected for the financial services industries. They also found many positives in the use of AI as well as the risks.  Whether you have begun to work with artificial intelligence tools or not, they will have an immense impact on mortgage lenders.  

The issues discussed here are mainly focused on the broader problems and opportunities facing us in the near future.  What we cannot ignore however, is the issues surrounding how our work is to be done. In the final article of the series, I will take a look at the potential processes and job functions that are the changes job functions will incur and the skills and knowledge necessary for management and employees. 

About The Author


Being touted as the answer to all the stresses and dissatisfaction currently found in the workplace, it is being developed as the advancement promised by technology prognosticators who have promised that work can be reimagined to be more flexible, faster and adaptable. A benefit, or so we are told, is the elimination of work that is standardized and repetitious.

So, is this good for mortgage lending?  Reading numerous articles in industry trade magazines and discussing these ideas with those who are currently mortgage lenders, it is apparent that we actually know very little about AI. Every article I have read claims to have found the answerandgives us a short, simplified explanation about what it actually does.  To add more confusion, artificial intelligence technology is sprouting a completely new set of acronyms which are anything but helpful unless you thoroughly understand it.  Another issue I found prevalent in the discussion is the fear about its usage and the elimination of jobs we now hold. But are there any risks in using AI that we don’t know about?  Do we understand how technology can make better decisions that humans?  What do we need to have to make sure the technology works accurately?  How can we implement it into our production and servicing processes?  So, before we rush head-long into the next great thing, we need to take the time to really understand what this technology is and does. 

Featured Sponsors:


This series of articles provides some basic definitions utilized by artificial intelligence technology providers, identifies the various iterations now in use and describes what critical elements are necessary for it to do what it claims to do. In addition to the technology itself, the articles will focus on the risks and opportunities associated with AI as well as discussing the potential for process and job function disruptions the industry will face as it is implemented. 

Beginning with the basics

“Work”, the basis of economic betterment, is changing again.  Known as the “Human Economy”, work involves people performing activities that are considered economically productive.  From the earliest days of economic activity humans have experienced the need to improve productivity.  Work, as we think of it today began with the development of wheels, wagons and animals for transportation as well as the use of small hand-tools for creating individual products one at a time. Next came the industrialization of product production, and then the initialization of automation to conduct the actual work. Ultimately this has led to better products, more satisfied customers and greater profitability while providing a better income and more time for leisure and family time for employed individuals. We are now in the initial stages of the next advancement, the use of artificial intelligence, to once again change what work is and it isn’t. Despite numerous nay-sayers, this new technology is bound to have a significant and fundamental change to our business. We, as an industry, need to know more. 

Featured Sponsors:


As defined by Paul Daugherty and James Wilson in their book, Human + Machine, Reimagining Work inthe Age of AI,artificial intelligence is a system that extends human capability by sensing, comprehending, acting and learning.  In other words, it is the simulation of intelligent behavior in computers or sometimes referred to as a machine with human cognition and the ability to carry out tasks as a human would. 

Within this field there are various levels of “intelligent” program types. They have names such as Expert Systems, Machine Learning, Deep Learning, Robotic Process Automation (RPA), and Neural Networks. Another common acronym describing how AI works is Natural Language Processing. All of these terms reflect a variation on the basic purpose of AI, that of having machines do the work of humans.  But how it is done and what functions it addresses are quite different.   

Featured Sponsors:


Defining Types of Artificial Intelligence

Today, the term is discussed by many in our industry as a single effort of implementing AI, yet they confuse terms like Expert Systems, Machine Learning and Deep Learning, seeing it all as one new technology.  When you hear these buzzwords tossed around in conversations, it is obvious that many do not recognize these terms as very different types of AI.  In order to clear up the confusion a definition for each type must be recognized for what it does and an understanding of how it can be used, beginning with simplest to the most advanced.  

All artificial intelligence products are based on one core requirement; the ability of the system to collect and learn facts.  Just as individuals are not born with the knowledge they have today, a machine must learn what it needs to know.  First these machines must be taught the basic facts. Even once these basic facts are learned, they are not innately able to associate those facts with the problem presented to it. It must be taught how to associate these facts appropriately with the problem it is trying to solve. As the complexity of the problem increases, the need for these machines to conduct this association becomes more complex, thus the varying levels of AI. 

Featured Sponsors:


Because AI is a relatively new technology, scientists do not necessarily agree on the terms and definitions utilized in describing the capabilities of current AI products.  The definitions and examples given below are most consistent with the terminology we use today.  In addition, there are terms that are frequently used within the scientific community for describing the abilities of specific AI functionality.  

Expert Systems– The most basic of AI products, this technology is a method of automated reasoning based on a very specific set of facts, rules and principles.  The automated underwriting systems we use today are an example. When we ask a question, such as does this loan meet our credit guidelines, the program takes the facts as provided in the application and from external data, compares them to the facts taught to it as an acceptable loan and filters this data to arrive at an answer. 

If any of the data is inconsistent with the “facts” taught to the machine, or there is no rule, it will be unable to decide.  These systems do not “learn”.  For example, if an AUS system has a rule that limits DTI ratios to 43%, it will not approve one at 44%, regardless of the fact that the human who then reviews the application does so 95% of the time.  The system has not “learned” that 44% is acceptable.  

Robotic Process Automation-RPA was designed and is utilized to automate those processes that are routine and labor intensive.  Today businesses find that work associated with repetitive tasks, such as inputting data, making calculations and answering standard customer questions is work that can be done by basic RPA technology.  Your home-based Alexis system is an RPA devise that answers questions and conducts some limited analyses based on the input.  For example, when you ask Alexis to play music, it will find a music program and begin to play it. The robot has learned what music is, the types of music and what performers are associated with each.  If you do not like the type of music selected you simply tell Alexis and the music is changed.  The same applies to individual songs.  Alexis then “learns” what type or piece of music you do not like and will not play it again.  The “bot” as it is now called, has learned something new and will apply it in the future.  

Today, companies with workflows that are consistently repetitive and simple are using this technology.  For example, a warehouse which contains thousands of products which need to be compiled into one order uses these bots to find and deliver the items to a central location.  These bots, which have the appearance of what we think of as robots, move up and down narrow rows of products and quickly and accurately collect the necessary items. This has resulted in the elimination of personnel who perform this function with much less accuracy and in a greater time period.  It even has the impact of allowing warehouses to be built higher, thereby eliminating the need to have long, low buildings with lower rows of goods so that humans can reach.  

One of the critical features for these bots is the ability for it to recognize what the user is saying in their own vernacular. Known as Natural Language Processing (NLP), its purpose is to allow bots to learn various languages, accents and idioms used by customers. The development of NPL requires that this occurs for industry specific terms as well.  Asking a question of a bot when the terms are used in different scenarios will most likely not get the answer you need.  Again, if the bot has been trained in the Northeast and is asked a question from someone in the South, without NLP adaptation to this accent, it will have problems answering the questions. 

Narrow A.I.- This term has come to be used by many of the individuals working in this field. Like its name suggests, it is focused on executing a single task.  Human interactions with a narrow A.I. are limited because Narrow A.I. can’t think for itself. This is why sometimes you’ll get a nonsensical answer back when attempting to use it because it lacks the ability to understand context.

General A.I. General A.I. or Strong A.I., as it is sometimes called, provides the ability to understand context and make judgments based on that context. Over time it learns from experience and is even able to make decisions, even in times of uncertainty.  Even with no prior available data it can use reason and be creative. Intellectually, these computers operate much like the human brain. This is where we are headed when we talk about its functioning in place of a human.  To understand the abilities of this technology and how it operates, the basic functioning of the human brain needs to be understood.  

Neural Networks. Each individual has within their brain a series of networks which transmit data. This can be a s simple as learning what tastes you like and what you don’t like.  Before you learn, by tasting, that you don’t like fried liver there is nothing in your thought processes that tell you it is not acceptable to you.  However, once you taste it and realize it tastes awful to you, you reject this food choice because your brain has learned from this observational data.  Our internal neural network processes the data we are observing and alerts us to the fact that fried liver tastes awful to us and alerts us to that fact before we make an unacceptable choice.  

Developing these neural networks in our brains takes time. My son recently bought a 2-month-old puppy.  Obviously the first thing he needed to do was house train him and so he has spent a great deal of time making sure Bogey understands what is expected and what he must do when he needs to go.  Developing this neural network in the dog’s mind is not an easy task.  Over time, the training took hold.  In other words, a neural pathway had been developed. 

Machine Learning. Machine Learning, “ML”, technology. This is the field of computer science with algorithms that learn from data that is “taught’ to the technology.  It also uses the data and incorporates algorithms that learn from and make predictionson data.  This technology can learn from humans or it can learn from other data. In other words, this technology is conducting some of the analytic thought used in bringing various components together and allows it, to some extent, to predict a result.  Another utilization of ML is known as “supervised learning”.  In this scenario data is broken into categories such as inputs and outputs and humans use it to develop an expected output from the technology. When the output is inconsistent with what is expected based on the algorithm, it sends out an alert to the user.

One of the most common uses of Machine Learning is found in fraud investigation. In these cases, the machine is feed information and searches for a pattern of data, such as spending habits or income, and utilizes its algorithmic tools to determine if the data is what would be expected.  If in using the previously received data, it determines that the new data does not fit the pattern the output reflects this inconsistency and potential fraud is identified. This is what happens before you get that phone call asking if you had made a recent trip to Saudi Arabia. 

Deep Learning. Deep Learning specifically refers to the ability of the machine to learn from other data. Deep Learning consists of a multi-layered network of algorithms in which data is processed through multiple layers with each layer as individual inputs.  It also allows the data to flow back and forth between these layers thereby expanding the systems knowledge base.    Deep Learning “DL” machines can analyze patterns and people to identify potential problems or opportunities far enough in advance to allow for early intervention or resolution.  Imagine the ability to know the probability that a borrower will default even before the loan has closed and been set up in servicing.  

These types of artificial systems, machine and deep learning, are both predictive systems. These systems find relationships between variables in the historical data, identify a pattern and then develop a model to predict future outcomes.  Developing these patterns requires the use all available data, including borrower information from sources other than the application, property information, both local, state and national, economic conditions, etc., that are now available provide the opportunities to find correlations never before considered.  By understanding these different approaches to artificial intelligence, it is much easier for management to evaluate the type needed to successfully implement such a program.  There are however other issues to be understood and addressed before taking the leap to artificial intelligence.  The next article will explore the risks and opportunities of AI along with new developments addressing these issues.  In addition, it will discuss what it means to “reimagine” how work will be done. Finally, we will discuss what a reimagined mortgage operation looks like and the impact its inclusion will have on the workflow and job functions.  After all, don’t you want to know if a robot will be taking your job?

About The Author


It’s been said that artificial intelligence is the future, but I’d argue that AI is very much a thing of the here and now. It’s playing an increasingly significant role in marketing efforts, and is taking marketing automation to the next level. And during an era when customers are demanding fast and hyper-personalized service, AI-based technologies couldn’t be more critical. AI-based technologies bolster marketing automation efforts through personalized interactions. Your business can benefit on a multitude of fronts by embracing these game-changing advancements.

Featured Sponsors:


Email marketing

It’s easy for emails to get lost in the shuffle of a busy day. After all, inboxes are overflowing with meeting reminders, work correspondence and plenty of promotions. Enticing a customer to simply open a marketing-related email can be challenging. But with the right tools, you can consider that challenge conquered. 

Featured Sponsors:


AI-based technologies are helping marketers win the attention of potential customers in numerous ways. For instance, they’re allowing small- and medium sized businesses to improve their email marketing automation strategies by knowing who to target, when and how frequently.

Know your target:AI helps businesses understand – and foresee – customer behavior patterns, and detect which strategies email subscribers best respond to. It can also help businesses more effectively automate emails to ensure customers are receiving personalized correspondence and promotions relevant to their needs.

Featured Sponsors:


Adding a personalized touch to emails pays off. A Statisa study shows that personalized emails had an open rate of 18.1 percent, compared to a 13.1 open rate for non-personalized emails. 

Engage thy customer:Of course, AI-generated emails should do more than attract customers – they should also engage them. Email subject lines can be tailored for each customer, for example, while the emails themselves might contain personalized offers. 

Featured Sponsors:


Lead nurturing 

When it comes to gathering and analyzing customer information, we mortals are no match for the pace and efficiency of computers. AI-based programs are capable of processing user data at lightning speed, and can be hyper-responsive to customer needs.

Automating the data collection process can help simplify the lead-nurturing process for businesses, while AI can set the course for future marketing strategies.

Capturing information:AI provides marketers with the valuable information they need to close a sale. It tells us when someone has visited their site, which pages they visited and how much time they spent perusing products. 

Marketing teams are using this data to drive sales, and if statistics are any indication, it’s working. AI could lead to an economic boost of $14 trillion in additional gross value added (GVA) by 2035, according to Accenture research.

Automated actions:As I mentioned earlier, AI-based software doesn’t just benefit businesses in the long term — it can also be of tremendous value in the present. Through AI, valuable customer information is being captured in real time, and can be instantly used to tempt shoppers with personalized offers and targeted advertisements. Forrester has predicted that 20 percent of enterprises will begin using AI to make decisions and provide real-time marketing instructions.


Successful businesses strive to equip customers with the tools and services they need. The trouble is, it can be a bit tricky to determine what those needs are.

AI aid:While automation can help track previous purchases and user experiences, AI can take that information and make predictions on the potential for additional sales. It can determine whether a customer is a good candidate for additional products or services and establish which products a customer might be in the market for next. 

Once your sales and marketing teams are made aware of your business’s cross-selling potential, they can make customers aware of those additional offerings.


A happy customer is a loyal customer. AI can help you retain your existing customer base by providing you with a better understanding of client needs. You might think you already have a handle on that front with your CRM platform, but AI can provide a more complete, more actionable picture of your current customers.

Meantime, customers will appreciate the personalized service AI-based software offers, and are more likely to become repeat customers.

Reengaging past customers

Customers come and go – and they’ll come again with the right approach. Whatever the reason for losing a customer, it’s never too late to win them back – especially with some artificial assistance.

Helpful insights:AI can provide insight into how and why a customer left, and help determine the best way to reconnect. For example, it can be used to send an email with product recommendations based on a customer’s purchase history, or inform them of any relevant price drops.

If you’ve spent any amount of time looking into customer engagement and sales/marketing success strategy, you’re undoubtedly already familiar with the ways in which automation can help your business. Now, it’s time to build on those benefits by incorporating AI-based technologies into your marketing strategy. 

Businesses are becoming increasingly competitive and to remain relevant, AI-based technologies must be embraced.

About The Author

The Evolving Roles Of Mortgage Originators And Loan Officers

For years, many in the mortgage industry have used the terms “mortgage originator” and “loan officer” interchangeably. Today however, a new trend is emerging and the two monikers are developing into their own unique, individual roles. To understand this changing dynamic, and what their jobs will look like in the future, we’re going to dissect the two and see where the differences lie.

Featured Sponsors:


While technology has helped simplify management of the more tedious processes of mortgage origination, sales and borrower engagement, there remains a great deal of information for lenders to keep up with. From creative marketing strategies to new regulations and products, it is here where loan originators tend to shine. Rather than engaging with borrowers on the minute details of their loans, loan originators tend to focus instead on their overall book of business and the “big picture” strategies for clients. I occasionally also refer to them as loan planners or mortgage consultants. 

Featured Sponsors:


On the flip side, loan officerstend to be more adept at driving business through their branch or office. For these professionals, stress tends to build as they gather together a large book of business, but then have to back-up their sales with high-level research and study. Their preference is often to work directly with their borrowers on an individualized basis and help them identify the best loan for their own unique financial situations. These professionals are also tend to have a background in processing or underwriting, which typically means they are highly attentive to small details and nuances of specific loan structures.

Featured Sponsors:


That’s not to say one is better than the other, or more necessary to a productive mortgage business. In fact, the most successful lenders have the two tracks running parallel with each other and functioning akin to a pilot and co-pilot, or even a counsel and co-counsel in legal fields.  

Featured Sponsors:


This changing dynamic is important as it puts the borrower first by creating a synergy between the loan originator and loan officer, with both playing to each other’s strengths. Frankly, I see the loan originator as representing the first steps of a borrower’s journey, helping him or her identify possible strategies for their mortgage, building trust and offering general guidance on issues like credit scores, for example. Once this framework strategy is in place, borrowers can seamlessly shift to working with loan officers who can help them with the minute details and identify their personal pathway to debt-free homeownership. 

Taking a more in-depth look, one can see that it’s the originator who lays out the ‘big picture’ or 30,000 foot view for the borrower, and helps them narrow their focus to one or a few select options. Then, the loan officer helps them narrow that view even further to create a smarter mortgage plan that allows them to ultimately save money and build personal wealth. This works to the advantage of borrowers as they have access to not one, but two experts in the mortgage industry, each with their own specialization to help them identify the best possible loan. 

This trend will continue to develop as technology impacts how lenders do business with borrowers, particularly as data, machine learning and AI come to play as an integral part in the lending process. Using data, lenders are able to identify and interact intelligently with borrowers at the start of their decision making process, establishing trust prior to them even starting the mortgage process. When the borrower does decide to move forward with originating a loan, lenders will be better prepared to provide them the best quality advice regarding the loan options available. While the role of originator and officer may be diverging in some respects, when a borrower is deciding where to take their business, they do not look for individual titles or positions. Rather, they look for businesses that respect their financial position, care about their future and work with honesty and character. That’s why this changing dynamic is important, not because of job titles, but because it will offer lenders more opportunities to demonstrate their expertise, educate borrowers and guide them down a homeownership path that supports their long-term financial wellness.  

About The Author

AI-Based Textual Analysis

In today’s mortgage market you can’t pick up an industry publication or attend a trade show and not hear someone talk about Artificial Intelligence (AI).  Many of the claims talk about how AI will disrupt the mortgage industry or radically change how mortgages are originated or processed.

Featured Sponsors:


Some of it is pure hype and some of it includes technology that can clearly streamline processes and reduce cost.  So how do you filter fact from fiction?  Let’s take a look at how AI-based textual analysis is actually at work in the mortgage industry and the different approaches some vendors are using.

Featured Sponsors:


High-Level Approach

The most fundamental difference in approach between an AI textual analysis and the majority of other “advanced” document recognition technologies is that AI textual analysis treats variable layout documents as unstructured documents whereas most other prominent solutions treat them more as semi-structured documents.

Featured Sponsors:


To illustrate the difference in methodology let us consider a customer wishing to process pay stub documents. An AI textual analysis implementation would typically be deployed with one set of completely generic rules designed to encompass all variations of the “Pay Stub” document type from any company.  Because alltext is evaluated by the AI engine, the rules can flex with the layout and verbiage changes just as a human does when reading the page. The solution is also capable of being configured to perform conditional processing for specific exceptions to generic rules (per-layout exceptions) but it is not typically necessary to do this.

Featured Sponsors:


Most other modern advanced document recognition technologies treat document variations as semi-structured documents. These solutions typically either: 

a) Remember as many of the variations as is practical and process each variation with layout-specific templates for processing Or b) Apply a mix of layout-specific processing and some generic processing (usually the higher incidence layouts are processed with layout-specific or templated processing) 

The obvious advantage of the AI textual analysis generic approach is that, as new layouts appear or existing layouts change, the software is better equipped to deal with these new variations. This advantage was recently validated at an account with over 35,000 layout variations where rules had been in place for five years with not a single modification. An audit of this client’s processes after five years revealed an identical automation rate to that when the system was first deployed. This outcome was observed despite the fact that a significant portion of the originally dominant layouts had been transferred to EDI processingand were therefore bypassing the system now. 

As you can see the varying approaches often produce different levels of success.  But how can you determine which is best for your organization? Get a demo and hope that the solution is more than smoke and mirrors?  Up until now that is usually the case.  To overcome much of the confusion and disappointment, a better evaluation process in many cases may go a long way towards greatly minimizing the risks involved in choosing a vendor who can actually deliver AI-based textual analysis.

In order to quickly understand AI-based textual analysis and its capabilities, a blind test with several sample files should be considered the gold standard for an evaluation.  This is especially true when it comes to the challenges presented with the many and varying document types and quality levels of document images found in the mortgage industry.  Asking vendors if they are willing to perform a test on a never before seen sample set of typical loan files on site and in sight of your evaluation team, is a great first step in separating fact from fiction.

Ideally, an evaluation should be setup as a one day event to hedge against any vendor refining their results.  This kind of test is intended to demonstrate the validity of the vendors’ out-of-the-box capabilities so that prospects can be assured that they are considering a proven, robust and scalable solution ready to deliver productivity improvements in weeks rather than months or years.

For qualified opportunities, Paradatec will perform this process, which enables prospective clients to quickly understand the overall levels of automation, and speed improvements they will be able to achieve with their technology.  Download our whitepaper on AI based textual analysis

About The Author

St. Louis Loan Originator Bets On FinKube To Help Increase Home Finance Business

FinKube, a company that provides AI-powered Platform-as-a-Service solutions for a range of industries, announced that St. Louis-based LenderCity has successfully deployed ELSA, FinKube’s Electronic Loan Services Assistant. The mortgage industry’s first chatbot is already interacting with prospective borrowers on the LenderCity website.

Featured Sponsors:


“Consumers want immediate answers to their home finance questions and ELSA is smart enough to provide the information they need and gather the information we need to prequalify the borrower,” said Gregg Harris, principal at LenderCity. “We know we need to respond very quickly to borrower requests for information, but we also want to capture as much information from them as we can, without taking up the loan officer’s time. FinKube’s ELSA is the answer.”

Featured Sponsors:


ELSA is an intelligent assistant that uses AI and machine learning to enhance the origination process from origination to close. Her AI is powerful enough to gather borrower information, render decisions, automate time-consuming tasks and help lenders produce fully compliant mortgage loans in as few as 20 days, though she is well versed in any form of consumer lending.

Featured Sponsors:


“American home buyers still want to visit with a live loan officer before signing on a new mortgage, but it’s not efficient to spend the loan officer’s time in conference with borrowers who do not qualify,” said Jorge Sauri, founder and CEO of FinKube. “At the same time, those borrowers who do qualify expect to have their questions answered immediately. They don’t want to wait for a call back. They want to feel like they are in control. Most chatbot technology cannot offer that, but ELSA does.”

Featured Sponsors:


With ELSA, LenderCity is able to hyper personalize the home finance transaction at every stage. Studies have shown that this:

>>Reduces customer acquisition costs by 50-80%

>>Increases engagement and conversion by 500%

>>Reduces customer service costs by 50-80%

>>Increases loan retention by a factor of 6

As an A.I. powered virtual assistant, ELSA works 24/7/365 pre-qualifying leads, communicating with customers and synchronizing outreach across chat, text, voice, email, and mobile wallet. In addition, FinKube can deploy ELSA ten times faster than generic chatbots that can’t speak mortgage out of the box.

Leading-Edge AI In Mortgage Lending

Previously CEO of HeavyWater Inc., the mortgage-focused Artificial Intelligence (AI) provider recently acquired by Black Knight, Inc, Soofi Safavi now serves as Managing Director of Black Knight’s Applied AI group, bringing leading-edge AI and computing capabilities to the Black Knight product portfolio. With over 20 years of experience in mortgage and banking technology, and deep expertise in IT strategy, architecture and machine learning, Soofi is uniquely suited to discuss AI’s role in the mortgage industry.

Q: It seems as though Artificial Intelligence (AI) has resulted in incredible advancements across so many industries, but we haven’t seen the same in the mortgage industry. Why is that?

SOOFI SAFAVI: I would say that we simply haven’t seen it –yet. The mortgage industry is a very complex environment that requires vast amounts of expert knowledge to navigate effectively. That is why there is still so much work to be to be done despite a never-ending quest for increasing efficiency through technology. AI will be key for not only increasing efficiencies, but also identifying and eliminating deficiencies.

Featured Sponsors:


Every day in our industry, we have thousands of experts leveraging technology to perform a series of activities and tasks that are essential to the mortgage process, from origination through servicing. The goal with a mortgage industry-focused AI is to capture that knowledge and replicate it within algorithms via machine learning.

From a high level, AI is about using computer systems that are able to perform tasks normally requiring human intervention. Machine learning is a type of AI that allows computers to learn without being explicitly programmed. It is a technology that uses algorithms to learn from and make predictions on data. It reads, comprehends and draws conclusions based on context to mimic human thinking and build expertise over time.  

An AI that can learn from the experts who are currently performing all of the many tasks involved in the mortgage process can offload much of the manual effort to technology. The mortgage professionals doing that manual work today can then shift into a different role, one more suited to the knowledge worker-based economy of the 21st century. They become teachers and guides of the AI, and make the decisions only humans can make. 

Featured Sponsors:


Q: That raises an interesting question. There is a lot of uncertainty associated with AI. Many feel that individual’s jobs may be at risk if AI takes over much of the manual processing associated with the mortgage industry. What would you say to these people?

SOOFI SAFAVI: My answer would be twofold. First, I’d say look to history. Repeatedly in our industry – and others – the introduction of new technology has spurred fears of machines replacing people. When automated underwriting systems first came on the scene, many underwriters feared for their jobs. In fact, what they did was reduce friction in the mortgage process, allowing for what had been an entirely manual process to become more streamlined. The end result was increased efficiency and throughput, calling for far moremortgage professionals, not less.

While AI is light years ahead of automated underwriting, I expect the same will be true today. In the current environment, there is a significant amount of rigidness in mortgage origination, and people tend to go through this process a handful of times in their life. The potential for reducing friction in the mortgage process has increased exponentially with the advent of AI. What has for many years been a long, exhaustive and laborious process – on the part of the homebuyer as well as those of us in the industry – will see reduced friction via the automation afforded by AI. 

Featured Sponsors:


An AI that has been taught to perform traditionally repetitive functions can do so more quickly and accurately than traditional methods. For example, verifying income, assets and insurance coverage; all traditionally manual activities that take hours to complete and are prone to error. Putting AI to work on these stare-and-compare tasks frees up highly-skilled mortgage professionals to focus on creating value, enhancing the customer experience and expanding production rather than simply executing repetitive functions.

Reduced friction equates to increased opportunity, for borrowers and mortgage professionals alike. If you remove that friction, and the underlying operational inefficiencies behind them, the home buying process will become much more fluid. A smoother, simpler process augmented by technology becomes one that can occur with more frequency throughout an individual’s life. And that opens the door to more innovation around products – loan products, technology products, credit products, and more – to support that increased frequency.

I would imagine that, much like was experienced with automated underwriting, a frictionless process will result in more loans being made, and more jobs for skilled professionals using AI-empowered tools. Not only that, but it will result in more jobs across the housing spectrum. All in all, we’re going to be looking at a much more interactive, more fruitful marketplace.

Q: The promise of AI in the mortgage industry seems incredible, but realistically, what sort of resources are required for an implementation of that level? 

SOOFI SAFAVI: The benefits of AI are not dictated by the size of an organization. In fact, mortgage industry players of all sizes can benefit from AI today. Black Knight’s own AI virtual assistant – AIVA – can be brought into an organization in much the same way as any other resource. An originator, or servicing shop, can “hire” AIVA to assist with specific functions or tasks. 

Much like any other employee, AIVA arrives for work with a certain skill set – it’s why she was hired in the first place. Of course, there is also an onboarding period where AIVA learns what is expected of her in this specific role, and is taught the specific process intricacies of a given organization, but after that, she is then deployed in the same way as any of the organizations other employees. And the skills she develops in the process become part of her knowledge base moving forward. 

Of course, at the enterprise level, when an organization’s operations reach across multiple verticals within the mortgage arena, the potential benefits increase exponentially. Rich, deep data is the fuel on which an AI runs, and the more data is available to AIVA, the more implementations become feasible.

But it’s important to stress that AIVA is not something that is only available to the largest lenders or servicers, but it is a resource that can be made available to organizations of all sizes. AIVA is as applicable in origination as she is in servicing, or in other facets of the industry.

Q: So an AIVA in every shop?

SOOFI SAFAVI: Let’s back up a bit, because I think this will be helpful in painting the entire picture. At the point of origination, a great deal of information is gathered on a prospective customer. That information, or some subset of it, is of use throughout the loan lifecycle – from application, through origination, settlement, closing, servicing, and if need be, modification or default management. All of these different players are gathering and processing information, and there is a great deal of overlap.

Each player in the mortgage process needs to receive a full file, and extract their own role-specific data from that document and then push it through their core system, to effectively complete their piece of the mortgage process. When you stop to think about it, for many of the players involved, roughly 60 percent of the information they need, or the calculations they make, mimic – or at least closely align with – activity the originator has already completed. 

Not only can an AI do that analysis and ascertain the 60 percent of information and analysis that has already been done, but it can go further. Rather than starting fresh each time, with the time and cost associated with these activities, those conclusions are presented by the AI, because it’s looking at the entire process holistically. Which adds to the unseen value. 

This points to the larger benefits of AI, its ability to learn. The more information an AI has at its disposal, and the more skills it is taught, the more places in the process it can add significant value. That same 60 percent share of information – perhaps more – that carries over from origination to, say, a title provider, also has value across the entirety of the loan lifecycle, to multiple players involved in the process.  

Not only can AI cut significant amounts of time from the process, it can also make data-based decision-making much more easily accessible for the all of the constituents involved. And that improves the process all across the board.

Q: Any final thoughts on AI in the mortgage industry, particularly for originators?

SOOFI SAFAVI: Black Knight’s first goal for AIVA is to drive down the cost to originate a loan by maximizing efficiencies and eliminating inefficiencies through the introduction of cognitive automation.

Whereas today’s workflow orchestration engines do a fantastic job of increasing efficiencies by alerting users to tasks that must be completed – a bank statement or paystub has arrived and is ready for review – AI can proactively evaluatethat document based upon its understanding of the mortgage lexicon. It leverages that expertise – which is continually expanding via machine learning – and a deeper understanding of associated data/behavior to see if there are any red flags or missing elements and inject a sense of urgency in getting those things addressed. 

Orchestration engines exist to help humans work more efficiently. The intent of AI, and particularly AIVA, is to help them to work less on mundane tasks so their capacity grows.  In short, work less, work better, by delegating some of the work to your virtual assistant. Then, the mortgage professional and his or her expertise shifts to verifying what the assistant has produced, providing the all-important human level of interaction our industry depends on.

Other goals focus on improving servicing functions and creating actionable intelligence, for improved efficiency across a company.

Ultimately, the name of the game is applied AI, not simply AI for AI’s sake. With applied AI, our goal is to bring cognitive automation to where the bulk of work and activity is happening. Our industry has come to accept a 45-day average mortgage cycle time, an $8500-$9500 average cost per loan, and the need for some 15-20 people having to touch a loan to get it through closing and beyond. What we’re trying to do is shake that acceptance and teach AIVA to automate the bulk of that work is being done. Again, rather than simply using AI for AI’s sake, we’re trying to introduce cognitive automation where it will have the biggest possible impact in terms of reducing cycle times and costs. And thatwill transform the mortgage industry.


Soofi Safavi thinks:

1: The industry will face a human resources challenge, as it will become increasingly difficult to entice the digital native college graduate who leaves her smart home to commute to work in her self-driving car into processing underwriting documents all day long. 

2: Technological innovation – and perhaps more importantly – adoption will continue to accelerate; expect more cutting-edge innovations. 

3: Along the same lines – much of the technological innovation in the mortgage industry has been on low-hanging fruit (point of sales systems, etc.); we will see innovators begin to tackle the more complicated parts of the process.


Previously CEO of HeavyWater Inc., the mortgage-focused Artificial Intelligence (AI) provider recently acquired by Black Knight, Inc, Soofi Safavi now serves as Managing Director of Black Knight’s Applied AI group, bringing leading-edge AI and computing capabilities to the Black Knight product portfolio. With over 20 years of experience in mortgage and banking technology, and deep expertise in IT strategy, architecture and machine learning, Soofi is uniquely suited to discuss AI’s role in the mortgage industry.