04-26-2023 Artificial Intelligence With Gabe Minton Of Mortgage Connect
Artificial intelligence (AI) is a rapidly growing and evolving field that has many exciting developments and applications. Some of the current trends and advancements in AI today which is a hot topic is ChatGPT. As technology continues to advance, we can expect to see even more innovative and impactful uses of AI in the future. Gabe Minton of Mortgage Connect sat with me to discuss AI and its accountability and how it will be useful in the mortgage industry.
---
Artificial Intelligence With Gabe Minton Of Mortgage Connect
I'm excited to have Gabe Minton joining us. He serves as Mortgage Connect’s Chief Information Officer and Executive Vice President of Information Technology. He has an extensive background in this area and has served. He provided leadership for many companies and provided vision for technology information strategy throughout the industry, including the development of the next-generation digital platforms to facilitate seamless experience between the consumer and the client and to maximize operational efficiencies. That's one of my hot buttons. We are going to get into that a little bit in this interview. Gabe also has many years of leadership roles inside mortgage companies and technology companies, with a special focus on developing software systems, products, and strategies. Throughout his career, he has led strategy and execution, communications, and vendor relationships. Gabe served as a Chief Information Officer at Guild Mortgage, where he led its technology and information strategy, including business systems, architecture, infrastructure, product and services technology. Gabe has served in senior management strategy and technology positions at Black Knight ServiceLink, Motivity Solutions, Accenture Mortgage Cadence, and the Mortgage Bankers Association. He has also been a leading architect and founding member of MISMO, the Mortgage Industry Standard Maintenance Organization, a subsidiary of MBA. He has earned a Bachelor of Science with a double major in Computer Science and Statistics, a minor in Mathematics from Radford University in Virginia, and a master's of Software Engineering degree from the University of Maryland. What an oppressive background. Here we go.
---
Gabe Minton, it’s good to have you back on the show with me. It's been a little too long.
It's great to be here.
We are excited about that. You heard me give a little bit about your background. I'd like to have you add about it. Most importantly, share with us what you are most passionate about over the things you have done throughout your career.
I started working for the US Navy as a software engineer and computer scientist. I bring an engineering background to my career. I have come up over the years and got into the mortgage and never looked back. I have held a variety of roles, from technologists to working at the Mortgage Bankers Association to working at lenders' shops to working at service providers. I work at Mortgage Connect, which is a national service provider to the mortgage industry.
I have seen all sides of the table around mortgages. It’s similar a lot to you, except you have a lot more experience than even I have with all the things that you have done. I look to you as an example of that. What am I most passionate about? I love solving problems and working with people. The mortgage industry is small. We have known each other for a long time. I like the familiar faces and working together to solve problems at an industry level, at a company level, and at a lender level.
You are one heck of a problem solver. Not only that, I love the six working geniuses. We don't have time to get into that, but one of the things it talks about is wonder and invention. You are always wondering how we could improve this industry. You are talking about inventing new things. You have been on the cutting edge of many things that we have thought about accomplishing. I will never forget one dinner we had in Washington, DC, where we sat around. We solved all the problems. What's interesting about that dinner is the forward-thinking that you had and others that were sitting around the table. It addressed a lot of things that were implemented that are similar to what we discussed. You are a thought leader, and I'm excited about getting into the topic, which is artificial intelligence.Readers go back and read the previous episode that I did with Gabe. This was done way back in 2019. You will think we could have recorded it the other day. There's that much new information, but it'd be good for those who are not going to go back to that episode, but they should. Help us define AI relative to RPA and ML, Machine Learning, and Robotic Process Automation. Walk us through that.
I encourage everyone to read that episode. That was a lot of fun working with you. I'm practical about using these technologies in the mortgage space to solve problems. We will probably get into some of those examples a little bit later in the conversation, but I like to think of them as building off of each other. Robotics process automation is where it all starts. This is if you have a rote process that you repeatedly do all the time. It's a human being that's doing this. I go to this website, find this file in this directory, put it in this directory on this laptop, and do this every hour on the hour.
If you have a repeatable and finite defined process, a great thing that you can use robotic process automation around is doing rote repeatable tasks. If you do the road repeatable task over and over again and maybe use fuzzy logic, you start to get into machine learning, which is moving you towards artificial intelligence.
Artificial Intelligence: Robotic process automation is used for doing repeatable tasks. Let it do something over and over again, and it will move you towards artificial intelligence through machine learning.Explain fuzzy logic a little bit. I understand RPA. From this standpoint, it is defined exactly. If you put in this, you want this. You are talking about fuzziness. Is it the definition of the answer? How would you describe that?
You and I have brought up examples of assembly lines in the mortgage space. I remember talking to Doug Duncan about this at length. You could be assembling a car or a jet. Imagine if you are assembling a car and you are using robots to assemble the car. They do this. This is all real. When they put the frame of the car together, the robot has to put the nut in the right part of the right hole to put the frame together.
It has to be exact millimeters to get the threading to work, like if you are going to do it with a socket wrench. The machine comes forward with the screw, and it's not quite at the hole to thread the screw. It's smart enough via fuzzy logic to move about 2 millimeters over. I'm lined up with the hole. I will start turning, spin the nut into the hole, and get the next.
It's where it makes some adjustments.
It can make adjustments usually around boundary conditions that you put around the decision. If you take that to the next level, you get to our topic, which is artificial intelligence. Artificial intelligence, which I and every engineer will tell you, has been around for a long time. These are self-taught systems. It's not doing something that you repeatedly do, and you tell it how to do it. It's thinking on its own and coming up with a solution to what you are asking.
[bctt tweet="Artificial intelligence is not doing something you repeatedly do. It thinks on its own and comes up with a solution to what you need." username=""]
ChatGPT is a great example. You give ChatGPT a construct. The construct is a middle-aged man talking to a middle-aged woman about a new job in Seattle. There's the construct. You give that to ChatGPT, and you say, “Write me a story.” It will create a story around the construct that you did not come up with. That is artificial intelligence.
What would that be in a mortgage construct in the sense? What would an example of that be? A document recognition system that never saw this document before. Most of them would fail and say, “I don't know what this is.” If the system is smart enough to look at its massive array of data, we will talk about big data more later in the conversation as well because that's important to where artificial intelligence is going. If it looks across that whole data repository and now can think up what this document is that you gave it the first time based on all 500,000 documents that it's seen in the past and it comes up with a name, that's self-teaching. That's an artificial intelligence doc solution.
When someone is trying to put this into the mortgage industry, give us another example other than docs. I'm thinking of sales and areas like lead generation. Has anything come to mind?
I will come back to that one, but another example that I like to use on the servicing side of a mortgage has to do with voice recognition. You can put artificial intelligence on a call center solution and detect and learn what the tenor or the manner of the caller is when they call in and route the calls better. Why is it artificial intelligence?
Gabe called this time, and he's in the call center for the first time. You didn't ever program anything of what Gabe is. Is he mad? Is he not mad? The system listens to the voice, looks at the waves of the voice, and determines that Gabe is upset. It's easy to determine that. It routes the call to a special group that handles upset borrowers because you want to give them more attention than somebody happy and call for a payoff balance on the servicing side. That's an example.
There are other examples that you could look at with ChatGPT about lead management. This is one of the more exciting things, like where you could go with your data from a Salesforce solution to set up a construct for ChatGPT to come up with your script to read to the prospect. You get your Salesforce transformational mortgage solutions or mortgage connect.
You look in your Salesforce, and you find this prospect. They are at 10% of the funnel. They haven't developed much yet. You get the company name, the geography of where they are, and these ten things. You give that to ChatGPT in the future, and you say, “Write me a script of how to introduce my product to this person.” ChatGPT came up with the script.
With integration to language packages that are already happening, they could say the script. You don't even have to call and read it. You could have a plug into ChatGPT that sounds like a woman, a man, an older person, or a younger person and have them call your pipeline and read a context-based message with the background of the person or whatever it is you put in the construct that gets the relationship going and doing all of this.
I'm ideating and creating something based on what you said, which is almost a little of artificial intelligence. What I'm thinking about is Slydial. It is a service where you can get past someone's phone. In other words, if I were to dial you, it wouldn't let you answer, but it goes right to your voicemail. What you could do is if I had a profile on Gabe and I never met you, I could put that in the computer or into ChatGPT, have it suggest that to use a natural voice, and have it drop a voicemail message that would be almost perfect. Is that correct?
Yeah. In ChatGPT technology or OpenAI, you can use it to come up with what is said, use a natural language package to make it sound like a person that doesn't sound like a bot, use Salesforce to seed the information interesting on how to build out the to-do pipeline for this automation and have everybody call.
It's amazing where all of this can go. I saw an example. I was looking at it with our legal department and what it had to do with putting together legal documents using artificial intelligence. You used OpenAI. It looked at the document that you had drafted or an MSA at Mortgage Connect or Transformational Mortgage Solutions. It looks at the sections of all of the documents within the legal library of documents and comes up with a more natural way to explain the document to the new company you are trying to sign up for.
The applications start getting exciting. It's limited to your ability to apply this and start experimenting with it. You bring up attorneys, which gets us to an interesting element about this. It is the regulatory environment regarding AI. We have already seen where AI has crossed some lines unintentionally. Talk about the regulatory environment.
I am not a regulatory expert like Alice or the others that you have frequently on the show. We have been paying attention to this a little bit because of interest. The Biden administration put out the blueprint for AI already some months ago. The Department of Commerce issued an RFC request for comment. The government is trying to say, “This is going fast in terms of where it's developing. We should be thoughtful about how we do this.” Being thoughtful, in my mind, has to do a lot with the accountability of the process that's running the artificial intelligence.
If you are going to use artificial intelligence to help you with hiring people, that's not necessarily right or wrong. If the artificial intelligence is self-taught and generates its own bias that your equal opportunity employment does not line up with and makes hiring decisions on your behalf, that's not okay. That's causing a lot of people to ask the question, “How far is too far? How do you set boundaries? What should those be? Who sets them? Is it the government? Is it the police? Is it a new AI group that doesn't exist yet? Is it the universities?” There is no governance yet around artificial intelligence. That's something that is a hot topic. I believe it's going to be a hot topic for years to come as this continues to advance.
Artificial Intelligence: If artificial intelligence can generate its own bias and is not aligned with your team's values, do not use it to hire people.As it continues to develop, there's no question that this enters into some of the biggest risks that we, as lenders, have. You watched the number of some of the top banks. Wells Fargo pulled out of the mortgage industry for the most part because of the regulatory fines they paid. The bigger the organization, the higher the risk. Gabe, when you look at the regulatory environment, you have already identified some things. There are some potential problems. Is that a reason to stay out of using ChatGPT, AI, RPA, ML, or any of these?
No, I don't think that any of these are a reason not to use them. I do think that these are all reasons to be thoughtful about how you use them.
Give us some guidance on how to be thoughtful. Give us a little insight and dive a little deeper into that.
Don't get artificial intelligence to do something that you wouldn't do.
[bctt tweet="Do not let artificial intelligence do something you would not do." username=""]
What if you tripped over the line you are doing, and it's starting to do it without you knowing it? For example, approving a particular group of loans or declining a group of loans.
Hire the right safeguards so that you catch and stop it. If you had a human being making loans to the wrong people, you have to catch that.
That's the point I wanted to get to. I'm asking questions somewhat on a rhetorical basis in my mind because you and I have talked about this enough, but there are practical things you can put in place. Stop caps that allow you to not get yourself in trouble as an organization. That thing is the most important. You and I have talked about this over many a dinner and opportunities of getting to gather. A lot of people look at the future. They look at what could be done and they either choose to move forward into it carelessly or they say, “It's too much risk, and we are staying away.” That's a ditch on both sides of this road. It can feel like this road is narrow, but the reality is it's a road that we all must get on and travel quickly, or we could find ourselves left behind. Would you agree with that statement?
I believe that innovation is going to happen regardless of what David and Gabe do. Put a different way. Another question might be, “What's game-changing with regard to artificial intelligence?” You say, “Gabe, it's been around for many years. What is different then?” What's different also ties to the point that you made. That's why I'm correlating here.
What's different is that there are many more people who have access to artificial intelligence than ever before. Everybody with a cell phone has access. Not only that, they have access to more data than ever before. It used to be that you had to have $100 million to go buy thinking machines and concurrent processing systems in the 1990s and hire a team of fifteen PhDs to write a neural network to do artificial intelligence. Now, all you need is an iPhone, and you can access ChatGPT and other AI projects. There are going to be more to come.
With 100,000 people and 1 million people that have access, 1 of them is going to be thinking about the next way of building the mousetrap, communicating about a mortgage, closing a deal, and plugging this together with Salesforce because there are 10 million that can have access to think about it. You don't have to go to a research firm to have access to the technology to come up with the solution.
When you talk about practical uses, we have already touched on a couple. Are there any other practical uses of AI in the mortgage lending environment?
For the main ones, we are using it for doc recognition and taking that to the next level at Mortgage Connect. We like to sign documents digitally, take our customers digitally, and progress down the digital landscape, which means getting out of paper across the board, electronic signatures, and electronic documents.
Artificial Intelligence: AI is mainly used in the mortgage industry for document recognition. It can help sign and sort out documents digitally.
If you haven't seen the document before, you have got to figure out what it is, classify it correctly, and put it in the right bucket and widget to move it down the process flow. We are using artificial intelligence and cloud and big data to do better document recognition. We are looking at voice solutions. We talked about that. That's more popular on the servicing and call center sides of the house, where they have lots of calls that are going in and need to do predictive analytics on the calls.
We talked about the example that I saw where they were utilizing ChatGPT in a legal context to help author and re-author next-generation MSAs and SOWs from your legal department or NDAs and being able to word them correctly and in a more natural context that's taking advantage of that natural language processing.
You can go from there to decisioning. Decisioning is where it gets a little bit more dangerous depending on what decision you are having the program make. You have to put guardrails around it like we talked about. You can have a system that learns to make its own decisions and apply that to a credit process or a mortgage process.
The skies are the limits, but within constraints and thoughtfulness about how you are using the technology and stepping, start with a minimum viable product and move it along where you can build data of all the hundreds of thousands of correct decisions that are being made. Even though they are self-teaching what's happening, you have to watch, monitor, and be close to what it's doing.
A lot of companies still don't even know how to manage their RPA, which is the box that is doing repeatable tasks. They are not even trying to decide, learn, and come up with their way. Monitoring is one of the most important things to think about when you are trying to deploy RPA solutions. AI takes that to a whole new level.
When we did the previous interview back in 2019, it seemed new. I read it while preparing for this interview. It seems as new as if we could have just done it. The only thing we weren't using was the language ChatGPT and OpenAI. That's the new element that's in there. Is there a way in which you can combine and use RPA and machine learning and put AI on top of that? Are we seeing an integration of all these various elements? Explain what that could do for a mortgage company.
We are because the RPA companies are constantly adding and innovating their solutions. They are combining machine learning and artificial intelligence into the RPA. The RPA box that you put in to solve a specific repeatable problem, if you turn on artificial intelligence, you can turn it on by having the bot recognize a document and do something with the document. That's using the artificial intelligence of the doc recognition on top of. You can also use artificial intelligence and start to do things like, “I'd like you to figure out the process. Here's a system. Watch what happens on the system for an hour and record that somewhere. Recommend to me what the process steps should be so that I can eliminate waste from the process.”
You are getting into business process re-engineering. I had never even thought about this, which was getting into the area I wanted to go. It is the practical application of ChatGPT to the mortgage industry. One of the things that are badly needed as a consultant and as someone who has been in the industry for many years, I see all these processes. We have been doing it the same way for decades and we don't have the innovation. It's because part of the time, we are busy, we don't have time to innovate, or we are in a contraction mode where we have to let go of all the people and extra people and our ability to look at how we could do things differently. As a result, business process improvement has suffered. You are saying ChatGPT or the whole AI, broadly speaking, can start going in and improving that. That is exciting.
I wouldn't profess myself to be a ChatGPT expert, but from what I understand, looked at, and played around with it, it's a natural language, artificial intelligence. They are updating and innovating to do a lot more things. That's what it's focused on. It by itself is not necessarily going to re-engineer a process for you yet. We will see where it is in 1 or 2 years from now. However, it will do natural language processing for you at a step in the process.
Let's say you get a loss mitigation process, which is it's not easy. You look at that, and you see all the contact points for the borrowers. You look at the scripts of what you developed, to your point, that's twenty years old, and you feed that to ChatGPT, and you say, “Give me a better way to communicate this to the borrower.” It will do that for you.
It goes back to what I love. When we had our last mortgage company in Southern California, we had an inbound call center. We used some technology that wasn't nearly as sophisticated to route phone calls to someone of like personality and increased the success of that inbound call experience for the inbound caller because they were talking to someone of a similar personality. Those are the things that we are going to expect to see. How is looking forward? Give us some insights into where this is going and the applications that we can anticipate in the months and years ahead. Things are moving quickly. Months will give you enough of a challenge.
They are constantly innovating ChatGPT because it has so much exposure, and many people are energized by what it does. They are using it for more tasks. Write me a college paper, admission test, or Salesforce script. While that's going on to innovate the uses of the system itself, it's building its data. It hears every single thing, and it knows every single thing that it's ever written. I can use that as the example for the next thing that you ask it to write.
Artificial Intelligence: ChatGPT is building its data. It hears every single thing you say and knows every single thing it has ever written.
It compounds over time. That's why the big data cloud, Tesseract document recognition, is a game changer. It's a game changer because it's accessible. You can subscribe and get it. It used to be $1 million to buy a doc recognition system. You recall this. You have to go to big providers to buy them and pay a lot of money. Now, you can subscribe to the cloud, and you have the solution.
The other game changer is that it's not your docs that it's looking at. It's looking at 100 million docs, all of them, and training off of all of them. Yours is only 100,000 docs if you are a large lender. You plug that in, and you have access to all of that data. You can move the solutions a lot faster because the data's compounding into the solution for you.
Where could it go from here? The sky is the limit. There's the innovation you talked about. You and I have gotten together at the dinner table in Washington and other places. Everybody is going to be getting around the table. They are going to be innovating with these new technologies that they have access to. I see lots of language applications, document applications, and progression towards digital applications that are going to happen in the mortgage industry as a result.
One of the things it is not going to do is replace people. It may replace processes that certain people do, but it will not replace people.
It's an interesting thing. I would agree with you because if you and I are sitting together, we talk about as management teams, we talk about, “Go back to the office and don't go back to the office. Work remotely.” It all relates to interacting and being social, which most human beings are. There was an article done that, with ChatGPT, somebody is using it to do voice voiceovers and doing the voiceover itself. A firm could be like, “I want a voiceover for this animated mascot that's running around, and it's selling my product.” Instead of going to a person, they are now going to ChatGPT and this partner to do the voicing of the mascot.
The article was like, “Is this killing an industry?” I don't know if it is or not. It's going to reduce that industry because if it's easier to do it the other way, it's less expensive, you don't have to schedule somebody, it's there all the time, and you go, it's going to provide benefit. That benefit is going to have to be weighed against getting that personal touch.
You have talked about this in many of your episodes. At the end of the day, mortgages are relationship-driven. Not only are they consumer relationship-driven, but they are also business-to-business relationship-driven. It's going to take a lot of software to go in, mess around, and replace relationships. It's hard to do.
[bctt tweet="Mortgages are relationship-driven. It will take a lot of software to mess around and replace relationships." username=""]
I love studying brain science. One of my favorite books is Rare Leadership. It talks about the limbic side of the brain. Simon Sinek talks about that Start With Why. I look at what ChatGPT can do. I look at the power of certain solutions, even with natural language programming, but it does not have the ability to have an emotional component. Some would say, “That's not true because it has been able to deceive others in chat conversations.” RPA can do that now if they have enough data on that. I love brain science. I'm studying it. I love looking at our logic center and how we process information, especially as you look at AI and look at what the opportunities are. Where can we better use AI to partner with the part of our logical brain? There are other parts of our brain, like the limbic system, which has known language.That's where we come to say, “I can't tell you why, but I have this hunch. We should or should not do this. I have this general feeling. My gut is telling me.” All those non-verbal signals, which we try to talk from. A lot of that comes out of our limbic brain system. How do you see AI partnering with us, and where does that partnership fall apart as it relates to the mortgage business?
Where AI is now, AI doesn't have much emotion. I would say emotion is learned. We learn emotion as kids through our lives, through getting married, through having kids. That's the family side of emotion, which is deep for most people through religion and emotion, through all of the different things that you can learn emotion from. Anything that is learned can be taught.
I would agree with you. I would never say that can't be done anymore. I don't think that there's a stopping of this. It's a question of how long until the next chatbot has emotion. You have 100 million people working on the problem now. Somewhere in there, somebody is going to put the pieces together to do that innovation faster. In this case, you will have an emotional, chatable, and natural language bot. You can direct it to new tasks to handle emotional people, read the emotion out of the people and put it on the right path, or learn all of them there.
Artificial Intelligence: There is no stopping AI from progressing and evolving. It is just a question of how long until the next chatbox acquires emotion.
There are a lot of applications here, whereas we have all of our experiences that we built over time. Look at how many experiences you have all of your whole life. When you say, “I have a hunch. My gut tells me that this is going to be like this.” Not always, but more often than not, it's some experience or previous combination of experiences that's driving the hunch.
If you could take all of your experiences and all of my experiences, the two of us, and put that into a database and make that database aware of artificial intelligence, they don't have to learn that and wait 50 or 30 years to gain all the experience anymore. They cut right to the chase. They have all of that experience right now to make it. It's not you and me. It's across 50,000 people's experience because the data is that deep.
That's where we are headed with this. It will have more applications. Is it going to replace a relationship? No. Will a bot be able to create a relationship with another bot or create and fool a human on the phone because they are not humans walking around? Yes. In the future, they are going to be able to do that. It's coming. It's a question of when.
This is such a deep topic, and it's evolving constantly. I can't thank you enough for coming on here and talking about this game-changing technology that's going on. We thought we had game-changing stuff going on back years ago when we were talking. It's moving at an extraordinary rate. As we wrap this up, what would you say our readers need to keep in mind as they are contemplating how they and their companies should use artificial intelligence or all three of them, whether it be RPA, machine learning, or AI?
Glad to be here. I always love being on your show and talking with you about all the topics. I look forward to where we go next. That being said, I think that you have hit on some important takeaways, and we have together hit on some important takeaways. Don't be scared by the technology. Embrace it. It's going to happen whether you embrace it or not, anyway.
If you are not embracing AI, the mortgage banker down the street is. They, the lenders, and GSCs are already asking about it. It's coming. It's the storm, and you can see it. The clouds are rolling towards you. You might as well get the umbrella out and walk into the rain. That's a big point. Don't overuse and overpower it. Be thoughtful about what you are doing. Don't do something and have something do something that you wouldn't do.
For an important one, across all these RPA, ML, and AI, make sure you are monitoring it like an unruly employee. You monitor them or the kid who graduated college and doesn't know. You are empowering them to do something in your office. You are putting belts and suspenders around them. You have to put belt and suspenders around any process that's making decisions for you.
In the end, you are accountable for the decision, not the AI. The AI doesn't go to a court of the United States and go on trial. A person who empowered the AI will go to court for the decision that the AI made. You have to keep that all in mind and be responsible and accountable, but don't be afraid. Be open and embracing.
[bctt tweet="You are accountable for the decisions artificial intelligence makes, not the software. " username=""]
AI is one of those examples. I love that you all recommend books. I recommend them to my leadership teams as well. The Power of Habit, and the list goes on. You should be reading about this because it is going to continue to innovate what they are doing with ChatGPT. That's not ChatGPT, as we thought. It is an application of artificial intelligence. Artificial intelligence itself is way bigger than that. That's a chatbot. You have to watch that. Don't lose the forest for the trees.
Look at the broader picture. There's so much wisdom here. How could people get ahold of you if they want to reach out to you to have you speak at events or ping you for your knowledge?
I work at Mortgage Connect. We are a company that's headquartered in Pittsburgh, Pennsylvania. We have offices around the nation. We are a services provider. My email address is GMinton@MortgageConnectLP.com, and www.MortgageConnectLP.com is our website. You certainly can reach out to me there if you have questions or if you would like to have me talk to you about anything on a go-forward basis. Most of the technologies that we talked about we are using in some form or fashion or researching the use of at Mortgage Connect because we are innovative and want to push the envelope as you do at Transformational Mortgage Solutions. That's the best way to reach out to me.
Thanks so much for your time. I appreciate it.
Thank you very much. It’s great to be here. You have a great day.
