Your Smartest Intern: ChatGPT

How To Use ChatGPT

The more time spent playing with the technology, the better people will become at using it. Shekhar says that it’s a good idea to learn some prompts to ensure commands or questions are understandable, and to get more accurate responses.

Here is a prompt provided in the ebook that works for marketing emails: “Write an email to a prospect for using me as a loan officer to get a mortgage for buying a home in Las Vegas, Nev., using the four Ps persuasive framework — promise, picture, proof, and push.”

ChatGPT then spit out an email Shekhar could copy, paste and send to his prospect, or he could make edits and craft it to reflect his voice.

For those who struggle with using social media to market themselves effectively, ChatGPT can be a helpful guide. In Shekhar’s ebook, he says it can create a content calendar to help plan what works best for you and writes the content. If you already have a written draft, ChatGPT can revise it to a specific tone, even adding a touch of humor. For videos, it can suggest just the title and the description, or it can write an entire script.

The most useful way to use ChatGPT is by having it help build relationships with referral sources, especially real estate agents.

Mortgage veterans will say the best way to get your foot in the door is by providing these Realtors value. Shekhar says this is typically accomplished by solving their problems, or by helping them create more opportunities and close more business. ChatGPT can do both.

Ask it to write listing descriptions for a three-bedroom, one-and-a-half bathroom house set in a quiet cul-de-sac. Obviously, the more detail added, the better the listing description will be. Then, ask it to shorten the description for flyers or social media posts.

In Shekhar’s book he says that the use cases for real estate agents far exceed that of a loan officer. If a loan officer were to teach classes on this topic to local real estate boards or officers, they would be building relationships with plenty of referral sources. Shekhar used the same strategy to build referral sources when he got started in the mortgage industry around 2008.

“I started teaching Realtors on using social media and video marketing in 2009 when it was a very new technology,” Shekhar says. “Very few people were using it so I could get in front of hundreds of real estate agents … and I think the same opportunity is there for loan officers now with ChatGPT.”

ChatGPT can also calculate mortgage payments and even run amortization scenarios and pre-qualifications for loan officers’ clients. Here is a prompt that Shekhar used: “What is the mortgage payment on a $425K loan at 6% for a 30-year fixed loan?”

ChatGPT’s response: “The monthly mortgage payment on a $425K loan at 6% interest for a 30-year fixed loan would be approximately $2,548.90.”

Limitations & Possibilities

Although this is an advanced supercomputer machine, ChatGPT can make mistakes from time to time. Shekhar says it can calculate basic pre-qualification when provided the debt-to-income ratio, even for more complicated pre-qualifications, but double-checking answers is recommended.

“I have tested it on a few numbers where it’s not too far off the correct answer, but it’s not the correct answer,” Shekhar says. “And so those are the things that we need to wait on. And we need to be giving it a thumbs down. That’s one of the ways ChatGPT learns is that if you give it a thumbs up, it knows that it’s the right answer. If you give it a thumbs down, then it knows it’s the wrong answer and maybe tries to figure it out by itself. Maybe you can teach it by saying, ‘OK, this is how you calculate debt-to-income ratio. How you’re calculating it is wrong.’”

Most of the cases that Shekhar provides in the ebook revolve around marketing, education, and content creation. But as newer versions of ChatGPT come out and improvements are made, it has the potential to make more accurate calculations and perform more complicated tasks.

In Shekhar’s ebook he notes how quickly the internet became obsessed with ChatGPT. An infographic shows that Netflix took 3 ½  years to reach a million users, Twitter took 2 years, Spotify took 5 months, and ChatGPT took only 5 days. Newer versions are already being made too. It was originally launched in November 2022 as GPT-3.5, then GPT-4 was launched March 14.

However, it will take time to enable ChatGPT to perform more advanced mortgage-related tasks, such as risk management, but Shekhar says it is theoretically possible. Unlike artificially intelligent platforms, ChatGPT uses machine learning technology so it can improve on tasks the more it is used. It does have the potential to detect fraud if it was trained to do so. Shekhar provides an easy-to-understand example of how this works:

“Let’s say that you want a platform to understand or be able to recognize what a cat looks like,” Shekhar says. “Now that machine will be fed hundreds and thousands of pictures of a cat and will be told what the basic features of what a cat is. They have four legs, whiskers, they’re small, and so on. ChatGPT will be shown a lot of other pictures that may look like a cat, but it’s not a cat. Maybe a baby tiger or lion or whatever. And then after that, it comes to a point where the machine has learned on its own and now it’s able to figure out which one is a cat and which one is not.”

In the mortgage world it can do the same to detect fraud.

“You can use the same machine learning to teach a platform like ChatGPT to understand, for example, which one is the correct pay stub and which one is not, or which one is the right W-2 and which one is not,” Shekhar continued. “And through learning millions of intakes, it can help with risk management for lenders where it’s able to figure out, ‘This pay stub is not correct, it’s a fraud.’”

Is it competent enough to replace experienced staff members such as underwriters, loan officers, processors, or risk managers? Shekhar says “no,” because it can make mistakes. It requires oversight on the tasks it performs, like an intern would. ChatGPT is not necessarily your best employee, but it can do some things for you and make completing your daily tasks easier.

A Competent Intern

Time is another currency that loan officers, lenders, and other mortgage professionals can never get enough of. Broker owners who don’t have a large support system around them or need to scale back on staff in a down market need help handling marketing, client management, and relations with referral sources. Interns are a great solution, but it’s also likely they would want to be paid more than $20 a month for assisting with all these tasks. They’re also not available 24/7 and aren’t talented in all of these areas.

“One of the challenges that I have always faced is the fact that you can’t find one person with all the talent,” Shekhar says. “If you’re looking for somebody who can write video scripts, write blogs, write social media content, research for me to figure out which content will actually rank and do better …  you are probably looking at three, four different interns because not all of them have the talent to be able to do all of this.”

However, having ChatGPT do the initial heavy lifting makes it easier for one person to complete all of these tasks.

“I mean, 50%, 60%, 70% of the work is done for you,” Shekhar says. “Now you can have an intern who can possibly then go and finish the rest of the task, or you can do it if you’re a one person shop.”

Users can measure the success of ChatGPT for their business the same way they normally would. For example, if the open rate on a marketing email was 15%, and after using ChatGPT it ticked up to 20%, then it’s working. You can measure success the same way with tracking likes on social media posts or visits to a blog web page.

Shekhar’s Gift To The Industry

Shekhar’s ebook ChatGPT For Lenders is free. Downloads are available at chatgptforlenders.com. Plug in your email, phone number, and NMLS ID and a free copy will be sent to you.

“It’s just my way of giving back or paying forward to the industry that has helped me quite a lot over the last decade,” Shekhar says.