CI&T

Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.

Mark Hansen (00:08):
Thank you everybody. My name is Mark Hansen. With me here is Leslie Chang. I run the product team for CMG Financial and Leslie runs our AI initiatives team. So you have product people and you have technology people. When we work together, sometimes good things happen. We'll find out today. This is a little bit different type of a demo because we are CMG, so we are a lender and we're not necessarily selling a product for a hundred dollars a month to you. We're showing you an example of what we're doing and the innovation that we have and that we're building in our studio alongside some of our partners, CI&T, who's here with us. Tim Vino helped us be successful with this project we're going to show you. CMG is AI forward. We are fully committed to building technology that uses all the possible models and we like to experiment.

(01:08):
We like to test things out. We have a laboratory where we're experimenting, taking product, and working in conjunction with our loan officers, with our processors, and with our underwriters to try to come up with best-case scenarios. One of the things that we've been doing in the last couple of months—if you can go to that screen—we have a CMG AI assistant desk. We're not going to go into that. This isn't the demo, this is just to show you some of the things we've been able to see. Leslie, you may need to kill your other PowerPoint window. We want to give loan officers the ability to have different chat opportunities. So we've built a bunch of technology in-house around a loan officer or processor being able to understand guidelines, upload bank statements, upload income documentation, and get results back.

(02:04):
I think CMG has something like 3,000 custom products. It's one of our specialties and it's impossible to know everything and how it all works. So having these different chatbots allows us to assist our loan officers in understanding what those capabilities are. The problem with that is you might think, "oh, that's just ChatGPT." So we wanted to show you something a little bit different. We're going to do a scenario today. I'm going to be the loan officer, Mark Hansen, licensed in 50 states, and I'm driving down this country road. Leslie is my borrower and I think she's going to call me with some good news.

Leslie Chang (02:43):
Hey Mark.

Mark Hansen (02:45):
Hey Leslie, how are you doing?

Leslie Chang (02:46):
I'm good, how are you?

Mark Hansen (02:47):
I'm great. I'm just driving down this road. I am almost hands-free, so I hope I don't crash. What's going on?

Leslie Chang (02:54):
Good. So it's safe to talk right now?

Mark Hansen (02:55):
It is. Go ahead.

Leslie Chang (02:56):
Awesome. Well I found a solution. My aunt is going to lend me another 100K for my loan, so I'd love to reduce my loan amount.

Mark Hansen (03:06):
Oh nice. Okay, so you have an extra hundred thousand dollars. Like we said, I'm driving right now and not in the office, so let me see what I can do. Check your inbox in a little bit and I'll let you know. Okay?

Leslie Chang (03:16):
Sounds good. Thank you.

Mark Hansen (03:19):
Alright, so now what I'm going to do is log into my phone and I'm using Teams. I'm not using any special software; this is just Teams. What we've done is we've created an AI bot that sits as an agent and has access to my LOS. I'm driving, so I'm going to do this hands-free. "Hi, can you please give me the details of the Leslie Chang file?" Now, as others have mentioned, with the internet, there is a slight delay. Keep in mind we have to open up the LOS, pull back pipeline data, do a verification, extract the data, and then bring it back through AI and present it in a readable fashion. So bear with us while this processes, but it's live so we can expect that. In the meantime, enjoy the scenery. Alright, so here are the loan details. We have loan number 1234. Property address is TBD. And it looks like she has a loan amount of 650,000. Let's go ahead and change that. That's one of the values: not only are we creating AI to access the data—that's old school, we just saw that on the last screen—I want to modify the data. "Great. Can you please change the loan amount to 550,000?" I spelled it wrong, but it's okay.

(04:51):
And again, I didn't have to be specific with what I asked; AI interpreted that for me. "Would you like to send an email confirmation after?" "Yes, that would be great. Please proceed." So just to confirm, you would like to process changing the loan amount to 550,000 and sending a notification to the borrower? Yes, I would. Thank you. Because we're changing loan data, it's important to have a little bit of extra checking, so we're just making sure. We do have extra checks just to ensure that it is the change that we want to make. Let's go back; we just want to wait for that confirmation.

(05:50):
For our AI models, we're using Anthropic, Gemini, and OpenAI. We use all three models for different purposes. You're seeing a combination of those in action depending on how we're extracting the data, what we're doing with the prompting, and how we're actually returning the information. We will sometimes run parallel models and compare them together in order to get what we need. Just waiting for it to finish up here. The loan amount for Leslie Chang has been successfully updated to 550,000 and a confirmation email has been sent. I'm done. I'm going to go enjoy my country drive and hopefully Leslie will get an email.

Leslie Chang (06:35):
I'm just going to check my email account and I have a new email from my loan officer. It says that my new loan amount went from 650,000 to 550,000 and I'm saving $287.89 per month with the new terms of my loan.

Mark Hansen (06:54):
That's our demo. With a few seconds left on the clock, we are available to have conversations. We were invited graciously by the CI&T team to stand at the booth. If anybody wants to talk to us about what we're doing at CMG or how we're using CI&T or other technologies, we're more than happy to have conversations with you guys for innovation. We live and breathe this stuff every day. Cloud code is in my brain here and it's all we use. Really excited to be here and for the opportunity to share. Leslie, any closing comments?

Leslie Chang (07:35):
No, thank you for having us. Mark and I love nerding out about AI, so we would love to come talk to you.

Mark Hansen (07:41):
Thank you. Awesome.