Andrew Wang joins us now, the CEO and co-founder of Valen, a mortgage servicing company and a major player in the world of software platform.
Nice to see you here today.
Thank you for having me.
Thanks for taking some time for us to talk to me about why AI adoption inside a big financial institutions seems to be moving.
As slowly as it is, so I think there's a couple of really large components.
Um, one big component of it is ultimately security and compliance.
There's a big amount of, uh, risk and worries that a lot of large financial institutions have when it comes to adopting AI and making sure that, you know, it's keeping their information.
Uh, private and, uh, it's holding it in the right places.
Then there's, I think, a different component which is there's just a much higher degree of scrutiny around regulated services, right?
If you use AI and it says, hey, I'm gonna foreclose upon this home, probably doesn't really check out, doesn't work, uh, versus if you had deterministic rules, which is how it used to work, then you have.
Some level of a guardrail and psychological safety.
And so I think the evolution of regulated services when it comes to AI just will take a very different and probably slower path than what we're seeing in, for example, engineering and coding and things like that.
That's a fascinating distinction.
What would you say are among some of the largest misconceptions about AI adoption right now?
I think the biggest misconception today on the tech side is just how fast they'll move.
I think it'll take a lot more change management and teaching people how to use the underlying product because so much of it ultimately is, you know, a matter of skill and understanding and really just changing.
How you work as opposed to just automating specific processes you used to do I think that's a really really big component I think uh another misconception uh on the financial services side and really the old school side it's just how much can be done right? there's a continued belief that.
AI, while it's really powerful and great, it's still just chat GPT, not cloud code or whatever else that people use today.
It's way more powerful than they imagined.
I think those two worlds can be true at the same time, but you know they need to meet in the middle.
And I've heard a lot of people have this conversation, maybe ask this question.
I wonder how fair it is.
Am I training my own replacement?
How legitimate, how fair is that concern from people in the space?
I think it's fair if you think you're never going to.
Change what you do, right?
The perfect example here is always, hey, if I used to, you know, be someone who does typewriting and I, you know, make mistakes here and there, but I've trained myself to be a very good typewriter so I can write these things with high proficiency and high speed.
Well, now it turns out that with computers and with, you know, voice AI transcription, you can write a lot faster.
Does that mean you're training your replacement?
Yes, in the sense that your job, you know, literally. what you were doing before has changed, but no, in the sense that perhaps your job actually was to take the messaging, the thoughts of whoever you're working with, and communicate more clearly.
So it really just depends on how you view your job.
And if you think about the evolution of that, then you're never really going to lose that.
Why is mortgage servicing such a difficult test for where artificial intelligence is right now?
So when it comes to AI and a lot of the underlying technology, it's been trained on the Vast, you know, corpus of public knowledge and information.
Mortgage servicing is a perfect example of where there's a good amount of public knowledge, but there's a tremendous amount of in-house tribal industry knowledge which is just not available in the public domain.
There's a lot of it that is, you know, written down in the lores of how you do mortgage servicing, the crisis, whatever else, and that is a very hard thing for AI to interpret and a very hard thing for AI to extract.
It is a perfect example that where if you've built all of that context in that domain, you can automate a whole lot more, but nobody's built that repertoire and corpus of knowledge.
Andrew Wang, CEO, co-founder of Valent, it's really nice to see you.
Thanks a lot for taking the time.
Awesome interview.
Thank you very much.
