OpenNashPodcasts

AI in Financial Services

Measuring the ROI of AI - with Jay Budzik of Zest AI

2 Jan 2023 · 22m · Daniel Faggella, Jay Budzik · 19 quoted facts

▶ Play episodeOriginal source ↗

In brief

Smaller lenders such as credit unions and regional banks can compete with big banks by using AI, but only if the hard parts are done for them. Jay Budzik of Zest AI argues that what scales is a repeatable process for building models, not one pre-trained model. That process is paired with pre-purchased third-party data and a guided, automated tool.

The tension is between vendor promises of push-button simplicity and the real work of adopting AI. Host Daniel Faggella keeps pushing back on the MailChimp comparison, noting that model building still involves research and tuning. Budzik concedes the user must still monitor results, but says automation lets business users, not data scientists, run the process.]

Key insights

  1. AI is a capability, not a single project

    Budzik says larger banks have spent years building a whole framework of data and analytics. Smaller lenders have mostly bought finished scores from vendors like FICO and Vantage.

  2. What scales is the process, not the model

    Zest AI reuses a careful model-building process learned from large banks, rather than a model trained on one bank. Faggella notes that moving one bank's data to another would raise ethical and legal problems.

  3. Third-party data fills small lenders' data gaps

    Small lenders often lack enough records for robust models. Zest AI maps which data sources predict well for each lending product, buys them in advance, and blends them with the lender's own data.

  4. National credit scores fit local populations poorly

    Budzik argues a score trained on a national population may not suit a credit union in Iowa or Florida. Lenders can tailor a model to their members and the segments they want to reach.

  5. Platform and vertical AI vendors now interoperate

    Because most models are built in Python, customers can train models on a general platform and then use Zest AI's tools to check compliance, improve fairness, report and monitor.

Questions

How can a small credit union use AI without a big data science team?

Budzik says Zest AI interviews the lender about its products and goals, sources suitable data, and builds the model automatically. This replaces the 100-person data science team some large banks have.

Why can't a vendor just train an AI on a big bank and reuse it everywhere?

Faggella says that raises ethical and legal issues and the lessons transfer poorly. Budzik adds that populations differ, so a model trained on one group may not fit another.

What if a small lender doesn't have enough of its own data?

Budzik says the lender's data is combined with other data to form a larger, more representative sample. That sample can include people who applied but were denied, or whom the lender hopes to reach later.

Is Zest AI's software really as simple as MailChimp?

Budzik compares its setup wizard and monitoring dashboards to MailChimp. Faggella disagrees, saying the underlying research and tuning are more complex than checking email open rates.

How does Zest AI work alongside other AI platforms?

Budzik says many customers train models on a general AI platform and then use Zest AI's software to check compliance, improve fairness, produce reports and monitor the model. This works because most models are built in Python.

Facts

Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.

Zest AI helps lenders, from large institutions to credit unions, train algorithms that detect risk and set loan approvals or terms.

“Zest helps to train algorithms to detect risk and determine approvals or terms for different kinds of lending products for all kinds of lenders, from gigantic financial institutions down to credit unions.”

0:01:23 · Daniel Faggella · Lending AI

Jay Budzik, CTO of Zest AI, had appeared on the podcast about three years before this episode.

“Jay was a guest on the show some three long years ago, so it's a pleasure to be able to have him back.”

0:00:32 · Daniel Faggella · Guest history

Zest AI was previously called ZestFinance.

“a company that was once called Zest Finance.”

0:00:32 · Daniel Faggella · Company history

Jay Budzik says there are tens of thousands of small lenders that need help accessing AI technologies.

“There are tens of thousands of them and they need help accessing these complicated technologies in order to win against those competitors.”

0:03:54 · Jay Budzik · Market size

Jay Budzik says large lenders use AI for marketing and loan origination, which lets them make more competitive offers.

“They're using AI for marketing. They're using AI to originate loans. And so they can make offers that are more competitive often.”

0:03:19 · Jay Budzik · Large bank AI

Jay Budzik says smaller lenders have mostly consumed analytics products from vendors such as FICO and Vantage.

“while the smaller ones have been kind of focused on consuming analytics products from vendors like FICO and Vantage.”

0:04:47 · Jay Budzik · Credit scoring

Jay Budzik says Zest AI has learned which data sources are relevant for different lending segments and products.

“we have the map of all the data sources that are relevant and that produce the performance in Lyft.”

0:06:52 · Jay Budzik · Data sources

Jay Budzik says the best data for marketing can differ from the best data for underwriting.

“So a marketing data source might be different than an underwriting data source”

0:06:52 · Jay Budzik · Data sources

Zest AI pre-purchases and aggregates third-party data and makes it available to customers through its tailored modeling process.

“we can aggregate or pre-purchase lots of data and make that data available to our customers through our tailored modeling process.”

0:07:24 · Jay Budzik · Data strategy

Jay Budzik says some large institutions employ data science teams of about 100 people to build and validate models.

“having a 100-person data science team like they do at some of the institutions you mentioned in order to build models and validate them”

0:09:49 · Jay Budzik · Data science teams

Jay Budzik says small lenders often lack enough observations to build robust AI models on their own data.

“They don't often have enough observations or data to be able to make robust AI models.”

0:11:29 · Jay Budzik · Data limits

Jay Budzik says the FICO score is trained on a national population.

“if you think about FICO, it's trained on a national population.”

0:13:36 · Jay Budzik · Credit scoring

Zest AI offers a wizard-like setup process for building a first lending model, which Jay Budzik compares to MailChimp.

“we've developed this wizard-like process, much like a MailChimp, where you can go and set up your first lending model.”

0:15:14 · Jay Budzik · Product design

Zest AI offers an API so larger customers can use its automation inside their own data science environments.

“We also offer an API, much like MailChimp does, for the larger customers to be able to access the power of the automation within their own data science environment.”

0:15:41 · Jay Budzik · Product design

Zest AI's model management system has three applications named build, adopt and operate.

“our model management system, which contains applications build, adopt, and operate.”

0:17:02 · Jay Budzik · Product design

Jay Budzik says most AI models are built in Python, which allows different systems to interoperate.

“most of these models are built in Python. And that provides a way of interoperating between various systems.”

0:19:06 · Jay Budzik · Interoperability

Jay Budzik says many Zest AI customers train models on another platform and then use Zest AI software to analyze them for compliance.

“we have many customers that train their model using a platform company and then use our software to analyze the model for compliance”

0:19:53 · Jay Budzik · Interoperability

Jay Budzik was scheduled to join the AI in Business podcast for a series on measuring AI return on investment.

“Jay is going to be part of a series on measuring AI ROI in the coming two months.”

0:21:33 · Daniel Faggella · Upcoming series

The AI in Business podcast's previous series on AI culture change began with an episode featuring IBM's chief AI officer.

“our kickoff episode was the chief AI officer at IBM.”

0:21:51 · Daniel Faggella · Podcast series

Chapters

  1. 0:03:19Why small lenders need AI to compete

    Sets up the problem: large banks use AI for marketing and loan origination, so credit unions and regional banks need similar tools.

    “the small lenders need to adopt the tools of those larger players in order to level the playing field”
  2. 0:04:47AI as an organizational capability

    Explains the data advantage of big banks and why small lenders have relied on vendor scores.

    “AI is really a capability. And so you need to think of not just the one AI project”
  3. 0:06:52Transferable process and pre-sourced data

    Budzik explains how Zest AI maps useful data sources and automates model building.

    “we have the map of all the data sources that are relevant”
  4. 0:11:29Blending thin lender data with broader data

    Covers how small lenders with limited records can still get robust models.

    “They don't often have enough observations or data to be able to make robust AI models.”
  5. 0:15:14The MailChimp analogy and its limits

    Compares configurable software to the real research effort underneath, with the host pushing back.

    “we've developed this wizard-like process, much like a MailChimp, where you can go and set up”
  6. 0:17:02Build, adopt, operate

    Describes the end-to-end system that lets business users build, review for compliance, and monitor models.

    “build, adopt, and operate”
  7. 0:19:06Platforms and vertical solutions together

    Shows how general AI platforms and focused vendors interoperate through Python tools.

    “most of these models are built in Python. And that provides a way of interoperating”

Transcript

0:00:04
Daniel Faggella

This is Daniel Fagella, Head of Research at Emerge Artificial Intelligence Research, and you're listening to the AI and Financial Services Podcast. The focus of this week's episode is on making AI more accessible for smaller enterprises. Every vendor company in the world will tell you how push-button simple their solution is, and of course here at Emerge, our job is to push back against that and to get to the truth of what it genuinely looks like to adopt AI and to find our way to ROI.

0:00:32

And as it turns out, many vendors actually are evolving and finding ways to get to value faster with clients based on their experience, based on the scalability of their technology and their processes. And our guest this week is Jay Budzik. He is the CTO of Zest AI, a company that was once called Zest Finance. Jay was a guest on the show some three long years ago, so it's a pleasure to be able to have him back.

0:00:54

This week, he focuses again on AI accessibility. And you'll hear Jay talk about the scalability of process versus the scalability of a single algorithm or some pre-trained system in some way, shape, or form. So I think scalability operates in a different way than many of us might presume from a vendor's perspective. How transferable their lessons are from one bank to another is something that I think will be interesting for people who are tuned in and want to get a sense of how quickly can we get up and running?

0:01:23

Again, a vendor is always going to say what a vendor is always going to say, and my job is going to be a poke into them and to get down to truth. But I think there's a lot of good nuggets shaken out of today's conversation, and I hope that you benefit from it. For those of you unfamiliar, Zest AI is in the lending space. Zest helps to train algorithms to detect risk and determine approvals or terms for different kinds of lending products for all kinds of lenders, from gigantic financial institutions down to credit unions.

0:01:48

So Jay's experience is in lending, but a lot of the insights he shares about the transferability of lessons and the ability to make AI accessible downstream towards smaller enterprises is something that would be relevant basically anywhere in financial services or even in the enterprise. So I hope you benefit a lot from this episode. If you're interested in knowing more AI use cases and finding new ways for AI to be able to streamline or automate some of your own processes, be sure to check out our AI and financial services executive brief.

0:02:14

You can go to emerj.com slash fs1. That's emerj.com slash fs1, fs like financial services, and then just the number one, and you can download our free PDF brief there where you'll not only get a glossary of terms for AI and financial services, but a short list of use cases that should be helpful in terms of being able to find where the low-hanging fruit may lie in your organization. So again, that's emerj.com slash fs1.

0:02:40

Without further ado, let's fly into this episode. This is Jay Budzik. It's great to be able to have him back on the show here on the AI and Financial Services Podcast. So Jay, good having you back with us. It's been a couple of years since you were with us on the program. And today we're going to be talking about something that you've continued to be buried in, which is the world of lending. And we're speaking today about the problems that potentially smaller lenders face when they're trying to win market share and grow, but they're also up against really, really big players.

0:03:13

Talk to us a little bit about what that business problem is and where AI fits in there. We'll set the table with the problem if you don't

0:03:19
Jay Budzik

mind. Yeah, sure. Our credit union and regional bank partners face a challenge. They have to compete with these larger players that are in their markets taking customers. Those large players are using AI to do that. They're using AI for marketing. They're using AI to originate loans. And so they can make offers that are more competitive often. And so in order to grow, the small lenders need to adopt the tools of those larger players in order to level the playing field and allow them to win.

0:03:54

We're focused on this segment because it just represents a huge opportunity for growth. There are tens of thousands of them and they need help accessing these complicated technologies in order to win against those competitors.

0:04:09
Daniel Faggella

Got it. And the competitors certainly have some advantages. I mean, you know, the classic advantages that have been around forever, way bigger marketing budgets, a bigger existing footprint, you know, brand recognition, maybe, et cetera, but also data, right? I mean, you know, a JP Morgan or a Citibank or a, I don't know, Wells Fargo or you name it, they might not all have great data infrastructure for AI, but potentially it's more mature and they have more of it to work with.

0:04:33

And that traditionally has been a bit of an edge too. I mean, it feels like the smaller players maybe aren't able to wield as much data or as well as the big guys. Would you categorize, would you color that differently when you describe kind of

0:04:47
Jay Budzik

the differences and the challenges here? Yeah, I think that's absolutely fair. There are data products out there and data platforms, but AI is really a capability. And so you need to think of not just the one AI project as a unit element and kind of what you can get from that, but what does it look like to sort of holistically adopt a framework of using data and analytics in your business? And so the larger banks have been doing that for years, while the smaller ones have been kind of focused on consuming analytics products from vendors like FICO and Vantage.

0:05:23
Daniel Faggella

Yeah, yeah. Certainly anybody who's in your space is going to be more than familiar with FICO and kind of their approach, which a little bit different than what some of the ML vendors are sort of moving towards. And I guess this brings us into where we're headed with the interview, which is what it looks like to get the edge against the big guys. Part of your work here is taking your experience in lending and making it something that the small folks can do.

0:05:45

And I think some of the barriers other vendors have run up against is, man, we often need a certain amount of data infrastructure compatibility with our product to be able to work here. We need a certain maybe volume of data. If we're talking about lending, in your case, maybe we could just say auto lending. I presume you guys are still involved there. Maybe the number of loans to the number of different kinds of demographic slices, maybe it's really hard to get a hold of that data and make sense of it in terms of it being harmonized.

0:06:12

And maybe also there's not as much of it. What does it look like to overcome that barrier? Because I know a lot of vendors wrestled mightily with that. What's been your approach to make accessibility real?

0:06:23
Jay Budzik

Yeah. So one of our solutions to that problem is, first of all, there's two pieces, right? First is the process of using the data to build a model that's going to be predictive, that's going to help our credit unions say yes to more members, to approve more loans, to safely expand into other markets, to add a dealership, for example, and not take on lots of risk. So that process is kind of a careful process that we've learned over the course of years working with the world's largest banks.

0:06:52

The data that's required to power that process is also pretty key. And so what we've done through our experience is learned what data is more or less predictive in which segments, which types of lending problems, lending products, lending functions. So a marketing data source might be different than an underwriting data source And an underwriting data source for auto and indirect might be different than in a direct auto type product line. And so we have the map of all the data sources that are relevant and that produce the performance in Lyft.

0:07:24

And so what we can do is we can aggregate or pre-purchase lots of data and make that data available to our customers through our tailored modeling process. And so what this does is it allows us to build models that represent the lender, the borrowers, the lender has today, but that we'll also have tomorrow. And the result of that is that by adding automation on top, we can automatically build these models, build reports that compare performances, and really take a lot of that manual effort and the consultant work that would be required to implement or access the power of AI out of the process, let the machine do the work.

0:08:10
Daniel Faggella

Got it. And yeah, we're definitely going to double down on this as a point. I'm going to try to nutshell what you've said, make sure I've got the right grasp and the listeners have the right understanding, and then we'll keep moving. What I'm sort of hearing here is, okay, to make this stuff accessible to the smaller guys, you know, we've gone through the process of unlocking data and value with very large companies, very large data sets.

0:08:30

We can't take Citibank's data and go help Wells Fargo with it. That's obviously not going to be legally kosher or kosher in a lot of different areas. And for the smaller guys, they very frankly, the data might be so, let's say the opposite of harmonized, so dirty, if you will, or disorganized, or siloed and hard to access, or maybe so low in its volume that it might not be that much for us to work with. But what you're saying is there's a lot of third party broader sources.

0:08:57

And there's a process for what does it take to take this demographic set, this set of products, and determine our pricing models, determine our risk models based on previous experience. So it sounds to me like the process is transferable. The data, there's some centralized sources that you can kind of bootstrap with, and that's what you take to the little guy to get them started. Am I hearing you correctly?

0:09:20
Jay Budzik

Absolutely right. So all the smaller banks and lenders, the credit unions, they have experience lending. They know their customer base. And so we can interview them and find out about what type of products they're offering, where they lend, what type of credit they accept and what they're looking for, what their expansion plans are. And then we can, because we have that map of the data, we can go source the appropriate data for their model, and we can build that automatically.

0:09:49

And so instead of them having to have all the data in-house already, have it all cleaned and organized and ready for modeling, go into their legacy systems and get the data out, and having a 100-person data science team like they do at some of the institutions you mentioned in order to build models and validate them and make sure that they're not going to run off the rails, they can rely on us and the automated process that we've built based on that experience to deliver a product that's going to transform their business.

0:10:20
Daniel Faggella

Got it. And one other quick thing to just touch on, is it possible or does it happen? Like I'm imagining some random credit union in the Midwest and our email readership, I know for a fact that credit unions are actually a decent chunk of the folks that read our stuff in the financial services world. So there's going to be some of them tuned in. I'm thinking about a smaller one, you know, however many branches, whatever they have for revenue.

0:10:41

Maybe they've got some data. It's harmonized enough to potentially use, but the volumes are just really low compared to what you'd be working with with the big guys. Is there a way of looping in and bootstrapping with some of that stuff? Or does it almost make more sense to start with the more centralized sources and the proven process and maybe not deal with every bespoke data set of a tiny, tiny company? What does that look like in reality?

0:11:03

Yeah. So, you know,

0:11:05
Jay Budzik

We've all had the experience back in the, maybe we've all had the experience back in the 90s training of speech recognition systems. So you download the software. It doesn't work very well out of the box, even with Alexa today. It says, you know, could you say a couple of words so it can configure itself that it works for you? That's a little bit like what we're doing to make AI work for these smaller companies. You're right.

0:11:29

They don't often have enough observations or data to be able to make robust AI models. And so we need to marry their data with other data in order to make it really work. And so what we can do is we can incorporate their data in an automated process so that it complements a larger sample that's more representative of the type of people that may have applied but got denied or that they might be able to do.

0:11:58
Daniel Faggella

reach next year or the year after. Got it. Okay. And I think for a conceptual grasp here, when I think about the scalability of a business and how this works, I think on the one hand, you'd have a company like a MailChimp that can just sell the same thing to everybody. We don't have to customize it more or less. I mean, maybe that's a bit of an exaggeration, but it ain't much of one, to be honest.

0:12:21

You've got that extreme. You've got the opposite extreme where everything is bespoke. The services companies like a McKinsey, what do they do? They come up with some great process for determining pricing for new cars and they think hard enough to codify it into a process. And they have a couple of smart people that have used that process. And then their scale is smart people and a good process. And then they can move along and replicate that revenue-wise.

0:12:44

You guys are maybe somewhere in the middle. I think what a lot of listeners might think is that scale for an AI company is, oh yeah, they train it on Citibank and then it knows everything about lending and we plug it into this company. When it's like anybody who's close enough to this tech knows, number one, there's ethical issues there. Number two, the transferability just isn't that hot. So you're really addressing the reality of this in a way I think listeners should grasp, which is that you can transfer the process.

0:13:08

You can bootstrap with some of the data. You can bootstrap with some broad data that's going to be representative. And from there, we can come up with the tailoring that's going to work for your pricing model as a smaller bank, as opposed to, Yeah, we already trained everything about lending. We just jack it in with you out there in Wyoming, and you're going to have perfect lending prices for your pickup truck auto insurance out in Wyoming.

0:13:30

It's that easy. I think you've done a good job of nutshelling what it's really like. Is there anything we've missed there?

0:13:36
Jay Budzik

No, absolutely. That's where a lot of the lift comes from. I mean, if you think about FICO, it's trained on a national population. Most of the world still lives in New York, right? And so it's probably really good in New York, right? And so when you think about a credit union in the Midwest, in Iowa or in Florida, for example, that population looks very different from the population that lives in Manhattan, where both of the people in New York live.

0:14:02

So kind of the idea that model that has all that representation from a population that looks nothing like the population that you're trying to use it on would work is kind of misguided. And so what we do is we enable the lenders actually to themselves, tailor the model to their population. They know, credit unions know their membership. They know their membership. They know where they want to go. Chief lending officers have plans for which segments they want to reach, how they want to bring people up in their community, how they want to serve the community better.

0:14:39

And so what we could do is we can construct a data set that incorporates some of their data, some third-party data, but that's representative of the reality that they face today, but also their aspirations and build a model that way. The contrast with MailChimp is kind of interesting, except when you think about MailChimp as a software, you're configuring that thing yourself. You pick which messages go out at the end of the day. And so if our software is a lot like MailChimp, it's the same, but it enables an end user to configure the model to their needs.

0:15:14

And so we've developed this wizard-like process, much like a MailChimp, where you can go and set up your first lending model. And it interviews you for kind of what your goals are and what you want to achieve, what products you're interested in using it on. And that then determines which type of data you need and what type of process to apply on the back end to get the model built and to compare that custom tailored model with an industry score.

0:15:41

And so in that way, our stuff is a lot like a MailChimp. We also offer an API, much like MailChimp does, for the larger customers to be able to access the power of the automation within their own data science environment. it because they don't want to, they want to have full control of the data, you know, all that other stuff. And so, so those are, those are kind of the ways that folks are accessing it.

0:16:03
Daniel Faggella

Well, you know, and, and it certainly behooves you guys to make the technology more and more accessible. Clearly a MailChimp doesn't have the same iterative R and D process, right? When, when we configure our MailChimp, we click send, it's going to send the damn email. You know, in your guy's case, we might realize, oh, by golly, whatever we're doing with pickup trucks or whatever we're doing with Corvettes or something, you know, it's just like, you We've got to tweak this.

0:16:26

We've got to adjust this. So there's still some, certainly not as much as if you were doing it from straight scratch, but the challenges of AI, even though you as the vendor are tackling most of them, are still a reality for what you're doing here. This is not, it's going to be magical as soon as we turn it on and everything's going to be improved. There's still R &D, but it sounds like the user interface side of it, you're making it as close to a MailChimp as we can.

0:16:48

I couldn't say the whole process. I'm not going to make that analogy, but it does feel like from a UX user's experience standpoint, all they have to worry about is more configuration. They don't have to juggle maybe some of the more R &D. Let me know if I'm putting my finger on the point here.

0:17:02
Jay Budzik

Yeah, that's about right, Dan. But even with the MailChimp type analogy, you still send that mail and then you look for your open rate. You look at your - Oh, sure. Sure, sure. No, you're right. You're right about that. Conversion rate on the landing page. You have a whole monitoring dashboard for those mail marketing campaigns. And much like that, our model management system, which contains applications build, adopt, and operate. To build the model, adopt, to review the model, and make sure that it's compliant and that it meets your needs.

0:17:30

And then operate to monitor the model's performance is much the same. It allows you to sort of look and see, okay, well, how many new people did I get? What would I have done if I had left my old model in place? How much better is my business doing today? So it's really that end-to-end solution to allow a business user instead of a data scientist to manage that process of building the model, reviewing the model for compliance issues, and then operating it in production.

0:17:59

Totally. And look, again,

0:18:00
Daniel Faggella

from the vendor standpoint, you've got to make the MailChimp analogy closer than I do. Again, I still think that the undergirded R &D is a slightly different animal than checking our open rates. I think that some of that is we could call safely complexity. It's not to say it's hard. It's not to say people shouldn't do it, but I would certainly draw a distinction there. Regardless, it is a process of making this much more accessible than if people start from scratch.

0:18:22

And we're going to close on this note, Jay, is around the interaction with other applications. So there are platforms that do other things. And I'm sure that a credit union above a certain size is getting a lot of inbound email from many, many companies that do AI. And maybe some of them think everybody's doing the same thing, but they kind of don't. You know, there's a lot of these AI platforms that might enable a broader use of AI and customer service marketing, a bunch of other areas.

0:18:48

And then there's more directed point solutions that are really process focused, like you guys at Zest AI. Talk a little bit about how those two interact, because maybe that's going to be helpful for the listeners who don't quite know, you know, how does this differ from a data robot or whatever? How do you talk about kind of the productive ecosystem where there's multiple players involved?

0:19:06
Jay Budzik

Yeah, for sure. And this has really evolved over the last few years, Dan. And you have systems that are really great at managing the infrastructure of AI. So training AI models and using massive compute resources to do that. And so those have gained adoption in the enterprise and in the industry. But at the end of the day, most of these models are built in Python. And that provides a way of interoperating between various systems. And so there's nothing that says that you couldn't take our tools, for example, and install them in a data science environment that provides a Jupyter notebook and an interactive data science workbench.

0:19:53

And in fact, we have many customers that train their model using a platform company and then use our software to analyze the model for compliance, make it more fair, produce reports. and monitor that model. And so you're seeing, I think, a lot of interoperability now between the platform companies and folks like us who offer these really vertically focused solutions that solve a particular business problem.

0:20:18
Daniel Faggella

Yeah. And I think that's important for people to understand is that there's different tools for different kinds of jobs or different tools with different kinds of scope. And it's been very interesting. Actually, Jay, it's really been since our last interview, two and a half, three years ago, whatever it was, where we've started to see the companies more pointedly focused like you guys really become something very different, much more ease of use focused, even more so than before.

0:20:42

And the platform vendors kind of go their different routes when in the earlier days, it kind of felt like everybody was ambiguous as to are we an AI platform for everything or are we focused? It sounds like you're also seeing different players with different, more carved out specific roles who are able to acknowledge fully that they need to interact in order to add value to the client. Absolutely. Awesome. Absolutely. Okay, cool. I think it's a good frame to bear in mind for some of the listeners who are tuned in.

0:21:06

I think there's a lot of transferable lessons from today, whether people are in lending or not. What does adoption look like? How do vendors adjust to it? What does scaling the learnings of an AI company look like? Tons and tons to jot down for the folks tuned in. Jay, this has been a great episode. I know that's all we have for time, but I sincerely appreciate you joining us again. Thanks so much. Hey, thanks, Dan. So that's all for this episode of the AI and Financial Services podcast.

0:21:33

A big thank you to you, our listener, for listening all the way through on this episode. And thank you to Jay for joining us again. When Jay was with us the first time, I should have mentioned in our intro here, he joined us on the AI in Business podcast. And he is actually going to be joining us there again. Jay is going to be part of a series on measuring AI ROI in the coming two months.

0:21:51

So if you want to learn more from Jay and you want more podcast episodes featuring big name guests about AI trends and return on investment, be sure to check out the AI in Business podcast. Very easy to find it on iTunes, easy to find it on Spotify, SoundCloud, wherever you listen to podcasts. Our last series was on AI culture change, basically making your enterprise ready for AI. And our kickoff episode was the chief AI officer at IBM.

0:22:15

We had executives from Facebook, from Intel. We had the founder of a unicorn startup company. We had a lot of exciting guests in that series. And our next series on the AI and Business Podcast is on AI ROI. Jay Budzik is going to be part of that. We're glad to have him with us. And I hope to have you as a listener there. So check out the AI and Business Podcast. And otherwise, thanks again for tuning in.

0:22:33

I look forward to catching you in the next episode here on the AI and Financial Services Podcast.