Skip to playerSkip to main content
  • 3 days ago
As artificial intelligence becomes embedded across the mortgage lifecycle, lenders are rethinking how they use data to drive decisions and automate workflows. Chris McEntee, Vice President of Corporate and Product Development at ICE, discusses how AI mortgage lending is transforming mortgage business intelligence, why data governance is becoming more important than ever and what organizations need to build AI-ready mortgage operations that can scale with confidence.

Learn how ICE helps lenders build AI-ready mortgage operations powered by trusted business intelligence and data governance. Click Here: https://mortgagetech.ice.com/index
Transcript
00:00Are we all looking at this new thing in the right way?
00:03Is it creating compliance risk?
00:04Is it creating regulatory risk?
00:07Is it creating performance issues?
00:14My name is Chris McEntee.
00:16I've been working for about 16 years in various capacities.
00:19I currently have the title of Vice President of Corporate and Product Development.
00:23And what that means is I'm looking still at integrations or potential deals
00:28where we do some of the M&A work,
00:30but more importantly, more in a product strategy role
00:32to see how the pieces fit together.
00:35And today, of course, we're going to be talking about the data business
00:38that I'm responsible for.
00:45Business intelligence because of AI is going through a pretty major evolution.
00:49If you think about historically,
00:51business intelligence was tracking metrics, getting data, cleaning data,
00:55and then trying to create report engines, right?
00:58Whether they were crystal reports, you know, they've evolved over time or Tableau.
01:03Business visualizations, so complex data being able to show,
01:07you know, clearly delineate decision-making.
01:10So that's where a lot of the underlying data, enterprise data,
01:14really in its more recent movement was, you know,
01:17going through very sophisticated visualization tooling.
01:21Now you bring in AI and you have data as well as reporting
01:26that is going to be acting on some of those reports
01:29and some of them will be sequenced.
01:31And I think the complexity of being able to say,
01:33I have data, the data is telling me something about the business.
01:38Can I create an automated tool that would react to that signal,
01:42right, in the data?
01:43And so really what a lot of people are doing nowadays
01:46is just trying to architect a future space
01:49for what they want to see, especially with AI.
01:52AI's got immense amount of promise.
01:54We all know that.
01:55And a lot of the POCs are working on, you know,
01:58trying to delineate where can we automate certain tasks,
02:01where are there certain compliance-driven tasks
02:03that we need to rope in.
02:05But most importantly,
02:06it's going to be driven by business intelligence, right?
02:09I'm going to see a signal.
02:10I'm going to see an update on a borrower or a consumer.
02:13I'm going to see some counterparty risk.
02:15I would like my automated tools,
02:17if they're driven by AI,
02:18to notify me as soon as that emerges.
02:20And that's going to require very direct connection
02:23into business intelligence and business data.
02:29Because lending is so diverse within the US
02:33and there's different business models,
02:35really a lot of the drivers of how they use the data
02:39and how they want to consume the data
02:41is very driven by, you know, again,
02:44counterparty risk, product offering,
02:47consumer direction, multi-channel.
02:49So when you look across each of the businesses
02:53and what their business model is,
02:54whether direct to consumer
02:55or what channel they're working on
02:57or what they want to grow,
02:58or the campaigns, for instance,
03:01being a marketing campaign
03:02and trying to automate that
03:03or in the midst of a refi boom,
03:06all those are driven by real-time data
03:08to the extent you can have it.
03:10A lot of the industrial consumers of our data
03:14are going out to third parties independent of us.
03:17So they have their enterprise data,
03:18call it their proprietary data,
03:20generated by their own volume and transactions.
03:23They're going to us
03:24because we might have proprietary market data
03:26that influences a decision.
03:28So it has to get combined in there
03:29and then go to a third party.
03:32So depending upon
03:33what the data infrastructures look like
03:35and how sophisticated they are,
03:36they can have probably a sophisticated role
03:38as a JP Morgan does,
03:40which, you know, the data teams are,
03:42you know, driven by data scientists,
03:44super complex environments.
03:45They can bring in all sorts of data real-time
03:47or just somebody saying,
03:49look, I want to look at a pipeline
03:50and I want to see what my customer,
03:52you know, click through
03:53and I get some reports
03:54and I want to see that
03:55through my marketing engine.
03:57So it really runs the gamut.
03:58The key thing is,
04:00is that getting the data accurate,
04:03which involves some internal governance, right?
04:05What is the source of truth
04:06within your enterprise?
04:08And then if I go out to third parties,
04:10am I going to get some conflict
04:12between how these data sets get put together,
04:14especially if they're going to be
04:15put into something that's automated?
04:17So a lot of decisions,
04:19we see the customers come to us collaboratively
04:21and ask us about what that looks like.
04:24Have you seen an implementation like this?
04:26We'd like to do this.
04:27We have this firewall issue.
04:28So we want to be collaborators
04:30as well as helping them get the best solution.
04:38Data governance,
04:40the initial perception would be,
04:42oh, that's putting a wet blanket
04:44on this novel innovation
04:45and what have you.
04:46And it's actually the opposite
04:49because a governance structure,
04:51and what I mean by that
04:52is explicit charters,
04:54what you're going to do with data,
04:55where are you going to store data,
04:56where are you going to update data?
04:57All those questions get answered
04:59in a more collaborative fashion.
05:01When I think about governance
05:02and I have a number of governance roles,
05:04you're bringing all the stakeholders
05:06initially and saying,
05:07what do you see here?
05:08So that's cyber risk,
05:10it's infrastructure,
05:11it's the systems engineering.
05:13So, you know,
05:15the word governance,
05:16kind of an abstraction,
05:17but from a practical standpoint
05:18is are we prepared
05:20to move this thing
05:21through a product design
05:23or what have you
05:23through a proof of concept
05:26into a limited release
05:28and then into production?
05:30The power of the tools
05:31that we're now encountering
05:33are so sophisticated,
05:34so complex.
05:35There's our concerns
05:36that people think,
05:37oh, well,
05:37either one,
05:38I get a false positive
05:40or I put out incorrect information
05:42to a bar,
05:43you know,
05:43to the extent
05:44that's a compliance issue
05:45or if I'm doing
05:47a business-to-business,
05:48talk about the reporting
05:49we did earlier,
05:50I'm getting inaccurate signals
05:52because, I don't know,
05:54somebody didn't update
05:55their model accurately,
05:56okay,
05:56some dependency
05:57on a third party.
05:59So I think that
06:00what's happening
06:01is a combination
06:02of available data,
06:05you know,
06:05real-time almost compute
06:06capability and automation
06:08as well as
06:09these kind of evolving
06:11things called LLMs
06:13and these large language models
06:14and these things,
06:15you know,
06:15the dynamic as to
06:16what is the output,
06:18right?
06:19I want an answer,
06:20which, you know,
06:20we're seeing through
06:21some of the query,
06:22but more importantly,
06:23I want to set up
06:24a series of tasks.
06:26So think about it.
06:27If I get the first task wrong,
06:29the following five tasks
06:30are going to be off.
06:31And so I think people are very,
06:34want to be as precise
06:34as they can be
06:35engineering-wise,
06:36but more importantly saying,
06:38are we all looking
06:39at this new thing
06:40in the right way?
06:41Is it creating compliance risk?
06:43Is it creating
06:43regulatory risk?
06:46Is it creating
06:47performance issues?
06:49So go back
06:49to the core of governance.
06:50I think it starts there,
06:52your entitlements,
06:53your controls,
06:55and then you think about
06:56what's the use case,
06:57distribution,
06:59are people ingesting
07:00this data today
07:01in a different fashion?
07:03And then specifically,
07:04go back to the point
07:05I brought up earlier,
07:06you probably have
07:08some nested activities there.
07:09What I mean by that is
07:10you're asking for data,
07:12then you're replying,
07:14and if that initial data
07:15response is called dirty,
07:17it's just going to
07:18kind of cascade through.
07:19So people are doing
07:20a lot of testing in QC
07:21to make sure these things
07:22work properly,
07:23and that's all governance.
07:24It's governance 101,
07:25so it's good hygiene.
07:33Primarily because
07:33we are a system of record
07:35for two large,
07:36you know,
07:37infrastructure systems.
07:38So obviously,
07:38your servicing platform
07:40as well as
07:40your origination platform.
07:42So you start there
07:43and you say,
07:44what is my system of record?
07:45What's my source of truth?
07:46And we have the applications
07:47that actually have that data.
07:49So once you have that data
07:51and you can say,
07:51look, I can manage it,
07:53I can give transparency to it,
07:54and then I can also produce reports
07:57that people can consume.
07:58The other benefit that we have
08:00is we also have
08:01a lot of data businesses.
08:03So to the extent
08:03that data gets presented
08:06to one of the users
08:07of the application,
08:08we can also put data
08:09into the workflow.
08:10So go back to my analogy
08:12either on the,
08:13or earlier on,
08:14the automation piece.
08:15I'm getting that real-time data.
08:17It doesn't necessarily
08:18have to be my data.
08:19It could be with some
08:20of the data businesses
08:21that are other parts of ice.
08:23Rates,
08:24product offerings.
08:26So again, rate sheets,
08:27but to the extent
08:27even market data.
08:29So I think some of the benefits
08:31that you see
08:31is our ability
08:32to be able to say,
08:33you have your enterprise data lender.
08:36We've kept good hygiene on it
08:39for that and availability.
08:41You have workflows
08:42that are in the application.
08:43We can inject data into that,
08:45whether it's your data
08:47or third-party data
08:48just to make the efficiencies go.
08:50And to the extent
08:51that we have a network
08:52where other third parties
08:53can provision data
08:54and when you see this
08:55in fees and fraud
08:56and some of the other products,
08:58there's a very robust market
08:59out there for it.
09:00So I think the idea
09:01that the data
09:03just gets sole-sourced,
09:05we think one of the benefits
09:06is we give you options
09:08and you're going to have
09:09multiple options.
09:10I want to go proprietary
09:11with some of my own data.
09:13I want to use this third party.
09:14We can accommodate it.
09:15So I think it's our flexibility,
09:17scalability,
09:18and just the extent
09:19of the amount of data
09:21that we do have.
Comments

Recommended