AI Is Moving Fast. Is Your Data Ready?

WEBVTT
00:00:00.000 --> 00:00:05.065
You don't buy a car and expect your mechanic to repair it for you for free.
00:00:05.065 --> 00:00:11.080
Sure, you might get three years, and you get six visits out of that, but, you're expected to pay the insurance.
00:00:11.080 --> 00:00:12.007
You put gas in the car.
00:00:12.007 --> 00:00:18.146
You go get it serviced, otherwise it gets repossessed or, the engine seizes and you can't drive it anymore.
00:00:18.146 --> 00:00:19.463
It's the same concept.
00:00:19.464 --> 00:00:33.368
You have to constantly be keeping an eye on it but the real benefit is gonna come from what I call data citizens across the organization, identifying issues and knowing who to go to raise the problem to find the solution.
00:00:33.731 --> 00:00:47.716
And that way it turns into this self-propelling data management engine that doesn't fall on anyone's shoulders because we all recognize we, we have table stakes in the management of our data because it, it drives our business decisions.
00:00:48.051 --> 00:00:50.954
It d- dri- it drives how we think about what we're doing.
00:00:51.262 --> 00:01:03.243
And you know, it makes you care a little bit more about what you're putting into your Excel spreadsheet or into your, your application because you un- understand the costs and the work that come with fixing it if you don't do it the right way,
00:01:15.864 --> 00:01:22.527
Hey, I'm your host, Ben Parker, and welcome back to this week's episode of Data Analytics Chat.
00:01:23.114 --> 00:01:28.055
Today we're looking at the question that is becoming more important as AI moves faster.
00:01:28.650 --> 00:01:30.703
Is your data actually ready for it?
00:01:31.391 --> 00:01:41.394
So there's a lot of organizations now obviously heavily investing in AI, but underneath that sits like, it's much more the fundamental challenge around data ownership, governance, and trust.
00:01:41.864 --> 00:01:55.986
So in this week's episode, we're gonna look at what it takes to strengthen those foundations, where organizations tend to get stuck what good data ownership actually looks like, and how getting this right can support AI and business value.
00:01:57.162 --> 00:02:04.269
So joining me this week is Erica Howard, who is the head of consumer data ownership at TD.
00:02:04.868 --> 00:02:06.590
So yeah, welcome, Erica.
00:02:07.816 --> 00:02:08.510
Thanks, Ben.
00:02:08.766 --> 00:02:09.818
I'm happy to be here
00:02:10.349 --> 00:02:10.675
Good girl.
00:02:10.675 --> 00:02:12.276
I'm looking forward to this conversation.
00:02:12.655 --> 00:02:19.669
So I guess before we dive in, do you wanna just give the listeners a quick introduction to who you are and the work you're doing?
00:02:20.813 --> 00:02:21.391
Absolutely.
00:02:21.411 --> 00:02:23.051
So like you said, I'm Erica Howard.
00:02:23.054 --> 00:02:35.594
I am the head of consumer data ownership at TD, and really functionally what I do is work with the various business segments in my area to enable the amana- data management practices within their respective groups.
00:02:35.597 --> 00:02:45.775
So that means helping them to understand when to identify a data issue getting them into some of our central pooling for reference data and metadata management.
00:02:45.776 --> 00:02:59.406
But ultimately guiding them through the process on how to understand and manage their data within their areas and making sure that we're tracking that centrally so that we're demonstrating progress on how we're maintaining it.
00:03:00.594 --> 00:03:01.049
Okay.
00:03:01.282 --> 00:03:08.801
And then obviously I know, obviously we had conversations you had a number of, obviously you've had a, yeah a long stint in your current employer, which is obviously unheard of in today.
00:03:08.826 --> 00:03:12.616
So I guess how did you end up focusing on data ownership and governance?
00:03:14.066 --> 00:03:16.883
What I could genuinely say is that I fell into the function.
00:03:16.883 --> 00:03:19.903
So you are fair in saying I've been with the bank for 18 years.
00:03:20.331 --> 00:03:21.663
I did start in the call center.
00:03:21.990 --> 00:03:32.169
Earlier in my career when I was still quite a young analyst, I had been working in the space that was building our early data governance model, and this was back in probably 2012.
00:03:32.569 --> 00:03:41.263
So I did that for a couple of years, and then I decided that I felt like I wanted to get back more to the analytical space, work with the data to help answer some strategic questions for the businesses.
00:03:41.756 --> 00:03:53.020
And then probably a little bit over a year ago, I had an opportunity to move back into the data management space and act as a bridge between tech and the business, helping them understand tho- those pieces.
00:03:53.418 --> 00:03:55.966
So I took the leap and I moved into this ownership role.
00:03:56.848 --> 00:03:57.153
Okay.
00:03:57.318 --> 00:04:02.907
I think it's I think it's gonna be an important one moving forward, especially with obviously the way LLMs and agents are going.
00:04:03.709 --> 00:04:07.032
And a- obviously a, a difficult challenge for businesses, shall we say?
00:04:07.633 --> 00:04:08.288
Absolutely
00:04:09.484 --> 00:04:16.290
So I guess when did it become clear that data ownership and governance were becoming, like, critical to your organization?
00:04:17.418 --> 00:04:29.415
So surprisingly, I would say data governance and data management really became a thing after the US housing market crisis in, the 20- 20-- 2008, 2010 timeframe.
00:04:29.911 --> 00:04:35.559
With all the systematic defaults and all of the issues around the, the mortgage lending.
00:04:35.865 --> 00:04:47.192
The regulators, at least in the US, kinda came down quite hard on a number of financial institutions'cause they, they stood up Dodd-Frank, and they stood up stress testing, and those were really significant bodies of work.
00:04:47.194 --> 00:05:00.685
And we had to also demonstrate not only, what we were doing in providing the end result in those reports, but we also had to demonstrate how we created those reports and ensuring that the information that we were giving them was actually correct.
00:05:00.685 --> 00:05:09.353
And so that's where some of our earlier data governance frameworks came into being not just at TD, but across a number of other institutions.
00:05:09.711 --> 00:05:20.048
And so over time and distance, that model has continued to evolve to where we are now, which it's definitely broadened under some of our more recent standard enhancements.
00:05:20.055 --> 00:05:26.935
So now we're starting to look beyond the regulatory scope into what does good governance around AI look like?
00:05:26.935 --> 00:05:29.182
What does good governance around models look like?
00:05:29.399 --> 00:05:37.649
It's a very big question, one that we haven't really focused on because of the regulatory nature of why data management what became important.
00:05:37.658 --> 00:05:51.012
But I think knowing the challenges we saw back during that h- housing crisis and knowing that, we're starting to see some hedge cases coming through in the media around challenges w- with AI, like the two dots are starting to connect.
00:05:51.444 --> 00:05:58.504
And I think folks are starting to ask a lot of questions around, what is going into these models that are producing these outcomes.
00:05:59.692 --> 00:06:02.853
Yeah, no, I think obviously it's funny, a lot of things happen, isn't it?
00:06:02.874 --> 00:06:15.459
When there's a big impact in the world, obviously changes happen, and I think now obviously, especially with AI, I think businesses are starting to see that you need your data, you need the data quality up there or you're just not gonna get the results you need.
00:06:16.059 --> 00:06:16.673
Absolutely.
00:06:16.673 --> 00:06:23.569
And it, I like, I think the biggest thing of it too is c-client customer confidence, right?
00:06:23.569 --> 00:06:24.934
You, the people...
00:06:25.187 --> 00:06:27.384
and it's not just financial organizations, right?
00:06:27.435 --> 00:06:28.819
It's any organization.
00:06:28.857 --> 00:06:35.937
We collect tremendous amounts of information about our clients, about the markets that they operate in.
00:06:36.182 --> 00:06:40.100
So there's an underlying assumption that we're doing everything that we can to care for that data.
00:06:40.100 --> 00:07:13.009
And when we produce or when we see AI outcomes where, something goes left of center or creates a negative client experience, they're gonna look at us and ask the question of why didn't you do it better?" And so that's where some of the, the data management and the governance is gonna play a big part, because if we can stick the landing on the governance piece, produce solid outcomes that don't have hallucinations, don't have bias skews, that will, one, create confidence with the customer that we are looking out for their wellbeing, that we are being responsible with AI deployment.
00:07:13.288 --> 00:07:15.985
But also, I think it's gonna create confidence in AI.
00:07:15.987 --> 00:07:27.560
It's really easy to create a negative experience for somebody versus coming back from that and creating a positive experience, especially depending on how significant the negative experience was.
00:07:28.785 --> 00:07:29.588
Yeah, no, definitely.
00:07:29.853 --> 00:07:35.569
So obviously you mentioned obviously the h- housing crisis was the, I guess the big, yeah, issue.
00:07:35.642 --> 00:07:38.867
So what gave-- at that time, what was the big problems you were seeing at that point?
00:07:40.463 --> 00:07:51.526
So I think the biggest thing is when we started to do the work, when it was getting the basis of the framework put together so that we could create accountability throughout the bank in terms of who owns what, right?
00:07:51.565 --> 00:08:05.607
When we're building, a regulatory report, it's not just putting widgets in a spreadsheet and shipping it out the door, but it's also creating the lineage, tracking how it's moving through the bank from the point of collection to the point that we send it to the regulators.
00:08:06.466 --> 00:08:14.925
To do that lineage work, what you really started to find was I got a lot of, I got a lot of things that are not necessarily a straight throughput.
00:08:14.968 --> 00:08:22.161
I've got manual processes I've got, different things that are getting interjected until we, we get to the final answer.
00:08:22.454 --> 00:08:32.884
Not to say that anything, any of it was wrong, but it did demonstrate that it was gonna become very difficult to govern when you have all of these little disparate pieces coming together to make the whole.
00:08:33.188 --> 00:08:47.009
And so it really demonstrated the need to really start to the investment in the data, getting things into appropriate data warehouses where you could do data quality testing more cleanly, monitor the data movement more accurately.
00:08:47.273 --> 00:08:54.567
And so it really started the impetus for a lot of change in our organization, starting with how we did our regulatory reporting.
00:08:54.567 --> 00:09:01.863
A lot of how it had been done almost after the, the housing crisis has changed very radically since then.
00:09:01.863 --> 00:09:03.825
We've built custom warehouses.
00:09:03.826 --> 00:09:07.423
There's timing, there's reconciliation, there's far more control.
00:09:07.721 --> 00:09:13.024
But it was that body of work that really triggered that transformation.
00:09:14.193 --> 00:09:14.653
Yeah.
00:09:14.948 --> 00:09:16.841
Obviously a lot of businesses do struggle with this.
00:09:16.841 --> 00:09:17.217
Do you feel...
00:09:17.289 --> 00:09:21.375
I guess back in that period, why, why didn't you have a thing in place already?
00:09:21.397 --> 00:09:22.674
Was it just a lot of data?
00:09:22.989 --> 00:09:24.230
Was it, what were the challenges, there?
00:09:25.982 --> 00:09:35.428
I think it's one of those things where if you're not asked the question, you don't know that you need to worry about it Yeah I think is the simplest way to put it.
00:09:35.429 --> 00:09:40.812
And then we got asked the question, and then we realized, oh, we really do need to worry about this.
00:09:41.263 --> 00:09:45.311
Now how the question has gotten asked has changed over the years, right?
00:09:45.311 --> 00:09:49.302
I think for many institutions, they started out with regulatory.
00:09:49.307 --> 00:09:56.466
How they expanded that I don't know p- per se, but I do know that the focus always has been largely on the regulatory.
00:09:56.467 --> 00:10:04.965
I think the concept now of scaling it out more broadly, looking organization-wide, is a relative- relatively new concept.
00:10:04.966 --> 00:10:12.440
It's also a lot easier, like once you ask the question, if the question is something unspecific, you can care for that, and you can make sure that stands tall.
00:10:12.441 --> 00:10:35.427
But when you start asking yourself the question of, how do I scale this to an entire organization- that may have existed for 30, 40, 50 years it becomes a really daunting question because now you have to look at all the other processes that are running your organization, and then figuring out, okay, how do I get them to the same place that I got for this one specific component?
00:10:35.833 --> 00:10:39.378
And that, I think is gonna be a, a, a broader challenge business-wide
00:10:40.448 --> 00:10:42.746
It's definitely a big headache, isn't it really?
00:10:42.749 --> 00:10:45.162
Obviously it's a m- just a, it's a massive change, isn't it really?
00:10:45.163 --> 00:10:53.008
'Cause obviously we've seen it a lot, a lot of businesses over the years have invested in data scientists, obviously now ML engineers, like building new models.
00:10:53.010 --> 00:11:04.105
But obviously I think now a lot more businesses are s- starting to understand that getting their data actually in the right format, obviously understanding their data is gonna be crucial to moving forward
00:11:05.860 --> 00:11:09.543
I, understanding the data is one of the, the pieces that's gonna be critical.
00:11:09.543 --> 00:11:15.783
The other under- the other thing that we're finding is a, is an interesting new challenge is just the data architecture itself.
00:11:15.783 --> 00:11:30.913
Depending on how your organization grew, you could have a nice lean data environment where you have a cloud and you have a data lake and you may have, a specific set of systems that you've consolidated everything down to, and so governance is an easy question.
00:11:31.320 --> 00:11:42.414
Now, you could be like other organizations where, you have a spaghetti soup diagram, as I like to call them, and you've got things here, there, and everywhere and it's a Gordian knot of how the data is moving around.
00:11:42.414 --> 00:11:49.993
And it's in those situations I think that data governance becomes far more challenging because now you have to figure out where do I rationalize down?
00:11:50.057 --> 00:11:51.019
Where do I simplify?
00:11:51.019 --> 00:11:54.910
Where do I consolidate things to make the governance piece of this a little bit easier?
00:11:55.100 --> 00:12:07.485
It has other upsides for the organization as well, because all of that residual tech debt that you're carrying, the cost to maintain it, that eventually goes away, so that's a great, productivity reinvestment opportunity.
00:12:07.827 --> 00:12:20.370
But the first piece is just understanding the landscape and then figuring out where all of the data flows through all of it to figure out how to rightsize it so that I can scale and move faster in some of the other spaces that I'm interested in.
00:12:20.371 --> 00:12:27.794
And I think that will be a broader scale challenge, not just- within the financial industry, but I think across the board, right?
00:12:27.796 --> 00:12:35.306
Because organizations didn't build their data architecture for AI, they built their architecture to run their organization.
00:12:35.700 --> 00:12:43.647
Trying to overlay AI now creates complexity because perhaps how you grew your organization is not consistent- with how AI needs to operate.
00:12:45.465 --> 00:12:47.917
So do you see it's like a...
00:12:48.341 --> 00:12:49.607
what's the b- the biggest hurdle?
00:12:49.618 --> 00:12:51.326
Do you think it's poor data quality?
00:12:51.336 --> 00:12:54.279
Is it unclear ownership, access or something else?
00:12:54.349 --> 00:12:57.768
'Cause obviously with you yourself, obviously you're working for a global enterprise and obviously...
00:12:58.250 --> 00:13:01.606
And especially now with AI, obviously there's so much more data available.
00:13:01.716 --> 00:13:05.719
What do you think has been the the, the hard, the, the most challenging part?
00:13:06.910 --> 00:13:08.697
It- and that's a really interesting question.
00:13:08.748 --> 00:13:11.200
I, honestly, I think it's the ownership piece.
00:13:11.241 --> 00:13:15.418
I s- we've tried the governance model since 2012 and it's evolved.
00:13:15.436 --> 00:13:17.922
Initially it was a- a data stewardship framework.
00:13:17.937 --> 00:13:20.907
It was federated in the business areas.
00:13:20.914 --> 00:13:26.443
It was only for specific groups, then it became a centralized function, now we've pushed it back to, to federated.
00:13:26.929 --> 00:13:43.681
The, the conversation on who owns the data has evolved over the years, but I think we've landed in a better place in this, this liter- la- this latest iteration where we're sitting in the business lines with the folks that are making the decision on the, the applications that house the data, that collect the data.
00:13:43.960 --> 00:13:51.485
We can demonstrate some influence there in, on how we're thinking strategically about data and capturing it going forward.
00:13:52.044 --> 00:13:56.461
But o- ownership i- is key alongside, alongside understanding.
00:13:56.827 --> 00:13:59.451
Data quality issues are always gonna happen.
00:13:59.616 --> 00:14:02.234
It's not something- that you can get away from, you can manage to it.
00:14:02.616 --> 00:14:04.475
And we're managing it to it very well.
00:14:04.533 --> 00:14:08.195
But there's also having a connected network as well.
00:14:08.202 --> 00:14:17.198
Having that clear ownership means that then you have, not just within my space, but across all of my partner spaces in the institution too, means now we have a connective tissue with one another.
00:14:17.485 --> 00:14:20.681
Because then we can start to say,"Okay, I'm doing this thing here.
00:14:20.681 --> 00:14:21.745
Are you doing the same thing?
00:14:22.572 --> 00:14:23.631
Does that make sense?
00:14:24.187 --> 00:14:28.726
Or should one of us do it?" And that's also allowing us to rationalize some of the processes.
00:14:28.947 --> 00:14:31.177
But ownership i- is absolutely key.
00:14:31.187 --> 00:14:36.745
You need that, someone to be accountable for this in this particular domain and space
00:14:37.336 --> 00:14:37.655
Yeah.
00:14:37.715 --> 00:14:41.908
I'll say I, I hear that quite a bit and it's, problem with data is it's never ending.
00:14:41.951 --> 00:14:42.941
The projects just keep...
00:14:42.986 --> 00:14:44.647
It's just, you're chasing your tail all the time.
00:14:44.686 --> 00:14:50.453
It's just it's not again, I think I've noted before in a p- podcast, like it's not like software projects where you have a finish.
00:14:50.490 --> 00:14:53.644
With data it's just, it's constant gr- like just constantly moving
00:14:54.245 --> 00:14:55.495
No, a-a-absolutely.
00:14:55.496 --> 00:15:06.144
And like the other piece of the equation as well is, if you are, operating in an environment where procedurally things have changed, like you also have to take a look at, like how does that impact the data?
00:15:06.181 --> 00:15:07.820
Did you go back and you corrected it?
00:15:07.831 --> 00:15:09.671
Did you decide to leave it as is?
00:15:09.706 --> 00:15:15.768
That's a decision that you can make if it doesn't create risk and it doesn't create a negative impact for a client, if it's like a minor change.
00:15:15.789 --> 00:15:19.780
But then depending on the new platform that you move on, a minor change could become a big issue.
00:15:19.794 --> 00:15:22.797
So it's those types of things too that now we're...
00:15:22.813 --> 00:15:23.611
Like as we're...
00:15:23.912 --> 00:15:24.875
'Cause you're absolutely right.
00:15:24.875 --> 00:15:26.551
There's through projects all the time.
00:15:26.897 --> 00:15:30.955
Sometimes there's things put through as break fixes that don't qualify as a project.
00:15:30.971 --> 00:15:31.966
They could trigger an issue.
00:15:31.970 --> 00:15:51.466
It's really an, an ever-evolving cycle of data change and movement, and it's really about taking a beat, thinking about it end to end, making sure you're doing the right things both for yourself, for the organization and for the people that you're supporting whatever your organization may be
00:15:52.067 --> 00:15:59.423
So was there a point where you realized your show was not really the technology, but the way, data was owned, managed, or governed?
00:15:59.722 --> 00:16:06.458
I think I probably came to that, that realization a couple of years ago.
00:16:06.835 --> 00:16:08.052
I was working on a project.
00:16:08.052 --> 00:16:09.697
We were trying to...
00:16:10.048 --> 00:16:11.962
And this was before I had stepped into this role.
00:16:12.013 --> 00:16:17.629
I was leading in an area that was supporting risk and regulatory data responses.
00:16:17.634 --> 00:16:22.455
We were building out some routines and reporting around proactively monitoring for controls.
00:16:22.835 --> 00:16:34.804
It was a really useful exercise, but what we were finding is it's not that we didn't have the controls, it's that we didn't have the access to the data to get at it, to demonstrate, to prove it to monitor for it.
00:16:34.804 --> 00:16:43.754
And like we knew we needed to do it, but getting to it was the problem, and I was trying to build a framework where we were bringing business and technology and governance together so we could solve the problem.
00:16:43.783 --> 00:16:47.514
The, the issue was, the governance team was centralized.
00:16:47.813 --> 00:16:51.301
They didn't have a lot of influence on getting things fixed.
00:16:51.485 --> 00:16:54.288
Tech required the business to prioritize.
00:16:54.291 --> 00:16:59.418
And the folks that I was working with at the business, they weren't necessarily sitting in the business proper.
00:16:59.418 --> 00:17:03.138
And the ability to fund and prioritize change was challenged.
00:17:03.138 --> 00:17:09.813
And it occurred to me"Who's gonna stand tall within a business area and say,'Hey, we need to do this because it's the right thing to do'?
00:17:10.161 --> 00:17:31.279
Because you're gonna get value in doing this thing, that you are gonna mitigate risk because you have visibility into something, so you can see it becoming a problem before it ac- the problem actually hits and stopping it in its tracks." And it was a thing that I was challenged with because I wanted to do the right thing.
00:17:31.280 --> 00:17:32.710
I always wanna do the right thing.
00:17:32.710 --> 00:17:38.340
And so when this role opened up and they had uplifted our standard to say,"Hey, we're putting data owners in the business.
00:17:38.717 --> 00:17:50.402
They're gonna be the people that are gonna stand tall for these," I immediately was like,"I wanna raise my hand." Because this is something that I've experienced, I've felt the pain of, for many years.
00:17:50.402 --> 00:17:58.127
And I wanna be, the impetus of change versus"Let's just keep doing the thing because it's the way we've always done it," good, bad, or indifferent.
00:17:59.377 --> 00:18:03.903
So it must be a difficult challenge yeah, a difficult challenge to get a message across internally
00:18:04.502 --> 00:18:22.316
It is, but i- generally speaking, when you are working in a corporate environment, getting a message across can generally be dif- difficult, and not because- the leaders aren't willing to listen, but in some cases you have to speak in a language that they understand.
00:18:22.316 --> 00:18:26.703
Like I have been told that I'm very technical, and that's probably very true.
00:18:27.251 --> 00:18:33.359
I understand things in a very technical way, and I can communicate in that way.
00:18:33.359 --> 00:18:35.208
The problem is not everybody speaks that language.
00:18:35.211 --> 00:18:39.534
So I start talking and it's like Greek, and I can see the lights die in their eyes.
00:18:39.875 --> 00:18:53.384
And so one of the things that I have been working on in the last year is how I'm telling the story, how I am breaking this into- digestible bites that these business leaders can understand and understand the impact of.
00:18:53.869 --> 00:18:56.730
And I think it's, I think it's making...
00:18:57.188 --> 00:18:58.766
I'm making headway with it, right?
00:18:58.767 --> 00:19:05.912
Because now leaders are starting to ask the right questions, challenge the right things, are getting to be a bit more conversant.
00:19:05.959 --> 00:19:09.351
Data in and of itself had, had been very far removed.
00:19:09.381 --> 00:19:15.691
These, these folks were not expected to, even be conversational in it, and I don't necessarily expect them to.
00:19:15.692 --> 00:19:16.978
I don't expect them to be me.
00:19:17.021 --> 00:19:17.900
That is my job.
00:19:18.255 --> 00:19:23.442
But I want them to be prepared, and I want them to feel comfortable in what they're speaking to.
00:19:24.057 --> 00:19:29.907
So I think I'm starting to kinda see that, that shift, but I can't say that it was not without its challenges.
00:19:29.913 --> 00:19:37.885
But I do work with a really great collection of people and they've done a tremendous job listening and supporting me and digesting what I'm telling them.
00:19:37.887 --> 00:19:44.105
So that, that has been a really great win despite all of the- the challenges up to that point.
00:19:45.167 --> 00:19:45.334
Yeah.
00:19:45.411 --> 00:19:46.819
Data literacy is gonna be...
00:19:46.862 --> 00:19:49.354
it's the big thing now for big businesses, isn't it?
00:19:49.354 --> 00:19:50.510
Especially with AI.
00:19:50.535 --> 00:19:53.190
'Cause businesses are getting taken upside down really, aren't they?
00:19:53.211 --> 00:19:55.102
'Cause you've gotta completely change the way it is.
00:19:55.109 --> 00:20:06.167
So it's getting every business unit to actually understand what's involved and obviously, not necessarily being an expert, but actually understand that why you're doing this and the impact it can have
00:20:07.688 --> 00:20:09.013
E- exactly, right?
00:20:09.038 --> 00:20:14.507
Because, in many cases in the past, you treat data management as a project.
00:20:14.507 --> 00:20:21.640
You have an issue, you realize that data is the root of it, you drum up a project, you tactically go and you fix the data, it's done.
00:20:22.169 --> 00:20:29.765
I have been preaching this philosophy with my partners and my peers that data management is not a tactical thing.
00:20:29.775 --> 00:20:31.444
It is a strategic investment.
00:20:31.861 --> 00:20:59.944
If you are caring for your data on a recurring basis, if it is part of your bottom line and it's baked into your strategy every year and you've got a plan to care for it, to maintain, but also focus on areas where you realize, hey, this, this spot needs to get tightened up, you'll find that you can move at speed a lot faster than if you maintain the status quo and then try to do something with data that you might not have looked at before.
00:20:59.994 --> 00:21:05.366
You collected it and you realized, hey, I could use it for this, and then realized, oh, i- it...
00:21:05.394 --> 00:21:07.659
the quality wasn't what I was expecting it to be.
00:21:07.659 --> 00:21:10.516
That can actually be a really huge hindrance.
00:21:10.523 --> 00:21:15.955
So the thing that I often use, or the analogy that I often use with folks is, it's like a car.
00:21:15.958 --> 00:21:21.023
You don't buy a car and expect your mechanic to repair it for you for free.
00:21:21.023 --> 00:21:27.038
Sure, you might get three years, and you get six visits out of that, but, you're expected to pay the insurance.
00:21:27.038 --> 00:21:27.965
You put gas in the car.
00:21:27.965 --> 00:21:34.105
You go get it serviced, otherwise it gets repossessed or, the engine seizes and you can't drive it anymore.
00:21:34.105 --> 00:21:35.422
It's the same concept.
00:21:35.423 --> 00:21:37.630
You have to constantly be keeping an eye on it
00:21:38.230 --> 00:21:38.512
Yep.
00:21:39.090 --> 00:21:44.047
And then I guess what, what are you changing as you build out a stronger data ownership and governance?
00:21:44.067 --> 00:21:45.694
I think it'd be a good point to ask
00:21:46.786 --> 00:21:57.935
So the one thing that I came to a really swift realization on, and I think being here for 18 years has been a tremendous benefit in that decision, is I'm the first person to say that I don't know everything.
00:21:57.938 --> 00:22:00.057
And I do often say I know a lot about a little...
00:22:00.268 --> 00:22:01.921
I know a little bit about a lot of things.
00:22:02.443 --> 00:22:12.151
But I'm well aware that, across all the business areas that I support, and I have oversight of about six or seven different very distinct business areas.
00:22:12.576 --> 00:22:17.660
I don't have the deep knowledge of their processes, their routines, how they collect their data.
00:22:18.442 --> 00:22:25.704
But so what I've opted to do is instead of staying centralized or hybrid or federated, I've opted for a hybrid model.
00:22:25.705 --> 00:22:26.619
I act as a hub.
00:22:26.684 --> 00:22:27.804
I support them.
00:22:28.513 --> 00:22:35.299
And we have designees in each of the business areas that we call segment data owners, and they're really my conduit into the business.
00:22:35.301 --> 00:22:39.114
If we have a problem in a specific area, I reach out to that person and their team.
00:22:39.114 --> 00:22:40.723
We start to bring the right folks together.
00:22:40.726 --> 00:22:42.435
We start to have a conversation.
00:22:42.912 --> 00:22:44.488
And it does two things, right?
00:22:44.488 --> 00:22:49.881
One, it helps me get in touch with the right people to figure out how to solve the problem and understand it better.
00:22:50.223 --> 00:22:59.575
And two, it also is a teach a fish philosophy because now these pe- now these folks in the business areas are getting involved in the work, understanding how it happens, how they need to think about it.
00:23:00.009 --> 00:23:06.307
And so over time, the concept becomes a muscle memory, and that's how I've, I opted to do it.
00:23:06.739 --> 00:23:09.078
And I think so far it's working well.
00:23:09.303 --> 00:23:16.221
But I think having only been deploying it for a year, there's a lot of opportunity for us to continue to mature
00:23:18.056 --> 00:23:22.357
Yeah, no, I think obviously, even hi- obviously hiring is a lot more...
00:23:22.385 --> 00:23:27.150
things are a lot more o- positions are opening up for a lot more specialist data governance type positions.
00:23:27.593 --> 00:23:40.315
So d- it's good that businesses are starting to recognize that this is a, yeah, a big change for companies, and the impact's gonna be massive'cause obviously, look, you got the better data quality, you, your LLMs, agents are gonna perform better
00:23:40.914 --> 00:23:41.647
Absolutely.
00:23:41.648 --> 00:23:48.771
And in some cases, there- if you deploy AI in the right way, it could even help you suss out weaknesses in your data.
00:23:48.779 --> 00:24:02.268
I see a lot of use cases where organizations are using agentic AI to help facilitate call center representatives when they're answering questions, and they need to go through process documents to figure out how to respond to a, a client inquiry.
00:24:02.269 --> 00:24:03.826
So they use the agentic AI.
00:24:04.305 --> 00:24:07.203
They, in plain language, pop in what they're looking for.
00:24:07.586 --> 00:24:24.451
The AI goes through the procedure database and says,"Oh here's a couple of documents that we think could work." The benefit in that instance, because it's well-designed and it's effectively being used as a, a query lookup tool is it's also gonna suss out where you don't have process documents, where you have deficiencies.
00:24:24.838 --> 00:24:29.513
And so then the owning business can go,"Oh, I don't have a process document for that.
00:24:29.525 --> 00:24:30.993
We don't know how to answer that question.
00:24:31.358 --> 00:24:32.371
Let's go close that gap.
00:24:32.579 --> 00:24:38.410
Let's make sure that there's a process in place." So AI is, can be very useful from a governance perspective.
00:24:38.410 --> 00:24:39.586
I know I've used it.
00:24:39.984 --> 00:24:46.328
We have Copilot at the ready for us at the bank, and I know that I use that a lot of times to, take documents.
00:24:46.586 --> 00:24:53.554
I will read it, but I'll also allow Copilot to tell me what are the risks and the gaps that, that you perceive with what we're doing here?
00:24:53.862 --> 00:24:55.104
Give me your perspective.
00:24:55.104 --> 00:24:59.415
And then I kind of fact check it against myself to see are we thinking the same thing?
00:24:59.415 --> 00:25:00.522
Are we thinking something different?
00:25:00.522 --> 00:25:02.132
Was there something I failed to consider?
00:25:02.461 --> 00:25:10.810
And there's even broader use cases on it for using it for data profiling, looking for anomalies in the data that you might need to consider uplifting a data quality role for
00:25:11.857 --> 00:25:17.288
So as you mentioned AI then, do you think it's actually been a big benefit for data governance?
00:25:17.325 --> 00:25:19.877
'Cause obviously it can, yeah, it can be a massive help
00:25:20.478 --> 00:25:23.230
I, I think it's early days.
00:25:23.268 --> 00:25:23.278
Okay.
00:25:23.334 --> 00:25:28.913
The one thing that I will say is my experience with AI has not been foolproof.
00:25:29.621 --> 00:25:34.810
For every great output, there is a,"Huh, how did we end up here?" output.
00:25:35.339 --> 00:25:38.159
So I think that there's a little bit of work to go.
00:25:38.191 --> 00:25:43.006
I think it also really depends on how good you are at prompting AI.
00:25:43.023 --> 00:25:44.797
That is really critical.
00:25:45.585 --> 00:25:52.253
And it, and when you think about it, prompting AI is the exact opposite of a Google search, right?
00:25:52.255 --> 00:25:54.252
When you go to Google, you're just like, keywords.
00:25:54.349 --> 00:25:55.640
These are the things that I'm looking for.
00:25:55.702 --> 00:25:59.646
But with AI- it's Ask Jeeves from 1997.
00:25:59.648 --> 00:26:04.971
I need you to write a full and complete sentence, and be very thorough and specific about what you're looking for.
00:26:05.318 --> 00:26:11.029
Because when you're even moderately vague, it just goes right out into left field.
00:26:11.455 --> 00:26:15.305
So I think that the, I think that the use cases are there.
00:26:15.641 --> 00:26:25.223
I think it's gonna be a matter of deploying it correctly so that we get the most value out of it, and that I think it'll alleviate a lot of time spent manually doing some of these activities.
00:26:25.830 --> 00:26:26.144
Yeah.
00:26:26.529 --> 00:26:27.469
I think it's obviously it's...
00:26:27.508 --> 00:26:28.440
like you said, it's early days.
00:26:28.463 --> 00:26:31.621
It's obviously, like with anything new, it's, there's a learning curve, isn't it?
00:26:31.625 --> 00:26:34.395
And I think it's, you, I think people have to...
00:26:34.395 --> 00:26:38.133
You gotta understand if you, if it's gonna get things wrong, you need to go on that.
00:26:38.471 --> 00:26:39.982
I guess everyone's on a journey, aren't they?
00:26:39.996 --> 00:26:42.901
So it's gonna take time to get to the level where we want.
00:26:43.613 --> 00:26:50.440
So I guess, from your perspective, what's been the hardest part changing how people think about data ownership and accountability?
00:26:51.039 --> 00:26:53.281
Oh, that is a really powerful question.
00:26:54.614 --> 00:27:03.643
I think the biggest challenge is the understanding that they- they own it, and that sometimes owning it is painful.
00:27:04.703 --> 00:27:24.877
I, I think there's a general acceptance that, I own the data and subsequently I own the risk, but sometimes what that means becomes a little bit daunting, especially when you're encountering issues or you come across a lot of issues that feel like, oh, I have this pile of problems that now I have to go and solve.
00:27:24.877 --> 00:27:26.548
And so it's...
00:27:27.220 --> 00:27:32.075
the process has really been about do- are there data quality issues?
00:27:32.211 --> 00:27:32.563
Sure.
00:27:32.782 --> 00:27:35.857
Are there data control movement pro- gaps?
00:27:36.174 --> 00:27:36.765
Probably.
00:27:37.530 --> 00:27:39.403
Do you have to solve them all at once?
00:27:40.105 --> 00:27:43.482
Not necessarily, because you can risk rate your problems.
00:27:43.483 --> 00:27:48.704
You can figure out what is the highest, most critical thing that you need to deal with, and then we focus on that.
00:27:49.185 --> 00:28:01.926
And so the biggest challenge I think is helping folks to understand how to prioritize and effectively take bite-sized chunks of the elephant versus trying to stare at the, let's solve all of the data.
00:28:02.027 --> 00:28:02.315
No.
00:28:02.368 --> 00:28:04.799
Let's parse it down and go problem by problem.
00:28:04.801 --> 00:28:21.539
And I think when you're doing AI if you don't have broad scale data governance in your organization and it hasn't expanded beyond your regulatory environment before, the thing that I would say is, don't boil the ocean when you're trying to deploy governance.
00:28:21.541 --> 00:28:23.719
Pair your use case with your governance needs.
00:28:23.719 --> 00:28:30.277
If you're, you know you're gonna use a specific data, set of data, then focus on governance on that specific data and pair it.
00:28:30.282 --> 00:28:41.518
And then eventually over time, as you build your use cases and your processes and your routine, you'll find that you'll achieve scale from a governance perspective and with AI versus trying to solve the universe at once.
00:28:41.522 --> 00:28:56.147
And I think that's what a lot of people think data management is, but it's really, it's a little bit more in the weeds and a lit- it's a little bit more let's get organized and figure out a plan of attack than it, than I think it, it's considered more broadly
00:28:57.221 --> 00:29:05.816
Yeah, and I think that's a good point,'cause I think there's so many people that they're trying to do 10, 20 different projects at once, and then they just get lost, and then obviously you've wasted time and then you're back to square one.
00:29:05.817 --> 00:29:11.409
So I think it is, like you said, focus on the most business critical problem and then just, yeah, go from there.
00:29:12.009 --> 00:29:18.030
So I guess, so what have you learned so far that you think other organizations could avoid or handle differently?
00:29:18.340 --> 00:29:25.356
I think the biggest thing that I've learned and this is a feeling that I've had for a while, is when you are...
00:29:26.564 --> 00:29:27.915
There's a couple of pieces to this.
00:29:27.921 --> 00:29:42.340
I think from a data management perspective, when you are assessing new platforms or uplifting an existing platform, you have to really challenge yourself to be uncomfortable with...
00:29:42.615 --> 00:29:43.932
To be comfortable with change.
00:29:44.497 --> 00:29:49.050
Do not focus on replicating what you did before.
00:29:49.153 --> 00:30:00.306
Focus on, how do I take this new application to tighten my architecture, to tighten my data capture, to remove these manual processes or the way that we're doing these things?
00:30:00.306 --> 00:30:10.031
I think when we look at platforms and data enablement and things like that, there's this v- this lens of, we're very comfortable with the process and routine we have.
00:30:10.031 --> 00:30:10.912
We don't wanna shake it.
00:30:11.596 --> 00:30:14.183
And maybe you can consider that people over process.
00:30:14.900 --> 00:30:24.883
But I think you have to be un- You have to allow yourself to be uncomfortable with the change, but get through the change, allow it to simplify.
00:30:25.273 --> 00:30:34.648
And then I think the other piece of the equation is as you start to evolve your ecosystems, always consider where the data's going, how it's being used, what it's being used for.
00:30:35.101 --> 00:30:38.894
We, we do a really strong job of that in many cases.
00:30:39.292 --> 00:30:41.167
But every once in a while, there might be a miss.
00:30:41.180 --> 00:30:45.643
It's a challenging environment, especially the more complex your architecture is.
00:30:46.174 --> 00:30:59.196
But if you're always thinking about not just, how do I get the job done, but also, do I have the, the data that I need to assess that I did the job correctly, that the things that I put in place are working, that I've...
00:30:59.963 --> 00:31:08.920
that there's no potholes, no scenarios where bad data can get entered or we've mitigated that, I think that you're gonna be on the right side of the law.
00:31:08.925 --> 00:31:22.372
But that attitudinal shift comes with time because you also are counterweighted with the idea of project timelines and budgetary costs and making sure that you're coming in on time and on budget or under budget.
00:31:22.428 --> 00:31:28.153
And so it's a very delicate balance depending on what you're trying to do and how much tech debt you carry.
00:31:28.450 --> 00:31:41.185
But it is a thing that folks probably should start to think about more clearly and more consistently, especially if they're using a data lake or they're mastering data at their organization and considering all those integration points.
00:31:41.471 --> 00:31:45.398
That I think is gonna be super critical for them in enabling them going forward.
00:31:46.910 --> 00:31:48.811
And you mentioned Lisa, you took on this role.
00:31:48.971 --> 00:31:53.374
Okay, so is any- has there been anything that's turned out to be a lot harder than you expected?
00:31:53.723 --> 00:31:58.950
I think the hardest thing is getting consensus amongst stakeholders, and I think that's a pretty common thing.
00:31:59.250 --> 00:31:59.426
Yeah.
00:31:59.476 --> 00:32:07.451
I'm not just the, the data owner of consumer, but I am also the data owner of our customer master, which is a really awesome thing that we're doing.
00:32:07.941 --> 00:32:09.593
But it has its own set of challenges.
00:32:09.594 --> 00:32:11.621
We're building a, a data management framework.
00:32:11.623 --> 00:32:17.425
It, it involves a lot of people, a lot of data, all, you know, and we wanna make sure it's all cared for correctly.
00:32:17.426 --> 00:32:25.933
But getting people to care about the same things that I care about at the same pace that I wanna care about them, that I think is the most challenging piece.
00:32:25.961 --> 00:32:26.769
But I...
00:32:27.498 --> 00:32:35.009
If you can create a good call to action that people can rally around, it makes all the difference.
00:32:35.015 --> 00:32:40.671
But sometimes that's not always there, and that is what I think creates the biggest challenge.
00:32:42.078 --> 00:32:42.834
Okay, cool.
00:32:43.075 --> 00:32:50.958
And I guess then as AI has accelerated, how has that changed the importance of getting the data foundations right?
00:32:51.258 --> 00:32:53.451
It, that is an interesting question.
00:32:53.461 --> 00:33:04.578
So I think there is an idea that we do need to get the data foundations right, but I think there's also a lot of thought going around what does good governance look like for AI?
00:33:05.771 --> 00:33:07.179
You have your data quality rules.
00:33:07.230 --> 00:33:13.634
You can make sure that the data is moving from point A to point A- B, but once it goes into the model, what does governance look like?
00:33:14.087 --> 00:33:19.339
So I think that there's a lot of conversation and thought happening about that.
00:33:19.830 --> 00:33:42.474
But I think at its core, what we're trying to make sure that we're doing is, at least at the bare minimum, if we've got use cases that we've prioritized to develop, making sure that the right people are involved from the areas that are supporting from a data perspective so that we can at least get a baseline sense of health before they get too far deep into the actual model build.
00:33:42.854 --> 00:34:02.665
Beyond that, I think we're gonna continue to evolve in terms of, what other steps of data management are required to make sure that once it starts to go through the model training, that it's still producing what we feel are correct and desirable outcomes versus, anything that looks like it's skewed or it's hallucinating.
00:34:02.909 --> 00:34:20.702
And I think the other, other-- the other piece of it too is depending on the use case, if we have to use synthetic data, in which is sometimes a thing we-- you can do in an AI model making sure that's not cur-- we're using the right com- composite of synthetic data in order to achieve the right results.
00:34:21.106 --> 00:34:40.543
I think there's a lot more thought around AI in the financial industry because it will impact a person in a real world way because we are caring for their money, and we are determining- their credit worthiness, if they wanted to get a house or a home equity loan or a credit card.
00:34:40.547 --> 00:34:50.943
So we really do have to get data management correct in this space because of the implications that it has on a person's day-to-day life.
00:34:51.293 --> 00:35:01.119
So we are thinking it through, but I would say that how to deploy it end to end from a management perspective is still something that we're hashing out the finer points on.
00:35:01.130 --> 00:35:06.800
We got the foundation of what we need to do, but to pull it through end to end I think is still a work in progress.
00:35:07.567 --> 00:35:07.878
Yeah.
00:35:08.494 --> 00:35:16.356
And like you mentioned, obviously, look, in the financial services, insurance, pharma, these are, the data is just so sensitive, isn't it?
00:35:16.356 --> 00:35:18.726
Like you can't have that coming out.
00:35:19.032 --> 00:35:21.661
It's just gonna- Well- yeah, cause all sorts of problems
00:35:22.681 --> 00:35:23.773
A- absolutely.
00:35:23.786 --> 00:35:35.675
There are so many implications, especially- Yeah when we think about the, the history and the housing market and the way that credit decisioning occurred with people of color versus, Caucasian individuals.
00:35:35.675 --> 00:35:43.119
There was a heavily skewed pattern towards people of cov- color where they got unfavorable lending rates in comparison to other groups.
00:35:43.119 --> 00:35:51.809
And if you try to train a model to determine credit worthiness and you're using historical information- Yeah you theoretically could skew.
00:35:51.826 --> 00:36:01.583
There's laws and regulations to avoid that, and, banks on a, the annual basis need to demonstrate that they aren't demonstrating that kind of bias.
00:36:01.606 --> 00:36:07.586
But that's not to say that if you're training a model on 20 years worth of data that it's not gonna inherently come through.
00:36:07.896 --> 00:36:32.353
And those are the things, that's the root of why we have to be so cautious, because historically we have done some of these things, and the, the federal government and regulators have come down and said,"No, that's not the right thing to do." And so we have to be very- cautious with AI in those types of spaces to make sure that we are not accidentally discriminating somebody without ooh, with no rhyme or reason.
00:36:33.474 --> 00:36:35.922
And even at that point you said that's, even that's a massive job, isn't it?
00:36:35.927 --> 00:36:42.019
In just going past historical data,'cause getting round that is it's a lot of thinking that goes into it, I guess
00:36:43.161 --> 00:36:51.106
It does, and depending on how much historical data you need, if what you're capturing today was what you were capturing previously.
00:36:51.344 --> 00:36:55.293
What we collect has evolved over time and distance, and if you're...
00:36:55.364 --> 00:37:01.072
You might need to plug holes in historical data depending on again, how much you need to train the model.
00:37:01.400 --> 00:37:04.175
And so that goes back to the point around synthetic data.
00:37:04.186 --> 00:37:12.255
You also have to be very con- very careful of the synthetic data that you use so that it doesn't detrimentally impact the model as well.
00:37:12.255 --> 00:37:12.445
There...
00:37:12.461 --> 00:37:26.101
it's quite a dance, I would say, when it comes to training the- these models because what we've needed to do with data over time has changed, and what we've needed from our data has changed
00:37:26.702 --> 00:37:27.936
I guess yeah, no, I agree.
00:37:27.936 --> 00:37:30.391
Obviously it's like Slava said earlier, it's constant evolution.
00:37:30.391 --> 00:37:32.949
It's just things con- constantly changes.
00:37:33.324 --> 00:37:41.186
And this question may be a bit difficult, but I guess what does good data ownership look like when it's actually working inside like a, a large organization like yourself?
00:37:41.786 --> 00:37:55.797
So while I, while we're only a year in to this experiment, what I think the way that I perceive it is good g- data governance is, and good data ownership is taking accountability.
00:37:56.141 --> 00:37:57.394
I'm a big believer in that.
00:37:57.394 --> 00:38:02.358
I'm the first person to stand tall and deal with a problem if no one else wants to raise their hand.
00:38:02.358 --> 00:38:18.918
And I think with the model that we are working to deploy, at least in my space,'cause my peer data owners may have different models to support their areas in the way that makes sense for them I think it's also helping to create a point of accountability within each of those areas too.
00:38:19.226 --> 00:38:24.733
I've already started to see folks coming to me and saying,"Hey I noticed this.
00:38:24.829 --> 00:38:26.851
Is this the right thing to have happen?
00:38:26.864 --> 00:38:27.615
Should we do X?
00:38:27.623 --> 00:38:42.733
Should we do Y?" And so that's actually made me really excited because that means that all of the work that we've done over the, the last 12 months, regardless of whether people were talking or quiet on the phone, like the message is starting to resonate.
00:38:42.751 --> 00:38:47.563
Now people are individually standing tall and taking accountability and saying,"Hey, I found an issue.
00:38:47.565 --> 00:38:55.559
Let's go figure out how to fix it." And that is how I would love to see data management continue to evolve.
00:38:55.559 --> 00:38:58.893
It's not a, it's not a one-person problem per se.
00:38:59.297 --> 00:39:03.123
There's one person to help rally the troops, and I'm happy to do that all day long.
00:39:03.621 --> 00:39:15.018
But the real benefit is gonna come from what I call data citizens across the organization, identifying issues and knowing who to go to raise the problem to find the solution.
00:39:15.380 --> 00:39:29.367
And that way it turns into this self-propelling data management engine that doesn't fall on anyone's shoulders because we all recognize we, we have table stakes in the management of our data because it, it drives our business decisions.
00:39:29.702 --> 00:39:32.605
It d- dri- it drives how we think about what we're doing.
00:39:32.913 --> 00:39:46.807
And it makes you care a little bit more about what you're putting into your Excel spreadsheet or into your application because you un- understand the costs and the work that come with fixing it if you don't do it the right way, or at least try to do it the right way
00:39:48.304 --> 00:39:56.849
Now you mentioned, you've obviously, you're earlier, obviously earlier in this journey, but are have you seen impacts, like, on whether it's customer trust, business value so far?
00:39:58.500 --> 00:40:05.827
I, I think that we are definitely starting to see the impacts on business value when we think about the, the, the customer level work that we're doing.
00:40:05.829 --> 00:40:09.996
It's going to be I think, a big game changer for us.
00:40:10.507 --> 00:40:12.175
And it's helping us to...
00:40:12.186 --> 00:40:15.507
It's gonna really help us to understand our customer base.
00:40:15.510 --> 00:40:37.936
I know that, we've socialized publicly some of our broader goals, and one of it is really around understanding and deepening our relationships with our clients, that the care and the maintenance of that master is gonna be huge in doing that because we're gonna really start to understand who they are, how they bank with us, what they bank with us for and helping us to suss out those strategic areas of opportunities.
00:40:37.936 --> 00:40:50.385
And so caring for that data, making sure it's healthy and whole is gonna be really critical in helping to create better products create better opportunities for clients and, increase satisfaction.
00:40:50.400 --> 00:40:54.594
It'll improve, it'll continue to im- to strengthen our reputation in the industry.
00:40:55.043 --> 00:41:02.088
So but it, it all comes down to, great partnership, but also making sure that you've got strong data management because it...
00:41:02.088 --> 00:41:10.306
you're bringing together a tremendous amount of information and, trying to make sure that you're pairing it to the right people to drive the right outcomes.
00:41:11.438 --> 00:41:11.766
Okay.
00:41:12.463 --> 00:41:18.891
And then if, say, if there's a le- another leader who's starting from where you started, what would you tell them to do?
00:41:18.891 --> 00:41:22.806
Say, over the next 90 days, what would be-- what would you encourage them to do?
00:41:22.916 --> 00:41:26.934
And what, what are the learnings you'd apply, say, starting fresh?
00:41:28.358 --> 00:41:35.195
I think if you're starting fresh from a data management perspective, I would really I would really spend the first 90 days listening.
00:41:35.476 --> 00:41:45.797
Meeting with partners, understanding the pain points, what their challenges are before you start developing a model, developing a strategy, developing a plan.
00:41:46.126 --> 00:41:54.954
I think if you are pursuing AI and you have use cases or things that you're thinking through, it's a similar fashion.
00:41:54.956 --> 00:42:01.809
You kinda, you listen to the ideas, you start to, to discuss what the data is that you're gonna be using, and then you figure out a plan for governor- governance.
00:42:02.423 --> 00:42:12.429
Consider an AI readiness strategy based on your use cases and how many data sources you need to use for it, and determining their readiness to actually be deployed in a model.
00:42:12.431 --> 00:42:26.880
And then you can figure out and plot the course, one, on how you wanna rally your data management function around that, and then two, what's the order of operations that you need to do to get the data where it needs to be to accelerate your AI use cases.
00:42:26.880 --> 00:42:29.963
And for some, it might be a little bit, for some it might be a lot.
00:42:29.976 --> 00:42:40.447
But I think if you start to put those things hand in glove, and then understanding what's going on within the business area itself, it's gonna be, it's gonna be a powerful thing
00:42:41.047 --> 00:42:48.467
So then looking ahead, do you think data ownership and governance will become one of the biggest areas of investment as AI continues to scale?
00:42:48.818 --> 00:42:55.594
I think it's going to depend on the organization and where they are at from a data management pos- per- perspective.
00:42:55.594 --> 00:43:01.483
I think right now we are investing in it, but there are still a lot of things that still need support.
00:43:02.318 --> 00:43:06.507
I am-- I'm less concerned with heavy investment.
00:43:06.534 --> 00:43:08.869
I'm more concerned with recurring investment.
00:43:08.876 --> 00:43:32.728
What I would love to see over time is not that we're throwing$50 million to solve our data management issues, but that every business area has a$10 million budget to manage their data and address issues in that year versus,"Oh, we have a problem, but we have to wait for a project," or we have to, we have to wait for an issue to crop up somewhere.
00:43:33.159 --> 00:43:34.438
The, I w- I would...
00:43:34.503 --> 00:43:40.519
An ounce of pre- prevention is worth a pound of cure, and that's where I would love to see it, it go over time.
00:43:40.905 --> 00:43:55.284
Until then, I think it's gonna be, a heavily projectized approach until everybody kind of plots out what they wanna do from an AI perspective and then align it to what they need to do from a data management perspective.
00:43:55.590 --> 00:43:58.478
And we, I think, are still very early days there.
00:44:00.184 --> 00:44:03.659
Yeah, no, I think, there's still gonna be investment within building models.
00:44:03.663 --> 00:44:13.331
I think that's obviously a big one for companies, but also I do think there'll be more investment within data governance, data ownership,'cause I think it's gonna become more crucial for businesses.
00:44:13.759 --> 00:44:14.315
But it's not...
00:44:14.315 --> 00:44:17.277
it's like you said it's gonna be a, it's g- it's a longer term project, isn't it?
00:44:17.277 --> 00:44:19.485
It's not one of them you can deliver quick
00:44:21.373 --> 00:44:22.393
A thousand percent.
00:44:22.393 --> 00:44:29.211
And what I would say also is I think the regulators are starting to ask the question as well is, what are we doing from a data management perspective?
00:44:29.211 --> 00:44:30.798
So I do agree with you.
00:44:30.798 --> 00:44:32.168
I think there will be investment.
00:44:32.277 --> 00:44:35.677
What it looks like will probably take different forms and shapes.
00:44:35.690 --> 00:44:38.974
Some of it might be architectural rationalization.
00:44:39.389 --> 00:44:43.672
Some of it might be going in and actually cleaning up fields and putting in controls.
00:44:43.677 --> 00:44:45.403
It takes a lot of forms and fashion.
00:44:46.050 --> 00:44:47.387
But it is gonna take time.
00:44:47.480 --> 00:44:52.230
But I think, we'll get spurred along because our regulatory partners care.
00:44:52.231 --> 00:44:54.246
We care about getting it right.
00:44:54.605 --> 00:45:02.577
So I think that we're-- we will but we have to, again, we have to prioritize what are the most critical things that have to get addressed first.
00:45:02.583 --> 00:45:09.059
So it might not necessarily be the stuff that I'm doing because we have to deal with bigger technological challenges first.
00:45:09.059 --> 00:45:09.644
But as long...
00:45:09.711 --> 00:45:19.000
A-as far as I'm concerned, as long as we're hitting where we wanna go from a target state, how we get there, it's the it's gonna be the journey that it's gonna be.
00:45:19.320 --> 00:45:23.039
I'm more interested in the outcome of landing where we need to be from a data perspective.
00:45:24.134 --> 00:45:24.452
Brilliant.
00:45:24.615 --> 00:45:28.911
And then I guess finally, if anyone's got any questions for you, is it best to hit you up on LinkedIn?
00:45:30.000 --> 00:45:30.755
Absolutely.
00:45:30.757 --> 00:45:40.119
I know I'm probably not the most timely person on responding- but if you hit me up in the chat that's probably the best place to reach out to me.
00:45:40.387 --> 00:45:44.753
My email sometimes routes things to, to junk mail, and I don't want anybody to get trapped in that limbo.
00:45:45.688 --> 00:45:46.085
No worries.
00:45:46.159 --> 00:45:51.797
And I'll, look, I'll come mention your put a link to your LinkedIn, so if anyone does have any questions, they can connect out to you.
00:45:52.309 --> 00:45:55.304
So yeah, I guess thank you for joining us, Seric.
00:45:55.320 --> 00:45:59.469
It's been a pleasure having you on and yeah, some great advice and insight shared
00:46:00.822 --> 00:46:01.246
Absolutely.
00:46:01.275 --> 00:46:03.338
Again, I appreciate you having me.
00:46:03.338 --> 00:46:04.608
This was really fantastic