WEBVTT
1
00:00:00.160 --> 00:00:03.439
Credit scores and credit reports are a record of what
2
00:00:03.520 --> 00:00:06.320
has happened. They are not a record of what will happen.
3
00:00:06.599 --> 00:00:08.800
You may not know exactly what the rules are. You
4
00:00:08.839 --> 00:00:11.119
may not know exactly what the rules will be. You
5
00:00:11.240 --> 00:00:14.800
do know that you should understand how models work.
6
00:00:14.880 --> 00:00:16.519
People want to know the answer right away if they
7
00:00:16.519 --> 00:00:18.640
don't want to wait ten minutes. Some days it's in
8
00:00:18.760 --> 00:00:20.839
some done frames just to then get declined, anything just
9
00:00:20.839 --> 00:00:21.440
to get to climb.
10
00:00:21.559 --> 00:00:24.519
Approval rates matter more than interest rate in today's work.
11
00:00:24.760 --> 00:00:26.960
I think I would say facts.
12
00:00:31.280 --> 00:00:33.200
Hi, Welcome to Leers and lending. Today. We're going to
13
00:00:33.240 --> 00:00:36.679
be talking about how credit unions can manage risk through
14
00:00:36.679 --> 00:00:40.200
smarter lending decisions and how that works within the regulatory
15
00:00:40.280 --> 00:00:44.479
environment that credit unions have. Talking to Drew and Linn today,
16
00:00:45.640 --> 00:00:48.840
managing lending risk has never been more complex we have.
17
00:00:49.679 --> 00:00:54.280
If you think of the how lending regulations have evolved
18
00:00:54.280 --> 00:00:57.079
through the years. You know, TILA came out in like
19
00:00:57.119 --> 00:01:00.640
the seventies, and then you know we all the way
20
00:01:00.679 --> 00:01:03.719
to sort of Dodd Frank in twenty ten, and it
21
00:01:03.840 --> 00:01:06.280
just seems there's always more and more regulations that are
22
00:01:06.280 --> 00:01:10.280
coming up for credit unis and banks to have to
23
00:01:10.319 --> 00:01:15.040
deal with. So, how do we get regulators comfortable with
24
00:01:15.200 --> 00:01:19.640
partnering with FinTechs and those FinTechs specifically that we leveraging
25
00:01:20.079 --> 00:01:23.560
AI based underrating models thinking of the regulatory environment that
26
00:01:23.760 --> 00:01:24.680
we're dealing with today.
27
00:01:24.799 --> 00:01:27.920
Yeah, it's a good question, and I think it's a
28
00:01:28.000 --> 00:01:31.680
shifting regulatory environment on a day to day basis. I
29
00:01:31.719 --> 00:01:33.840
think all of us have certainly watched the changes with
30
00:01:35.000 --> 00:01:40.640
the various financial regulators this year than now some court
31
00:01:40.719 --> 00:01:42.680
challenges to the leadership and who's on the board and
32
00:01:42.680 --> 00:01:44.400
who's not on the board, and what the rules are.
33
00:01:44.519 --> 00:01:48.400
So I would recommend for a credit union or really
34
00:01:48.400 --> 00:01:52.120
any lending institution, like just focus on the fundamentals. You
35
00:01:52.200 --> 00:01:54.799
may not know exactly what the rules are, You may
36
00:01:54.799 --> 00:01:57.640
not know exactly what the rules will be, but you
37
00:01:57.799 --> 00:02:02.560
do know that you should understand how models work. You
38
00:02:02.560 --> 00:02:05.480
should understand who owns which part.
39
00:02:05.280 --> 00:02:05.920
Of the decision.
40
00:02:06.439 --> 00:02:09.280
You should do your due diligence on any vendors that
41
00:02:09.319 --> 00:02:12.639
you partner with in a traditional way where you're looking
42
00:02:12.680 --> 00:02:20.039
at compliance risk, financial health, reputational risk, like even though
43
00:02:20.080 --> 00:02:22.400
those things are may not be explicitly called out anymore,
44
00:02:22.400 --> 00:02:26.199
like reputational risk as an example, that doesn't mean that
45
00:02:26.240 --> 00:02:27.000
you don't want.
46
00:02:26.840 --> 00:02:27.919
To do that for your business.
47
00:02:27.960 --> 00:02:31.479
So think about from a core business standpoint, what do
48
00:02:31.560 --> 00:02:33.919
you care about? What do you think the key areas
49
00:02:33.919 --> 00:02:37.039
are that you should be able to both understand and
50
00:02:37.080 --> 00:02:40.479
document and then how can you up level your organization
51
00:02:40.719 --> 00:02:45.039
to really be confident in understanding and monitoring AI models?
52
00:02:45.280 --> 00:02:48.639
And I think that's a very big challenge. And I
53
00:02:48.639 --> 00:02:51.319
think similar to a lot of conversations we've had about
54
00:02:51.360 --> 00:02:54.919
different sized organizations, Like there's choices of partnering. There are
55
00:02:54.960 --> 00:02:57.439
many vendors, there are many consultants who.
56
00:02:57.280 --> 00:02:58.560
Help in this space today.
57
00:02:58.639 --> 00:03:02.400
So you may not be able to hire deep AI
58
00:03:02.560 --> 00:03:05.240
experts internally you don't need to, but can you find
59
00:03:05.280 --> 00:03:07.719
trusted partners that can help you do that evaluation and
60
00:03:07.879 --> 00:03:11.719
help and opportunities for your team to become to kind
61
00:03:11.719 --> 00:03:14.240
of deepen their subject matter expertise as it relates to.
62
00:03:14.479 --> 00:03:15.080
AI and.
63
00:03:17.000 --> 00:03:20.479
New and more innovative ways of evaluating risk and lending.
64
00:03:20.719 --> 00:03:23.039
Yeah, and that's a good like having a good general
65
00:03:23.159 --> 00:03:26.000
high level knowledge base of all things AI. I mean,
66
00:03:26.000 --> 00:03:29.840
it's the topic at hand most often anymore, right, but
67
00:03:30.039 --> 00:03:32.120
digging even deeper. So if you are partnered with some
68
00:03:32.240 --> 00:03:35.840
type of institution that is driving AI and some metric.
69
00:03:36.319 --> 00:03:39.199
How can you get more knowledge base into your internal
70
00:03:39.240 --> 00:03:42.879
teams on this company, does X, Y, and Z we
71
00:03:43.000 --> 00:03:45.039
kind of understand the full rails of it. Maybe we
72
00:03:45.039 --> 00:03:48.520
don't understand all of it, but we have confidence in
73
00:03:48.560 --> 00:03:51.759
going into a regulatory exam or speaking to regulators of
74
00:03:51.800 --> 00:03:55.159
why we're comfortable with this approach and having that for
75
00:03:55.240 --> 00:03:58.319
every single vendor that you have that leverages AI, so
76
00:03:58.360 --> 00:04:00.599
you have the high level knowledge base and then you
77
00:04:00.639 --> 00:04:02.960
have all the double clicks and everything else you're using
78
00:04:02.960 --> 00:04:03.199
it for.
79
00:04:03.879 --> 00:04:06.199
Right, So you don't have to be the expert thing.
80
00:04:06.280 --> 00:04:09.240
You just have to partner with experts in some respects.
81
00:04:09.319 --> 00:04:10.919
Yeah, well, I mean I think you have to have
82
00:04:11.719 --> 00:04:13.879
some level of expertise. But like if we go back
83
00:04:13.919 --> 00:04:18.839
to like thinking about how say phycostore example is created,
84
00:04:20.639 --> 00:04:22.839
could you, as a credit union executive, tell me how
85
00:04:22.920 --> 00:04:26.240
that number is calculating exactly? So I think think about
86
00:04:26.240 --> 00:04:28.879
it that way, like you know, putting an oversight of
87
00:04:28.920 --> 00:04:32.160
an AI lender, an AI vendor, or an AI model
88
00:04:32.480 --> 00:04:35.639
into your overall model risk management framework, Like how do
89
00:04:35.639 --> 00:04:37.079
you think about managing model risk?
90
00:04:37.120 --> 00:04:38.800
How do you think about evaluating.
91
00:04:38.319 --> 00:04:40.920
Model risk that does not mean that you need to
92
00:04:40.920 --> 00:04:43.439
be able to build, build, and recreate the entire thing,
93
00:04:43.800 --> 00:04:46.240
but you have to be able to articulate and understand
94
00:04:46.279 --> 00:04:49.439
it and are doing enough kind of due diligence on
95
00:04:49.480 --> 00:04:53.040
your partner to feel confident in their capabilities and their depth.
96
00:04:53.199 --> 00:04:55.399
Yeah, and if you get the further digged in details,
97
00:04:55.439 --> 00:04:58.720
you're able to create risk assessments if XYZ and happen.
98
00:04:58.879 --> 00:05:02.279
If XYZ happened, we mitigate this way, and that just
99
00:05:02.319 --> 00:05:05.839
gives more comfort to regulators on why you're able to
100
00:05:05.879 --> 00:05:08.399
adopt those types of rights of partnerships.
101
00:05:08.600 --> 00:05:11.240
So shifting gears a little bit. How do you believe
102
00:05:11.240 --> 00:05:14.240
the credit unions are using data today to sort of
103
00:05:14.279 --> 00:05:17.120
balance unsecured versus secured lending?
104
00:05:17.279 --> 00:05:20.600
I would say from an organic perspective, if it's an
105
00:05:20.600 --> 00:05:23.920
existing member base that's coming through to get unsecured. The
106
00:05:24.000 --> 00:05:27.240
biggest thing from a data perspective today is the credit
107
00:05:27.360 --> 00:05:30.319
unions have a large source of data for that consumer.
108
00:05:30.759 --> 00:05:33.040
They have their in and outs via AH, they know
109
00:05:33.120 --> 00:05:37.120
their cash flows, they understand their spending habits. Just because
110
00:05:37.199 --> 00:05:39.720
DTI percent is X does not mean that they're an
111
00:05:39.720 --> 00:05:42.879
actual DTI percent of X, right, So they're able to
112
00:05:42.959 --> 00:05:45.360
leverage that from a decisioning basis. Then you think about
113
00:05:45.399 --> 00:05:47.120
all of that data and different rails that you can
114
00:05:47.120 --> 00:05:50.360
put it in. You're getting an output of pre approval
115
00:05:50.480 --> 00:05:53.360
or denial at a snap of a finger. So I
116
00:05:53.360 --> 00:05:56.879
think that's one real good aspect of comparing underwriting today
117
00:05:57.040 --> 00:05:58.639
compared to the prior.
118
00:05:59.040 --> 00:05:59.720
Yeah, I'm true.
119
00:06:00.079 --> 00:06:04.560
If you have their main deposit account, their main transaction account,
120
00:06:04.879 --> 00:06:08.040
you can learn a lot about a person's behaviors or
121
00:06:08.240 --> 00:06:10.399
next like does it you know, are they buying it?
122
00:06:10.480 --> 00:06:12.560
Did they recently get married? Did they maybe open a
123
00:06:12.639 --> 00:06:13.360
joint account?
124
00:06:13.399 --> 00:06:16.480
Like can you did they move? And can you you know?
125
00:06:16.519 --> 00:06:20.959
Will they need potentially additional types of of lending to
126
00:06:21.000 --> 00:06:23.680
support that. So I think using that, using that data
127
00:06:23.720 --> 00:06:26.720
that they've had and have rich sources of, and using
128
00:06:26.720 --> 00:06:28.839
it to figure out how they're lending or engaging with
129
00:06:28.879 --> 00:06:29.399
those people.
130
00:06:29.519 --> 00:06:32.199
And even if they're not a depository or using a
131
00:06:32.279 --> 00:06:35.079
depository product, you can paint a picture from a credit
132
00:06:35.120 --> 00:06:37.600
report and you don't need an underwriter to sit there
133
00:06:37.600 --> 00:06:40.519
and look at the fifty different tradelight lines that were
134
00:06:40.560 --> 00:06:42.839
opened over the past twenty years to paint that picture.
135
00:06:43.240 --> 00:06:45.920
You can tell that that borrower has never missed a payment,
136
00:06:46.120 --> 00:06:49.920
or maybe they'll miss a payment at month on book
137
00:06:50.079 --> 00:06:52.560
thirty six, but things start to level back out. You
138
00:06:52.560 --> 00:06:55.120
can kind of get sense of that. And what's nice
139
00:06:55.160 --> 00:06:57.959
now is with the use of AI and with automated fashions,
140
00:06:57.959 --> 00:07:00.160
you don't have to sit and read it. You can
141
00:07:00.199 --> 00:07:02.439
get an output again going back to kind of like
142
00:07:02.480 --> 00:07:04.800
the cash flows of their depository product to be able
143
00:07:04.800 --> 00:07:07.040
to paint that picture again on a snap up, because.
144
00:07:06.839 --> 00:07:07.439
It's just data.
145
00:07:07.519 --> 00:07:08.079
It's just there.
146
00:07:08.199 --> 00:07:10.079
It's just a bunch of data. And it's how you
147
00:07:10.120 --> 00:07:12.480
actually ingest that data into some sort of an AI
148
00:07:12.600 --> 00:07:16.920
system that will take data and turn into really live information.
149
00:07:16.920 --> 00:07:18.759
That's yes, were for you. You know, it's really good at
150
00:07:18.759 --> 00:07:22.000
this are the credit card companies right, absolutely, because and
151
00:07:22.160 --> 00:07:25.079
we're talking on secure lending, right, so they have a
152
00:07:25.160 --> 00:07:29.319
lens into purchasing behavior. They know that you booked a
153
00:07:29.319 --> 00:07:33.199
flight to Columbus, Ohio, and so they would expect, well,
154
00:07:33.199 --> 00:07:35.279
there should be a hotel charge in Columbus, Ohio, and
155
00:07:35.360 --> 00:07:38.920
restaurant charges and so on, and so instead of now
156
00:07:38.959 --> 00:07:42.040
I'm getting declined because I'm no longer and you know
157
00:07:42.240 --> 00:07:44.720
close to home. I'm two thousand miles from home. Credit
158
00:07:44.759 --> 00:07:46.839
card companies have sort of figured this out. I mean,
159
00:07:46.959 --> 00:07:49.600
is this something that is there an opportunity here for creditings?
160
00:07:50.600 --> 00:07:54.959
Many creditings don't own their credit card balances anymore, but
161
00:07:55.040 --> 00:07:57.800
they do have partnerships where there should be some sort
162
00:07:57.839 --> 00:07:59.079
of sharing of that data.
163
00:07:59.120 --> 00:08:01.879
Sure, and at about years ago, when you, particularly if
164
00:08:01.879 --> 00:08:05.160
you traveled internationally, you had to notify your credit cards.
165
00:08:05.680 --> 00:08:06.439
Sometimes you still do.
166
00:08:06.480 --> 00:08:08.040
You had to go in and say here's where I'm
167
00:08:08.040 --> 00:08:10.519
going on these dates, or you risk your credit card
168
00:08:10.519 --> 00:08:13.920
not working when you got on vacation. And now that
169
00:08:14.079 --> 00:08:16.639
it happens organically because they're using the data to make
170
00:08:16.680 --> 00:08:18.279
those make those assumptions.
171
00:08:19.079 --> 00:08:20.360
Yeah, I do think that's interesting.
172
00:08:20.360 --> 00:08:22.079
I think particularly with me, and I don't want to
173
00:08:22.399 --> 00:08:25.160
don't want to segue too far into the open banking
174
00:08:25.839 --> 00:08:29.800
kind of quagmire that's happening today because there's a lot
175
00:08:29.800 --> 00:08:32.840
of competing conversations around it. But I think that's a
176
00:08:32.879 --> 00:08:35.519
good question of like who owns the data? So they
177
00:08:35.559 --> 00:08:39.200
may be partnering with a credit card vendor to extend
178
00:08:39.200 --> 00:08:43.320
that product to their members, but that is still their member.
179
00:08:43.360 --> 00:08:45.840
They still hold the economic risk. And so because they
180
00:08:45.879 --> 00:08:48.159
hold that economic risk, what level of data do they
181
00:08:48.159 --> 00:08:50.840
have access to and how can they use that, particularly
182
00:08:50.919 --> 00:08:52.960
to make new offers. And I think that's I think
183
00:08:53.039 --> 00:08:57.519
is a lot of this open banking. Like the legal
184
00:08:57.600 --> 00:08:59.799
challenges play out too, I think we'll see more clarity
185
00:09:00.120 --> 00:09:02.200
und what they can do in the future.
186
00:09:02.279 --> 00:09:07.799
Do you think overall that data driven models outperform traditional linderwriting.
187
00:09:07.600 --> 00:09:11.080
Yes, being biased since we come from upstart, of course
188
00:09:11.080 --> 00:09:13.759
we use data driven models, but we're able to see
189
00:09:13.879 --> 00:09:17.440
and it's not just upstarted across the board is you
190
00:09:17.559 --> 00:09:22.120
have borrowers that under traditional underwriting guidelines probably wouldn't qualify
191
00:09:22.159 --> 00:09:24.639
for a credit union type of credit, right, whether that
192
00:09:24.720 --> 00:09:28.240
be a secured versus an unsecured type of asset. Whereas
193
00:09:28.559 --> 00:09:31.519
these types of models are able to separate risk much
194
00:09:31.559 --> 00:09:33.679
better than an underwriter sitting in front of a computer
195
00:09:34.000 --> 00:09:38.159
or some automated prequal type of logic that, Okay, maybe
196
00:09:38.159 --> 00:09:42.200
this six forty FICO borrower we would decline, but they're
197
00:09:42.279 --> 00:09:44.720
really transacting more of like a prime type of borrower.
198
00:09:44.759 --> 00:09:47.519
But they're too new into the system take a chance
199
00:09:47.519 --> 00:09:49.960
on them, right, So I would be very biased in
200
00:09:50.000 --> 00:09:52.320
saying that, yes, they're better separators of risk.
201
00:09:52.519 --> 00:09:54.279
Yeah, And you think about and We've talked about this
202
00:09:54.320 --> 00:09:58.600
a lot at at our company. Of course, where credit
203
00:09:58.720 --> 00:10:01.480
scores and credit report sports are a record of what
204
00:10:01.559 --> 00:10:04.000
has happened, they are not a record.
205
00:10:03.720 --> 00:10:04.679
Of what will happen.
206
00:10:05.799 --> 00:10:08.600
And while you know, we'll put the disclaimer like historical
207
00:10:08.639 --> 00:10:09.639
results are in an.
208
00:10:10.080 --> 00:10:13.559
Indicator of a future, there certainly.
209
00:10:13.200 --> 00:10:16.639
Is that idea that that you know, if you've behaved
210
00:10:16.639 --> 00:10:19.000
a certain way, particularly if you behave really risky you
211
00:10:19.039 --> 00:10:22.240
could expect risky behavior. But like you're to the point
212
00:10:22.279 --> 00:10:25.080
you used, what if it's a person who is, you know,
213
00:10:25.399 --> 00:10:29.240
a new citizen in the US or a recent college
214
00:10:29.240 --> 00:10:32.480
student who maybe didn't have the benefit of their parents
215
00:10:32.519 --> 00:10:34.840
helping them build that credit profile and that credit score
216
00:10:34.919 --> 00:10:38.600
growing up, but they have a a good job and
217
00:10:38.639 --> 00:10:42.799
they're making money, and they are are most likely going
218
00:10:42.840 --> 00:10:44.639
to pay in the future. How do you use that
219
00:10:44.759 --> 00:10:48.200
data to lend to those borrowers where they may in
220
00:10:48.279 --> 00:10:52.039
traditional kind of rate rate based models may may not
221
00:10:52.080 --> 00:10:52.600
be lent to.
222
00:10:53.200 --> 00:10:55.360
Another example of that, too, is going back to my
223
00:10:55.799 --> 00:10:59.639
previous kind of comment, is the credit report has this
224
00:10:59.639 --> 00:11:03.039
this whole slew of everything that this individual has transacted on.
225
00:11:03.399 --> 00:11:05.240
Maybe there was a life changing event that took them
226
00:11:05.240 --> 00:11:07.480
into that six point forty. But they've always been a
227
00:11:07.519 --> 00:11:11.039
super prime type of consumer. But something happened along along
228
00:11:11.080 --> 00:11:14.679
the way, and to your point, forward thinking, this borrower
229
00:11:14.759 --> 00:11:18.480
is probably going to transact that same way. But death, divorce,
230
00:11:18.639 --> 00:11:20.879
there's a bunch of different things that could bring you
231
00:11:20.960 --> 00:11:22.840
into a lower fight Coban, Yeah.
232
00:11:22.679 --> 00:11:25.039
And I think that trending data of like which direction
233
00:11:25.120 --> 00:11:25.840
are you trending?
234
00:11:26.240 --> 00:11:28.080
That can definitely be seen in the.
235
00:11:28.039 --> 00:11:32.279
Credit report data. Are your balances increasing, are they decreasing?
236
00:11:32.440 --> 00:11:35.679
Are the number of tradelines you have increasing decreasing? Is
237
00:11:35.679 --> 00:11:38.120
your mix improving of the types of tradelines? I think
238
00:11:38.159 --> 00:11:41.399
all of those things, not just like the snapshot of
239
00:11:41.440 --> 00:11:44.000
what your credit report looks like today, but what does