WEBVTT

1
00:00:00.225 --> 00:00:02.845
- We just think that human
connection is so integral

2
00:00:02.845 --> 00:00:05.125
that AI will mostly be supercharging

3
00:00:05.125 --> 00:00:06.445
them, not replacing them.

4
00:00:09.745 --> 00:00:12.225
- So Zillow's got a couple
of different AI tools.

5
00:00:12.225 --> 00:00:15.345
Can you walk us through how
the platform has changed

6
00:00:15.345 --> 00:00:17.145
and where it's going from here?

7
00:00:17.145 --> 00:00:18.885
- So Zillow's been, you know,

8
00:00:18.885 --> 00:00:20.915
an AI focused company for the beginning.

9
00:00:20.915 --> 00:00:23.025
Obviously many people
think of the estimate

10
00:00:23.025 --> 00:00:24.765
and also how we think
about search ranking.

11
00:00:26.005 --> 00:00:28.075
I think the current wave is not hype.

12
00:00:28.075 --> 00:00:29.605
It's definitely very real for us.

13
00:00:30.465 --> 00:00:32.925
The improvements we're making
are not only affecting,

14
00:00:32.925 --> 00:00:34.845
you know, buyers and renters as they go

15
00:00:34.845 --> 00:00:36.865
through the whole flow,
but also we're doing a lot

16
00:00:36.865 --> 00:00:40.225
behind the scenes with
agents and loan officers.

17
00:00:40.225 --> 00:00:43.395
And it's a mix of AI that's very visible.

18
00:00:43.395 --> 00:00:46.685
Like, hey, you could fly
through a floor plan as well as

19
00:00:46.685 --> 00:00:49.245
behind the scenes making the kind of tasks

20
00:00:49.245 --> 00:00:52.005
that you don't wanna be spending
time on a lot easier. And

21
00:00:52.005 --> 00:00:53.205
- What kind of tasks are we talking about?

22
00:00:53.205 --> 00:00:55.005
Like what things are being simplified?

23
00:00:55.005 --> 00:00:57.285
Is it like the search tool is getting more

24
00:00:57.285 --> 00:00:59.355
conversational, or what
does this look like?

25
00:00:59.355 --> 00:01:01.365
- Searching and finding a home is still

26
00:01:01.365 --> 00:01:02.545
incredibly complicated.

27
00:01:02.545 --> 00:01:04.345
You have to think about
all your preferences,

28
00:01:04.345 --> 00:01:05.875
all your requirements,
what you can afford,

29
00:01:05.875 --> 00:01:07.045
what neighborhoods are out there,

30
00:01:07.045 --> 00:01:09.505
places you haven't been,
all, all the above.

31
00:01:09.505 --> 00:01:11.085
And so one of the biggest things

32
00:01:11.085 --> 00:01:13.585
that we're investing in is
natural language search.

33
00:01:13.585 --> 00:01:15.245
Before generative ai, a lot

34
00:01:15.245 --> 00:01:17.045
of the techniques were old school

35
00:01:17.045 --> 00:01:18.505
natural language processing.

36
00:01:18.505 --> 00:01:19.845
And now with transformers

37
00:01:19.845 --> 00:01:22.905
and the like, we can make
that experience much smarter.

38
00:01:22.905 --> 00:01:24.825
So it's just more intuitive.

39
00:01:24.825 --> 00:01:28.785
And then on the agent
side, you know, agents

40
00:01:28.785 --> 00:01:31.885
really sell themselves by
that personal relationship

41
00:01:31.885 --> 00:01:35.105
and that guidance that
everyone needs to move

42
00:01:35.105 --> 00:01:38.805
and yet have to spend a ton
of time basically coordinating

43
00:01:38.805 --> 00:01:41.445
and nurturing people who
aren't ready to move, as well

44
00:01:41.445 --> 00:01:43.105
as in paperwork mode.

45
00:01:43.105 --> 00:01:46.025
You know, if you've moved all the title

46
00:01:46.025 --> 00:01:48.305
and escrow stuff, all
the transaction stuff.

47
00:01:48.305 --> 00:01:51.275
And so AI can help in both polls,

48
00:01:51.275 --> 00:01:53.445
helping people like get going, as well

49
00:01:53.445 --> 00:01:55.925
as helping the transaction
actually finish so

50
00:01:55.925 --> 00:01:59.155
that the agents can focus on
the, Hey, what should we tour?

51
00:01:59.155 --> 00:02:01.005
What should we offer that kind of human

52
00:02:01.005 --> 00:02:02.405
to human connection. So one thing

53
00:02:02.405 --> 00:02:04.125
- I think is interesting
here is that, you know,

54
00:02:04.125 --> 00:02:06.065
people probably don't think of real estate

55
00:02:06.065 --> 00:02:08.525
as a tech business, but I
think it's probably fair to say

56
00:02:08.525 --> 00:02:10.965
that Zillow is a tech company, right?

57
00:02:10.965 --> 00:02:13.265
Like it's an online platform.

58
00:02:13.265 --> 00:02:18.145
How much of the user experience
of going to zillow.com

59
00:02:18.145 --> 00:02:21.145
is interacting with AI now

60
00:02:21.145 --> 00:02:23.365
and where's that going
like five years from now?

61
00:02:23.365 --> 00:02:25.085
Do you think AI and you know, interacting

62
00:02:25.085 --> 00:02:26.205
with chatbots is gonna be a much

63
00:02:26.205 --> 00:02:27.385
bigger part of the platform?

64
00:02:27.385 --> 00:02:30.645
Or is it something that creeps
in more slowly depending on

65
00:02:30.645 --> 00:02:32.045
how people react to it?

66
00:02:32.045 --> 00:02:33.775
- I think we're at the
kind of like teenager

67
00:02:33.775 --> 00:02:35.545
phase of this wave.

68
00:02:35.545 --> 00:02:39.025
And you know, anytime
there's a new AI technology,

69
00:02:39.025 --> 00:02:41.525
it usually starts with
what we've witnessed

70
00:02:41.525 --> 00:02:44.005
with generative AI of
everything's gonna change.

71
00:02:44.005 --> 00:02:47.385
Like hold onto your hats,
tomorrow's gonna be different.

72
00:02:47.385 --> 00:02:50.325
And then you get to phase two,
which is like, hold on, all

73
00:02:50.325 --> 00:02:52.905
of our ideas aren't really coming to pass

74
00:02:52.905 --> 00:02:53.945
as fast as we thought.

75
00:02:53.945 --> 00:02:55.485
Or maybe in the waves that we thought,

76
00:02:55.485 --> 00:02:58.585
and we gotta figure out what
the long-term path here is.

77
00:02:58.585 --> 00:03:00.725
And so I, my take is, I think we're kind

78
00:03:00.725 --> 00:03:03.285
of at the Ulta Vista
level of generative ai

79
00:03:03.285 --> 00:03:06.145
where there's something there,
there's definitely value,

80
00:03:06.145 --> 00:03:08.605
but we haven't figured out the right

81
00:03:08.605 --> 00:03:09.925
UX model for this long term.

82
00:03:11.065 --> 00:03:12.485
We are are trying to figure out

83
00:03:12.485 --> 00:03:16.225
how much a chat experience can
help guide people as well as,

84
00:03:16.225 --> 00:03:19.685
you know, the more traditional
property search information

85
00:03:19.685 --> 00:03:22.345
retrieval story, we need both.

86
00:03:22.345 --> 00:03:24.425
And it's sometimes
complicated to figure out

87
00:03:24.425 --> 00:03:25.845
how they work in parallel.

88
00:03:25.845 --> 00:03:28.345
And so that's something
we're looking at a lot.

89
00:03:28.345 --> 00:03:31.845
And then also we think there's
huge amounts of opportunity

90
00:03:31.845 --> 00:03:33.005
behind the scenes with agents

91
00:03:33.005 --> 00:03:34.915
and loan officers in their workflow.

92
00:03:34.915 --> 00:03:37.705
They're already spending a
lot of time clicking buttons,

93
00:03:37.705 --> 00:03:39.225
you know, doing these tasks.

94
00:03:39.225 --> 00:03:41.645
And so we have, we're more bullish on

95
00:03:41.645 --> 00:03:44.205
how generative AI can
adopt that kind of UX

96
00:03:45.075 --> 00:03:48.025
than necessarily a a pure chat experience.

97
00:03:48.025 --> 00:03:50.565
- So this is, you know, a much
broader question, you know,

98
00:03:50.565 --> 00:03:51.565
really about the state of the world,

99
00:03:51.565 --> 00:03:55.005
but if we, you know, compare
Zillow to how real estate used

100
00:03:55.005 --> 00:03:56.915
to work, I think a lot of the work

101
00:03:56.915 --> 00:03:58.405
that Zillow does is something

102
00:03:58.405 --> 00:04:00.185
that a real estate agent would do.

103
00:04:00.185 --> 00:04:02.525
But if, you know,

104
00:04:02.525 --> 00:04:06.215
18 months from now OpenAI says
we've solved the AI problem,

105
00:04:06.215 --> 00:04:07.605
we've got, you know,

106
00:04:07.605 --> 00:04:09.165
a general intelligence available can

107
00:04:09.165 --> 00:04:10.865
do everything a person can do.

108
00:04:10.865 --> 00:04:12.645
Do real estate agents go away then?

109
00:04:12.645 --> 00:04:15.305
Like, would that be a, a good thing?

110
00:04:15.305 --> 00:04:16.365
Is that even something

111
00:04:16.365 --> 00:04:18.885
that comes into the consideration
when you're thinking

112
00:04:18.885 --> 00:04:20.345
about the future of the product?

113
00:04:20.345 --> 00:04:21.925
- My belief is that real estate,

114
00:04:21.925 --> 00:04:25.185
like human real estate agents
will always be at the center

115
00:04:25.185 --> 00:04:27.485
of a, of a buying transaction.

116
00:04:27.485 --> 00:04:29.685
I think the incentives are
very aligned with this one

117
00:04:29.685 --> 00:04:31.205
because, you know,

118
00:04:31.205 --> 00:04:34.125
what slows these human
agents down is a lot

119
00:04:34.125 --> 00:04:37.025
of these more basic tasks
that we're talking about.

120
00:04:37.025 --> 00:04:39.545
And their goal is to rocket forward

121
00:04:39.545 --> 00:04:42.195
how many successful
transactions they get a year.

122
00:04:42.195 --> 00:04:44.245
It's almost like the
parallel of the Microsoft

123
00:04:44.245 --> 00:04:45.595
and Google co-pilots.

124
00:04:45.595 --> 00:04:49.875
They want AI basically unblocking
them to spend more time

125
00:04:49.875 --> 00:04:51.345
with their customers

126
00:04:51.345 --> 00:04:54.585
and then all the customer
research, all the actual buyers

127
00:04:54.585 --> 00:04:57.655
we see time and time again
after they successfully moved.

128
00:04:57.655 --> 00:04:59.725
Their biggest, most
important element was having

129
00:04:59.725 --> 00:05:01.945
that great agent guiding them.

130
00:05:01.945 --> 00:05:02.685
And these, these

131
00:05:02.685 --> 00:05:04.115
conversations are happening
at the dinner table.

132
00:05:04.115 --> 00:05:06.105
They're happening after the soccer game.

133
00:05:06.105 --> 00:05:09.045
We just think that human
connection is so integral

134
00:05:09.045 --> 00:05:11.285
that AI will mostly be supercharging

135
00:05:11.285 --> 00:05:12.785
them, not replacing them.

136
00:05:12.785 --> 00:05:14.635
For most people, 90% of buyers,

137
00:05:14.635 --> 00:05:16.865
this is the biggest financial transaction

138
00:05:16.865 --> 00:05:19.425
and half of them are doing
it for the first time.

139
00:05:19.425 --> 00:05:21.685
And so there's a lot of
uncertainty, there's a lot

140
00:05:21.685 --> 00:05:24.705
of questions, and having someone
a human be able to explain

141
00:05:24.705 --> 00:05:27.625
and guide you through is, is so important.

142
00:05:27.625 --> 00:05:30.045
- In general. Can you
tell us what types of AI

143
00:05:30.045 --> 00:05:31.645
or machine learning tools Zillow is using?

144
00:05:31.645 --> 00:05:33.665
Like, is it mostly language models

145
00:05:33.665 --> 00:05:35.795
or, I imagine there's a
bunch of different things.

146
00:05:35.795 --> 00:05:37.805
- It's the whole gamut.

147
00:05:37.805 --> 00:05:40.825
There's everything from
deep learning neuro models

148
00:05:40.825 --> 00:05:44.305
to more traditional rank ranking
recommendation models to,

149
00:05:44.305 --> 00:05:45.855
of course generative ai,

150
00:05:45.855 --> 00:05:48.265
large language models to your point.

151
00:05:48.265 --> 00:05:51.545
And even some more basic ones
to like figure out scheduling

152
00:05:51.545 --> 00:05:54.265
and which tour you should
go to in which order

153
00:05:54.265 --> 00:05:57.225
to maximize your time
actually in the house.

154
00:05:57.225 --> 00:05:59.525
And so a lot of our, our job is trying

155
00:05:59.525 --> 00:06:01.505
to figure out which tool for the job

156
00:06:01.505 --> 00:06:04.205
and not just assume
that that one model out

157
00:06:04.205 --> 00:06:06.105
of the gate is the right answer.

158
00:06:06.105 --> 00:06:10.275
- We hear so much about, you
know, this apocalyptic stuff

159
00:06:10.275 --> 00:06:12.625
that once you know AI gets
good enough, it's gonna,

160
00:06:12.625 --> 00:06:17.625
you know, once replace the need
for human labor in general.

161
00:06:17.675 --> 00:06:19.085
Does that keep you up at night?

162
00:06:19.085 --> 00:06:22.065
Like, I'm not really convinced
that, you know, we've gotten

163
00:06:22.065 --> 00:06:26.115
to the place where that technical
problem is gonna be solved

164
00:06:26.115 --> 00:06:28.115
even in the medium term necessarily.

165
00:06:28.115 --> 00:06:29.005
What do you think? Is that

166
00:06:29.005 --> 00:06:30.565
something we should be thinking about?

167
00:06:30.565 --> 00:06:34.965
- I, I am like you, I'm
not losing sleep over this.

168
00:06:34.965 --> 00:06:37.245
I mean, you remember the previous AI waves

169
00:06:37.245 --> 00:06:40.025
with deep learning and
convolution neural nets.

170
00:06:40.025 --> 00:06:42.105
We were like, oh goodness,
you know, here we go.

171
00:06:42.105 --> 00:06:45.115
No one will have a job. We all still do

172
00:06:45.115 --> 00:06:47.905
when you're in it day to
day, to your point, you see

173
00:06:47.905 --> 00:06:51.145
how brittle it can be, how
much it needs to improve

174
00:06:51.145 --> 00:06:54.225
to even be good at, at what
we'd expect it to do today.

175
00:06:54.225 --> 00:06:56.045
And so, no, I, I don't,

176
00:06:56.045 --> 00:06:59.125
I don't see us all losing
jobs or beginning replaced.

177
00:06:59.125 --> 00:07:02.065
I think it will more be helping
us on the side guidance,

178
00:07:02.065 --> 00:07:03.725
you know, co-pilot type stuff.