1
00:00:02,720 --> 00:00:14,040
Speaker 1: Bloomberg Audio Studios, Podcasts, Radio News.

2
00:00:18,560 --> 00:00:22,360
Speaker 2: Hello and welcome to another episode of the Odd Locks podcast.

3
00:00:22,400 --> 00:00:24,759
Speaker 3: I'm Jill Wisenthalm and I'm Tracy Alloway.

4
00:00:25,400 --> 00:00:28,640
Speaker 2: Tracy recording this January eighth. We ran a good piece

5
00:00:28,720 --> 00:00:31,639
Speaker 2: in the newsletter this week from Sconda about AI spend

6
00:00:31,920 --> 00:00:35,400
Speaker 2: is like a sort of like meaningful macro driver or

7
00:00:35,479 --> 00:00:38,040
Speaker 2: getting close to where it starts to move the dial.

8
00:00:38,479 --> 00:00:41,239
Speaker 3: Yeah, that was a really good piece, and I have

9
00:00:41,320 --> 00:00:45,920
Speaker 3: to say some of it is slightly worrying. But I

10
00:00:45,960 --> 00:00:48,199
Speaker 3: think one of the big things that's happening now is

11
00:00:48,240 --> 00:00:52,280
Speaker 3: okay ai has become such a big pillar of the market.

12
00:00:52,400 --> 00:00:54,760
Speaker 3: Y right, like the entire S and P five hundred basics.

13
00:00:54,840 --> 00:00:55,640
Speaker 2: It's like an AI play.

14
00:00:55,960 --> 00:00:58,240
Speaker 3: Yeah, it is an AI play. And so at some

15
00:00:58,480 --> 00:01:01,960
Speaker 3: point the hype passed to be match by reality, i e.

16
00:01:02,120 --> 00:01:05,640
Speaker 3: All that investment has to be matched by some sort

17
00:01:05,680 --> 00:01:08,120
Speaker 3: of revenue, right, you have to get money out of

18
00:01:08,160 --> 00:01:09,200
Speaker 3: making this investment.

19
00:01:09,440 --> 00:01:11,480
Speaker 2: I think there's three ways things could go. I've been

20
00:01:11,520 --> 00:01:14,120
Speaker 2: thinking about this. There's three ways this could play out.

21
00:01:14,440 --> 00:01:18,480
Speaker 2: One is companies don't get a lot of productivity gains

22
00:01:18,480 --> 00:01:21,480
Speaker 2: from these tools, they cut back spending. There's a bunch

23
00:01:21,560 --> 00:01:26,160
Speaker 2: of other projects get delayed, market goes down. Another possibility

24
00:01:26,360 --> 00:01:30,600
Speaker 2: is there's this great productivity breakthrough, companies are more efficient

25
00:01:30,640 --> 00:01:33,679
Speaker 2: than ever, incredible boom. That's great. And then the other

26
00:01:33,720 --> 00:01:35,840
Speaker 2: possibility is that none of this matter is and they

27
00:01:35,880 --> 00:01:38,680
Speaker 2: build God at one of these labs, and then everything

28
00:01:38,720 --> 00:01:41,360
Speaker 2: we know about economics doesn't even make any sense. So

29
00:01:41,440 --> 00:01:43,440
Speaker 2: to even talk about productivity or the S and P

30
00:01:43,520 --> 00:01:47,320
Speaker 2: five hundred or earnings in that regime is just like

31
00:01:47,440 --> 00:01:49,840
Speaker 2: it's like a secondary concern to like the way the

32
00:01:49,840 --> 00:01:51,920
Speaker 2: world has changed when they achieve AGI.

33
00:01:52,280 --> 00:01:53,280
Speaker 3: Joe, are you okay?

34
00:01:53,480 --> 00:01:53,560
Speaker 1: No?

35
00:01:53,680 --> 00:01:56,080
Speaker 2: I think those those are like the three Yeah.

36
00:01:55,920 --> 00:01:59,080
Speaker 3: You might as well go big with your scenario analysis. Yeah,

37
00:01:59,080 --> 00:02:01,400
Speaker 3: but you're right. It could go either way. And obviously

38
00:02:01,440 --> 00:02:04,240
Speaker 3: there's a lot of talk and nervousness about a bubble

39
00:02:04,400 --> 00:02:07,000
Speaker 3: in AI at the moment, So I think it would

40
00:02:07,040 --> 00:02:09,120
Speaker 3: be a good idea to maybe try to get an

41
00:02:09,160 --> 00:02:14,040
Speaker 3: understanding of how much companies are actually spending and benefiting

42
00:02:14,240 --> 00:02:14,840
Speaker 3: from AI.

43
00:02:15,200 --> 00:02:18,040
Speaker 2: That's right, and we have the perfect guest because he

44
00:02:18,080 --> 00:02:21,520
Speaker 2: has a great view into what companies are spending money on,

45
00:02:21,720 --> 00:02:25,280
Speaker 2: including AI, and also as the CEO of a company

46
00:02:25,360 --> 00:02:28,400
Speaker 2: himself in the tech space, but not directly in the

47
00:02:28,440 --> 00:02:31,280
Speaker 2: AI space, the sort of user of AI tools, and

48
00:02:31,320 --> 00:02:34,160
Speaker 2: maybe could talk about what's being used, what's not being used,

49
00:02:34,200 --> 00:02:37,480
Speaker 2: where productivity gains are being had. We're going to talk

50
00:02:37,480 --> 00:02:39,160
Speaker 2: about all of this stuff. We're gonna be speaking with

51
00:02:39,320 --> 00:02:42,560
Speaker 2: Eric Glyman. He is the founder and CEO of RAMP,

52
00:02:42,639 --> 00:02:45,040
Speaker 2: which is a New York City based company. It's a

53
00:02:45,040 --> 00:02:48,480
Speaker 2: spending management platform for companies help them deal with expenses.

54
00:02:48,600 --> 00:02:51,119
Speaker 2: We might also talk a little bit about expense management

55
00:02:51,160 --> 00:02:55,880
Speaker 2: platforms because I have I have headaches about dealing with expenses.

56
00:02:55,919 --> 00:02:56,880
Speaker 4: I think a lot of people do.

57
00:02:57,080 --> 00:02:59,359
Speaker 2: Yeah, I guess that's why there's a business, a new

58
00:02:59,360 --> 00:03:01,440
Speaker 2: business to be made. Eric, thank you so much for

59
00:03:01,760 --> 00:03:02,640
Speaker 2: coming on Odd Laws.

60
00:03:02,960 --> 00:03:05,040
Speaker 4: Joe Tracy, thanks so much for having me. It's great

61
00:03:05,040 --> 00:03:05,440
Speaker 4: to be here.

62
00:03:05,919 --> 00:03:08,600
Speaker 2: Real quickly, what do you describe a little bit? What's Ramp?

63
00:03:08,639 --> 00:03:09,600
Speaker 2: What's your story here?

64
00:03:10,160 --> 00:03:13,200
Speaker 4: Well, you can think about RAMP as a financial operations platform.

65
00:03:13,240 --> 00:03:16,560
Speaker 4: It's a single place where companies can issue cards, make

66
00:03:16,639 --> 00:03:20,560
Speaker 4: payments of all kinds, and even automate both expense management,

67
00:03:20,560 --> 00:03:23,560
Speaker 4: which we'll get to, and accounting. But the ethos of

68
00:03:23,600 --> 00:03:25,919
Speaker 4: the company and why we exist is actually to help

69
00:03:26,000 --> 00:03:29,359
Speaker 4: companies save time and money. We're the only company in

70
00:03:29,400 --> 00:03:31,920
Speaker 4: our space that actually reports back to our customers how

71
00:03:31,960 --> 00:03:34,160
Speaker 4: much money and time did we save you Over the

72
00:03:34,160 --> 00:03:36,680
Speaker 4: past four years. We've saved our customers over two billion

73
00:03:36,680 --> 00:03:39,480
Speaker 4: dollars twenty million hours, and half of that has actually

74
00:03:39,480 --> 00:03:42,080
Speaker 4: come in the past year. And so we serve thirty

75
00:03:42,120 --> 00:03:45,360
Speaker 4: thousand plus companies from small, medium to publicly traded.

76
00:03:45,760 --> 00:03:46,120
Speaker 2: Got it.

77
00:03:46,360 --> 00:03:49,040
Speaker 3: What was the gap in the market that you saw,

78
00:03:49,240 --> 00:03:52,160
Speaker 3: Because on the one hand, as Joe just laid out,

79
00:03:52,400 --> 00:03:55,400
Speaker 3: a lot of people hate expenses and they're clunky and

80
00:03:55,560 --> 00:03:58,880
Speaker 3: very bureaucratic and they take forever to do. But on

81
00:03:58,920 --> 00:04:01,800
Speaker 3: the other hand, this is a space that is dominated

82
00:04:01,880 --> 00:04:05,960
Speaker 3: by some very powerful legacy players, right I'm thinking about

83
00:04:06,000 --> 00:04:09,800
Speaker 3: American Express, for instance, and you're basically going up against them.

84
00:04:10,640 --> 00:04:13,280
Speaker 4: These are companies that are great in their own right.

85
00:04:13,320 --> 00:04:14,720
Speaker 4: But I think we're built for a bit of a

86
00:04:14,720 --> 00:04:17,560
Speaker 4: different era. In the company you mentioned, the founders quite

87
00:04:17,600 --> 00:04:20,320
Speaker 4: literally wore top hats, you know, and are you know,

88
00:04:20,440 --> 00:04:23,159
Speaker 4: thinking not so much about I think really the needs

89
00:04:23,160 --> 00:04:25,760
Speaker 4: of the twenty twenties. Where I think luxury in the

90
00:04:25,760 --> 00:04:28,760
Speaker 4: twenty twenties is actually having an hour to yourself at

91
00:04:28,760 --> 00:04:31,200
Speaker 4: the end of a long workweek versus you have expense

92
00:04:31,240 --> 00:04:33,599
Speaker 4: reports to do at the end, and so we thought

93
00:04:33,600 --> 00:04:35,839
Speaker 4: the gap was a few folds. First, could you actually

94
00:04:36,200 --> 00:04:39,000
Speaker 4: infuse technology not to make an easier to use expense

95
00:04:39,000 --> 00:04:41,880
Speaker 4: report that only took an hour instead of two hours,

96
00:04:42,080 --> 00:04:44,479
Speaker 4: but actually an expense report that does itself, books that

97
00:04:44,560 --> 00:04:47,359
Speaker 4: keep themselves, and so the difference was we saw an

98
00:04:47,360 --> 00:04:49,479
Speaker 4: opportunity to create a card where you can tap it

99
00:04:49,800 --> 00:04:52,680
Speaker 4: make a purchase. We pulled a receipt from the merchant

100
00:04:52,760 --> 00:04:55,360
Speaker 4: or your email automatically, so your expense report is done

101
00:04:55,400 --> 00:04:57,520
Speaker 4: for you, your books and records are done for you,

102
00:04:57,640 --> 00:05:01,159
Speaker 4: and more. For business owners, we found the strange distortion

103
00:05:01,400 --> 00:05:04,839
Speaker 4: where they were trying to market products like spend more money,

104
00:05:04,880 --> 00:05:07,440
Speaker 4: earn more points. But every business owner I ever met

105
00:05:07,480 --> 00:05:10,040
Speaker 4: actually wanted to spend less and be more profit. So

106
00:05:10,080 --> 00:05:12,039
Speaker 4: we just try to keep it simple and build a

107
00:05:12,080 --> 00:05:14,239
Speaker 4: company just on those principles.

108
00:05:14,560 --> 00:05:17,680
Speaker 2: I mentioned in the intro. Because you have this expense

109
00:05:17,760 --> 00:05:21,120
Speaker 2: management platform, you have some insight into what companies are

110
00:05:21,160 --> 00:05:24,479
Speaker 2: spending money on these days on AI, Like what can

111
00:05:24,520 --> 00:05:27,320
Speaker 2: you see you know, what are you able to see

112
00:05:27,360 --> 00:05:30,200
Speaker 2: about AI spend within large corporations?

113
00:05:30,240 --> 00:05:32,920
Speaker 4: I mean it's real It is dramatically increasing, but in

114
00:05:33,520 --> 00:05:35,440
Speaker 4: actually interesting ways. And to give you a sense of

115
00:05:35,440 --> 00:05:37,960
Speaker 4: the panel data that we're looking at to get these insights,

116
00:05:38,000 --> 00:05:41,560
Speaker 4: we see over fifty billion a year in spend buy companies.

117
00:05:41,800 --> 00:05:44,640
Speaker 4: Some of these are publicly traded, most of these are private,

118
00:05:44,920 --> 00:05:47,400
Speaker 4: and often these tend to be on the bleeding edge.

119
00:05:47,400 --> 00:05:50,960
Speaker 4: So these can be from AI research labs themselves to

120
00:05:51,279 --> 00:05:55,160
Speaker 4: farms to nonprofits, mom and pop shops. And this is

121
00:05:55,200 --> 00:05:58,599
Speaker 4: across both credit card data as well as bill payment data,

122
00:05:58,680 --> 00:06:01,240
Speaker 4: so it's a pretty good subset of it. And what

123
00:06:01,279 --> 00:06:04,000
Speaker 4: we've seen is maybe twofold. First, just in terms of

124
00:06:04,080 --> 00:06:07,480
Speaker 4: raw and aggregate numbers, an average customer on Ramp from

125
00:06:07,760 --> 00:06:10,360
Speaker 4: the start of twenty twenty three to the end was

126
00:06:10,400 --> 00:06:13,320
Speaker 4: spending about four times the number of raw dollars on

127
00:06:13,440 --> 00:06:16,960
Speaker 4: AI based products, and so there's real budget that's starting

128
00:06:16,960 --> 00:06:20,320
Speaker 4: to go to this increased ways. And next the products

129
00:06:20,360 --> 00:06:25,640
Speaker 4: themselves are starting to actually go from experimental to operations.

130
00:06:25,640 --> 00:06:27,200
Speaker 2: How can you see that? How do you what do

131
00:06:27,240 --> 00:06:29,480
Speaker 2: you see? What do you see? Your data that bags

132
00:06:29,480 --> 00:06:29,880
Speaker 2: that up.

133
00:06:29,920 --> 00:06:32,000
Speaker 4: So the best way we think to know this is

134
00:06:32,000 --> 00:06:35,240
Speaker 4: that if you looked at an average AI purchase, maybe

135
00:06:35,279 --> 00:06:38,240
Speaker 4: you purchased some software seat In twenty twenty two, there

136
00:06:38,279 --> 00:06:40,679
Speaker 4: was a fifty percent chance that within the next month

137
00:06:41,040 --> 00:06:42,919
Speaker 4: a customer that bought it would no longer be a

138
00:06:42,920 --> 00:06:46,280
Speaker 4: customer they were experimenting with it. In twenty twenty three

139
00:06:46,480 --> 00:06:50,039
Speaker 4: that had jumped to a seventy percent chance. In twenty

140
00:06:50,080 --> 00:06:53,560
Speaker 4: twenty four, it's continuing to go higher, and we'll release data.

141
00:06:53,680 --> 00:06:56,280
Speaker 2: Wait, sorry, fifty percent change, You continued at the next

142
00:06:56,360 --> 00:06:57,760
Speaker 2: that's okay, got it, got it.

143
00:06:57,800 --> 00:07:00,200
Speaker 4: And so there was a radically higher chance that you

144
00:07:00,240 --> 00:07:03,280
Speaker 4: were keeping this around. And so it went from tinkering

145
00:07:03,400 --> 00:07:06,480
Speaker 4: to this is starting to become a real part of

146
00:07:06,520 --> 00:07:10,320
Speaker 4: engineering processes, sales tools that teams were using to be

147
00:07:10,400 --> 00:07:14,360
Speaker 4: more productive, to even back office tools to manage accounting,

148
00:07:14,400 --> 00:07:16,880
Speaker 4: managed expenses. And so I think we're still on the

149
00:07:16,880 --> 00:07:19,200
Speaker 4: trajectory best in class of products are going to be

150
00:07:19,240 --> 00:07:22,360
Speaker 4: in the nineties of percent, but the jump was dramatic

151
00:07:22,560 --> 00:07:23,680
Speaker 4: in twenty twenty three and four.

152
00:07:24,160 --> 00:07:27,200
Speaker 3: How granular does your data go? Like can you see

153
00:07:27,320 --> 00:07:32,080
Speaker 3: people spending on I don't know, a basic LM subscription

154
00:07:32,320 --> 00:07:33,800
Speaker 3: versus something else?

155
00:07:34,440 --> 00:07:37,440
Speaker 4: Very much so. So the interesting part about what we

156
00:07:37,480 --> 00:07:40,680
Speaker 4: do is because we automate the expense report process, we

157
00:07:40,720 --> 00:07:44,200
Speaker 4: can see not just that a company spent on open ai,

158
00:07:44,360 --> 00:07:47,280
Speaker 4: but specifically was it an API call, was it a

159
00:07:47,440 --> 00:07:51,800
Speaker 4: chat GBT license? And so even among products, you're seeing

160
00:07:51,840 --> 00:07:54,960
Speaker 4: itemized and skew level data, and so you can start

161
00:07:54,960 --> 00:07:58,000
Speaker 4: to get really interesting insights of even in terms of

162
00:07:58,080 --> 00:08:00,640
Speaker 4: sub markets. One of the emergent themes that people are

163
00:08:00,640 --> 00:08:03,520
Speaker 4: talking about now wasn't twenty twenty three, there was only

164
00:08:03,600 --> 00:08:05,760
Speaker 4: one name in AI that mattered, and that was open Ai.

165
00:08:06,280 --> 00:08:10,480
Speaker 4: In twenty twenty four, suddenly twenty percent of developer market

166
00:08:10,480 --> 00:08:13,280
Speaker 4: share was going to Anthropic, which was I think at

167
00:08:13,320 --> 00:08:15,520
Speaker 4: three percent in data in twenty twenty three, And so

168
00:08:15,560 --> 00:08:17,480
Speaker 4: you can start to get very granular of how is

169
00:08:17,520 --> 00:08:20,679
Speaker 4: this even being used across which models are being called,

170
00:08:20,680 --> 00:08:23,600
Speaker 4: and so it's actually this interesting level of insight that

171
00:08:23,680 --> 00:08:26,120
Speaker 4: hasn't quite been seen in these markets.

172
00:08:26,440 --> 00:08:28,960
Speaker 3: And then, sorry to focus so much on the data,

173
00:08:29,000 --> 00:08:32,960
Speaker 3: but how do you actually classify an AI use versus

174
00:08:33,000 --> 00:08:36,160
Speaker 3: something else? Because I imagine there's a lot of software, for instance,

175
00:08:36,200 --> 00:08:40,600
Speaker 3: out there now that incorporates some sort of AI component. Right,

176
00:08:40,640 --> 00:08:43,520
Speaker 3: it feels like the ven diagram of AI and basic

177
00:08:43,559 --> 00:08:46,600
Speaker 3: tech spending is kind of starting to come together.

178
00:08:47,520 --> 00:08:50,000
Speaker 4: I think you're totally right, and there's a variety of

179
00:08:50,559 --> 00:08:52,800
Speaker 4: I would say it was an easier question in twenty

180
00:08:52,880 --> 00:08:55,640
Speaker 4: twenty three there was only a few strange companies calling

181
00:08:55,679 --> 00:08:58,560
Speaker 4: themselves AI. Now you see kind of AI washing of

182
00:08:58,600 --> 00:09:01,320
Speaker 4: companies that are not your love stock to pop. But

183
00:09:01,480 --> 00:09:04,000
Speaker 4: I would say that we tend to classify these based

184
00:09:04,040 --> 00:09:06,240
Speaker 4: off of kind of self identification of the companies. These

185
00:09:06,240 --> 00:09:09,079
Speaker 4: tend to be large language models labs. These tend to

186
00:09:09,120 --> 00:09:12,199
Speaker 4: be companies that are pure play AI products, maybe in

187
00:09:12,240 --> 00:09:15,199
Speaker 4: eleven labs. If you want to generate an AI digital voice,

188
00:09:15,360 --> 00:09:18,439
Speaker 4: a cognition or Devin, you can hire an AI developer

189
00:09:18,760 --> 00:09:20,839
Speaker 4: and these type of tools. I think you're right though

190
00:09:21,200 --> 00:09:23,200
Speaker 4: my sense, And if you talk to too many people

191
00:09:23,200 --> 00:09:24,560
Speaker 4: in the valley, they'll tell you there will be no

192
00:09:24,640 --> 00:09:26,720
Speaker 4: company that sells technology in five years that is in

193
00:09:26,760 --> 00:09:29,280
Speaker 4: an AI company. And we'll see how the jury goes there.

194
00:09:29,280 --> 00:09:31,599
Speaker 4: But it's this basis two things.

195
00:09:31,720 --> 00:09:34,040
Speaker 2: First, as a statement, I've never coded in my life,

196
00:09:34,080 --> 00:09:36,280
Speaker 2: So I've made a goal for twenty twenty five to

197
00:09:36,400 --> 00:09:39,200
Speaker 2: like use AI to like build an app, And I

198
00:09:39,280 --> 00:09:42,480
Speaker 2: actually built a really rudimentary app, but it wasn't really

199
00:09:42,559 --> 00:09:44,000
Speaker 2: doing what I wanted it to do. I'm not going

200
00:09:44,040 --> 00:09:45,520
Speaker 2: to talk about what it is. It wasn't really doing

201
00:09:45,520 --> 00:09:47,560
Speaker 2: what I want to do. And then I like fixed

202
00:09:47,720 --> 00:09:49,440
Speaker 2: the code and I tried to re upload it and

203
00:09:49,440 --> 00:09:51,160
Speaker 2: I broke it. So I had an app for about

204
00:09:51,160 --> 00:09:53,760
Speaker 2: five minutes and then it died. But I'm gonna this

205
00:09:53,800 --> 00:09:55,319
Speaker 2: is like my like go, I really want to like

206
00:09:55,400 --> 00:09:57,800
Speaker 2: learn to use technology. I want to talk more though

207
00:09:57,880 --> 00:10:01,520
Speaker 2: about you know me, when I use quote use AI,

208
00:10:01,960 --> 00:10:05,360
Speaker 2: it's just meaning like going to chet OpenAI dot com,

209
00:10:05,400 --> 00:10:08,760
Speaker 2: just like the most rudimentary user interface. Talk to us

210
00:10:08,800 --> 00:10:11,720
Speaker 2: more about what you can see on the gap between

211
00:10:12,120 --> 00:10:15,520
Speaker 2: just someone subscribing to the website versus someone paying for

212
00:10:15,600 --> 00:10:18,960
Speaker 2: API calls, which I imagine is sort of a deeper

213
00:10:19,040 --> 00:10:22,760
Speaker 2: level of sophistication. Building these models into a workflow in somewhere.

214
00:10:22,520 --> 00:10:25,839
Speaker 4: Oh for sure. And I actually think this is the

215
00:10:25,880 --> 00:10:31,120
Speaker 4: most interesting development that if this happened successfully, it unlocks

216
00:10:31,160 --> 00:10:33,880
Speaker 4: what many of the value are talking about of historically

217
00:10:33,920 --> 00:10:36,760
Speaker 4: technology with software as a service. Yea, you know, you

218
00:10:36,800 --> 00:10:38,760
Speaker 4: could sell seats to people and would do it. And

219
00:10:38,800 --> 00:10:42,560
Speaker 4: there's this growing idea of service as software where suddenly

220
00:10:42,600 --> 00:10:46,040
Speaker 4: there are workflows where if AI is not just a

221
00:10:46,080 --> 00:10:48,800
Speaker 4: window you chat into, you get a response. But actually

222
00:10:48,880 --> 00:10:52,240
Speaker 4: at every step of it becomes very very interesting where

223
00:10:52,240 --> 00:10:55,000
Speaker 4: you have kind of end to end products, videos being created,

224
00:10:55,240 --> 00:10:59,400
Speaker 4: books being done automatically for finance teams, even art kind

225
00:10:59,400 --> 00:11:01,400
Speaker 4: of getting created in And so what I would say

226
00:11:01,480 --> 00:11:04,200
Speaker 4: is the way you can see it mechanically is often

227
00:11:04,240 --> 00:11:07,120
Speaker 4: the type of license. Usually these are consumer licenses if

228
00:11:07,120 --> 00:11:10,880
Speaker 4: you're buying like a chat GPT pro subscription versus there

229
00:11:10,920 --> 00:11:13,760
Speaker 4: are let's say enterprise or even kind of developer plans

230
00:11:13,960 --> 00:11:15,200
Speaker 4: where at the end of a month you get an

231
00:11:15,240 --> 00:11:18,000
Speaker 4: invoice from one of these vendors and you see, okay,

232
00:11:18,040 --> 00:11:20,760
Speaker 4: there were this many calls to this endpoint, these many

233
00:11:20,800 --> 00:11:24,240
Speaker 4: tokens were ultimately used, and will the specific use we

234
00:11:24,320 --> 00:11:27,520
Speaker 4: aggregate anonymous and don't report down to that level. You

235
00:11:27,559 --> 00:11:30,360
Speaker 4: can start to see suddenly this is much more similar

236
00:11:30,400 --> 00:11:34,120
Speaker 4: to how maybe a company would call Amazon Web Services

237
00:11:34,280 --> 00:11:37,520
Speaker 4: or Microsoft Azure, and this is core compute for some

238
00:11:37,840 --> 00:11:40,679
Speaker 4: service that is reliant and the different things about these

239
00:11:40,720 --> 00:11:43,560
Speaker 4: graphs is done right, and what all of these companies

240
00:11:43,559 --> 00:11:45,720
Speaker 4: are betting on is that they're going to grow exponentially

241
00:11:46,080 --> 00:11:48,640
Speaker 4: and that it's going to be deeply embedded. There are

242
00:11:48,640 --> 00:11:51,840
Speaker 4: some distinctions. One thing that I think makes this market

243
00:11:52,000 --> 00:11:54,760
Speaker 4: very different and in some sense more vicious than anything

244
00:11:54,840 --> 00:11:58,440
Speaker 4: I've ever seen is usually there's this idea of lock in.

245
00:11:58,600 --> 00:12:02,160
Speaker 4: You coast all your clouds services on one provider and

246
00:12:02,480 --> 00:12:05,200
Speaker 4: you can't change from one cloud to another, only for

247
00:12:05,240 --> 00:12:07,719
Speaker 4: small use cases. But in AI there's this practice of

248
00:12:07,800 --> 00:12:11,880
Speaker 4: kind of multiplexing, and so what developers are often doing

249
00:12:11,920 --> 00:12:15,400
Speaker 4: and sort of why Anthropic came out of nowhere seemingly

250
00:12:15,520 --> 00:12:17,880
Speaker 4: in a way that wouldn't be possible, was people would

251
00:12:18,240 --> 00:12:21,960
Speaker 4: try some knowledge work or some response on multiple libraries

252
00:12:22,000 --> 00:12:24,240
Speaker 4: open source opening, I see which one's the best, and

253
00:12:24,520 --> 00:12:26,960
Speaker 4: the winning call starts getting more calls, and so things

254
00:12:26,960 --> 00:12:28,760
Speaker 4: are just the markets are moving faster.

255
00:12:43,559 --> 00:12:46,640
Speaker 3: So the specific companies are anonymized, But can you see

256
00:12:46,679 --> 00:12:48,520
Speaker 3: stuff on a sectoral level.

257
00:12:49,000 --> 00:12:50,400
Speaker 2: We both went to the same but that's what I

258
00:12:50,440 --> 00:12:51,120
Speaker 2: was thinking about.

259
00:12:51,000 --> 00:12:53,560
Speaker 3: Like, for instance, can you see which industries seem to

260
00:12:53,600 --> 00:12:57,160
Speaker 3: be ahead in AI and which ones are perhaps lacking behind?

261
00:12:57,880 --> 00:13:01,560
Speaker 4: We can, And I think that there's a few things

262
00:13:01,600 --> 00:13:04,959
Speaker 4: that were obvious and others that have started to jump

263
00:13:05,000 --> 00:13:09,360
Speaker 4: out in interesting and unexpected ways. The obvious ones of

264
00:13:09,440 --> 00:13:13,800
Speaker 4: certainly the earliest adopters are technologists like technology. These are

265
00:13:13,800 --> 00:13:17,200
Speaker 4: engineering offices, startups. A lot of folks doing training are

266
00:13:17,720 --> 00:13:20,400
Speaker 4: trying and adopting very very quickly in our nimble to

267
00:13:20,520 --> 00:13:23,200
Speaker 4: use this. The interesting thing there is that have actually

268
00:13:23,240 --> 00:13:26,840
Speaker 4: most surprised us maybe correlate to what you might see

269
00:13:26,880 --> 00:13:29,679
Speaker 4: in experience. A lot of newsrooms are actually using kind

270
00:13:29,720 --> 00:13:31,920
Speaker 4: of recording tools, so if you're having a call, notes

271
00:13:31,960 --> 00:13:34,160
Speaker 4: are taken for you. A lot of sales teams are

272
00:13:34,240 --> 00:13:37,480
Speaker 4: using this tools that listen to sales calls, take down

273
00:13:37,520 --> 00:13:40,400
Speaker 4: the notes, suggest the next steps, and start to go

274
00:13:40,520 --> 00:13:44,160
Speaker 4: into this. But you also see I would say historically,

275
00:13:44,480 --> 00:13:47,120
Speaker 4: if companies wanted to cut costs and sort of focus

276
00:13:47,120 --> 00:13:50,360
Speaker 4: on efficiency, their choice was higher lower cost labor. Often

277
00:13:50,400 --> 00:13:53,200
Speaker 4: in spaces where some of this can be if you

278
00:13:53,240 --> 00:13:56,400
Speaker 4: can get the rules right, things like healthcare, which tends

279
00:13:56,440 --> 00:13:59,480
Speaker 4: to be late adopters. The rates of use of increase

280
00:13:59,559 --> 00:14:02,480
Speaker 4: is actually starting to be much faster because when costs

281
00:14:02,520 --> 00:14:05,080
Speaker 4: and margins are very low, if you can start doing

282
00:14:05,120 --> 00:14:08,040
Speaker 4: network calls to do more work when there's few two people.

283
00:14:08,360 --> 00:14:11,760
Speaker 4: These are some interesting emerging industries, but it's still quite early.

284
00:14:12,080 --> 00:14:14,920
Speaker 4: I would say the big question that people have out

285
00:14:14,960 --> 00:14:17,600
Speaker 4: there about the valuations of these companies is still very present.

286
00:14:17,920 --> 00:14:21,520
Speaker 4: The relative increases dramatic four x year over year in

287
00:14:21,680 --> 00:14:24,800
Speaker 4: raw dollars. Spend per company is very very large. But

288
00:14:24,920 --> 00:14:27,600
Speaker 4: to start to pay back some of these large cathic cycles,

289
00:14:27,880 --> 00:14:29,640
Speaker 4: it's going to be early. And so the bet that

290
00:14:29,680 --> 00:14:32,440
Speaker 4: needs to happen from where we sit is AI can't

291
00:14:32,480 --> 00:14:35,160
Speaker 4: just be a product skew that people buy. But I

292
00:14:35,200 --> 00:14:37,120
Speaker 4: think to pay this back, you probably start to need

293
00:14:37,200 --> 00:14:38,280
Speaker 4: everyone to use it.

294
00:14:38,640 --> 00:14:41,400
Speaker 3: Actually, this was going to be my other question. So

295
00:14:41,440 --> 00:14:44,680
Speaker 3: we're talking about AI spend, can you see the other

296
00:14:44,920 --> 00:14:48,080
Speaker 3: side of it, in the form of I guess savings

297
00:14:48,240 --> 00:14:52,200
Speaker 3: either by being more efficient or perhaps cutting jobs.

298
00:14:53,160 --> 00:14:56,440
Speaker 4: I would say it's very interesting questions, and I would

299
00:14:56,520 --> 00:15:00,360
Speaker 4: say generally, as AI has taken place, unemployment has started

300
00:15:00,400 --> 00:15:02,920
Speaker 4: to come down, and so I think these are very

301
00:15:02,960 --> 00:15:05,760
Speaker 4: real questions for the long term. And in fact, I

302
00:15:05,840 --> 00:15:08,320
Speaker 4: actually think one of the biggest misses in twenty twenty

303
00:15:08,320 --> 00:15:11,280
Speaker 4: four for those promising AI was it was going to

304
00:15:11,280 --> 00:15:14,200
Speaker 4: be the year of the AI agent. Was whatever I pitched,

305
00:15:14,600 --> 00:15:18,360
Speaker 4: you'd have the AI CFO, the AI engineer, the full

306
00:15:18,440 --> 00:15:20,280
Speaker 4: kind of jobs. And I think that what you start

307
00:15:20,320 --> 00:15:24,280
Speaker 4: to see is you see slices of things happening. For example,

308
00:15:24,560 --> 00:15:26,840
Speaker 4: I hope it's not anyone's job in twenty twenty four

309
00:15:27,000 --> 00:15:30,520
Speaker 4: to do just expense reports. But actually an AI can

310
00:15:30,600 --> 00:15:32,600
Speaker 4: do your expense reports. It can kind of look at

311
00:15:32,640 --> 00:15:36,560
Speaker 4: your invoices, it can do kind of the lowest value tasks,

312
00:15:36,880 --> 00:15:40,320
Speaker 4: where that's starting to become a present thing in these tools.

313
00:15:40,600 --> 00:15:42,800
Speaker 4: And so I think in the short run, I actually find,

314
00:15:43,160 --> 00:15:46,080
Speaker 4: generally we hear from customers work is getting more interesting

315
00:15:46,360 --> 00:15:48,960
Speaker 4: in some sense when AI is being adopted. I think

316
00:15:49,000 --> 00:15:51,800
Speaker 4: long term there are real questions of might it actually

317
00:15:51,800 --> 00:15:53,760
Speaker 4: be able to take a workflow? And and I think

318
00:15:53,880 --> 00:15:58,280
Speaker 4: practically speaking, aif and has very limited context, it gets

319
00:15:58,320 --> 00:16:00,680
Speaker 4: a question, it can prompt out a re sponsor doesn't

320
00:16:00,680 --> 00:16:02,280
Speaker 4: see the rest of the knowledge work, but as it

321
00:16:02,320 --> 00:16:04,720
Speaker 4: starts being everywhere, it might be possible.

322
00:16:04,960 --> 00:16:08,560
Speaker 2: I need that and and agentic AI for this app

323
00:16:08,600 --> 00:16:11,800
Speaker 2: that I'm building, because like what's really annoying is it'll

324
00:16:11,800 --> 00:16:13,880
Speaker 2: be because like I'll like write some code and like

325
00:16:13,920 --> 00:16:16,240
Speaker 2: Google collab, and then it'll be like I'll try to

326
00:16:16,240 --> 00:16:18,000
Speaker 2: push it to get hub and then I have to

327
00:16:18,000 --> 00:16:20,000
Speaker 2: like go find my token and I don't want to

328
00:16:20,120 --> 00:16:21,320
Speaker 2: do it, and then I'm like wait, where do I

329
00:16:21,360 --> 00:16:23,320
Speaker 2: find the token in GitHub so that I could put

330
00:16:23,360 --> 00:16:25,200
Speaker 2: it in here? And that's the stuff that I don't

331
00:16:25,200 --> 00:16:27,080
Speaker 2: want to I mean, I guess I have to do it,

332
00:16:27,120 --> 00:16:30,120
Speaker 2: but I need the AI to just like go find.

333
00:16:30,480 --> 00:16:32,920
Speaker 3: Does that count as coding if the AI is doing

334
00:16:32,920 --> 00:16:36,120
Speaker 3: the ending, if you're literally just typing in design and

335
00:16:36,200 --> 00:16:37,480
Speaker 3: AC I have.

336
00:16:37,520 --> 00:16:39,440
Speaker 2: Like, you know, I have some ideas. I'm like trying,

337
00:16:39,560 --> 00:16:41,760
Speaker 2: but it's like it's still a little bit annoying, like

338
00:16:41,800 --> 00:16:44,840
Speaker 2: all these different windows and everything. Okay, but speaking of

339
00:16:44,880 --> 00:16:48,040
Speaker 2: all that, let's talk about your what you're seeing in

340
00:16:48,080 --> 00:16:51,760
Speaker 2: your company. I imagine every engineer that I talked to

341
00:16:51,960 --> 00:16:54,640
Speaker 2: is like, yes, in twenty twenty four or twenty twenty five,

342
00:16:55,080 --> 00:16:57,120
Speaker 2: they have a window open with their AI and they

343
00:16:57,120 --> 00:17:00,840
Speaker 2: have their coding window, and it's improved their productivity. So, like,

344
00:17:00,840 --> 00:17:03,400
Speaker 2: I assume that's happening in your company, that a lot

345
00:17:03,400 --> 00:17:06,399
Speaker 2: of code is being written either directly or indirectly or

346
00:17:06,400 --> 00:17:11,639
Speaker 2: with assistance from AI. Where else besides engineering, what actually

347
00:17:11,800 --> 00:17:14,879
Speaker 2: are you spending money on in terms of AI resources?

348
00:17:16,000 --> 00:17:19,080
Speaker 4: So the three big places, and you nailed it. Number

349
00:17:19,119 --> 00:17:21,480
Speaker 4: one is in engineering. It is one of the most

350
00:17:21,480 --> 00:17:24,080
Speaker 4: digital jobs. All code is digital and so in a

351
00:17:24,119 --> 00:17:27,400
Speaker 4: strange way, that is actually arguably maybe the first industry

352
00:17:27,520 --> 00:17:31,680
Speaker 4: that is closest to AI. Second is sales and Growth. RAMP,

353
00:17:31,760 --> 00:17:34,440
Speaker 4: beyond having a large amount of spend data, is one

354
00:17:34,480 --> 00:17:37,720
Speaker 4: of the fastest growing companies or startups in history. And

355
00:17:37,760 --> 00:17:39,480
Speaker 4: part of what's allowed us to do this is the

356
00:17:39,600 --> 00:17:42,720
Speaker 4: average salesperson at RAMP is about four times as productive

357
00:17:42,760 --> 00:17:44,960
Speaker 4: as an ex closest competitor. A lot of that is

358
00:17:45,000 --> 00:17:48,160
Speaker 4: a heavy use of AI to automate aspects and lower

359
00:17:48,160 --> 00:17:48,840
Speaker 4: specific job.

360
00:17:48,960 --> 00:17:52,160
Speaker 2: Let's spice so one of it makes a sales caller.

361
00:17:52,240 --> 00:17:53,679
Speaker 2: So what are they using AI for?

362
00:17:53,920 --> 00:17:56,399
Speaker 4: So one of the most important functions in the first

363
00:17:56,480 --> 00:17:58,800
Speaker 4: rule for anyone going into sales is this role called

364
00:17:58,800 --> 00:18:01,640
Speaker 4: a sales development rep. And the job of that person

365
00:18:01,840 --> 00:18:05,400
Speaker 4: is book meetings, find people out there who maybe are

366
00:18:06,000 --> 00:18:08,840
Speaker 4: doing expense reports the old way, and let's bring them in.

367
00:18:08,960 --> 00:18:12,000
Speaker 4: And if you kind of decompose what that task actually is,

368
00:18:12,359 --> 00:18:15,240
Speaker 4: it's first asking who are these businesses? What is a

369
00:18:15,320 --> 00:18:18,199
Speaker 4: relevant moment? Have they just raised funds, have they just

370
00:18:18,320 --> 00:18:21,359
Speaker 4: hired someone on their finance team, Maybe they're posting an

371
00:18:21,359 --> 00:18:23,840
Speaker 4: open role in some needs. There are these signals out

372
00:18:23,840 --> 00:18:26,600
Speaker 4: there in the world then you notice, Okay, you've got

373
00:18:26,600 --> 00:18:29,080
Speaker 4: your lead lists assembled, you have to go write them.

374
00:18:29,200 --> 00:18:31,520
Speaker 4: Maybe you have a junior person just at a college

375
00:18:31,560 --> 00:18:34,160
Speaker 4: going and joined a Guess what's this person's email, what's

376
00:18:34,200 --> 00:18:36,960
Speaker 4: this person's phone number? How do I get a mail

377
00:18:37,000 --> 00:18:38,760
Speaker 4: or in front, or how do I knock on their door?

378
00:18:39,280 --> 00:18:41,679
Speaker 4: Then they write kind of the message, what's going to

379
00:18:41,680 --> 00:18:45,000
Speaker 4: be compelling. There's all these little steps. What makes a

380
00:18:45,119 --> 00:18:49,040
Speaker 4: human seller great and what makes sales interesting is the

381
00:18:49,080 --> 00:18:52,239
Speaker 4: genuine human connection, someone who can go deep, understand all

382
00:18:52,320 --> 00:18:55,359
Speaker 4: the context and actually close that great sale. And yet,

383
00:18:55,440 --> 00:18:57,960
Speaker 4: if you looked at the task in twenty twenty two,

384
00:18:58,119 --> 00:19:00,840
Speaker 4: how most people were spending their time with things that

385
00:19:00,920 --> 00:19:05,600
Speaker 4: algorithms are great at. Finding people's email address, testing which

386
00:19:05,680 --> 00:19:09,920
Speaker 4: copy will ultimately work better, detecting across vast swaths of data,

387
00:19:09,960 --> 00:19:13,400
Speaker 4: what's the signal in the noise, and so effectively. Part

388
00:19:13,400 --> 00:19:16,000
Speaker 4: of what's powering this level of growth is a broad

389
00:19:16,040 --> 00:19:19,760
Speaker 4: set of AI tools which do exactly that. Where AI

390
00:19:20,040 --> 00:19:23,920
Speaker 4: is finding the person's email, AI is detecting these signals

391
00:19:23,960 --> 00:19:26,960
Speaker 4: for intent sending the message. In the job of an

392
00:19:27,119 --> 00:19:31,480
Speaker 4: entry level salesperson now is the majority to respond to interest,

393
00:19:31,800 --> 00:19:36,000
Speaker 4: to close people on the call, and so that's one example.

394
00:19:36,400 --> 00:19:38,359
Speaker 4: We can go through through several kind of throughout the

395
00:19:38,440 --> 00:19:40,960
Speaker 4: sales cycle, but it's changed the role to be much

396
00:19:41,000 --> 00:19:41,600
Speaker 4: more interesting.

397
00:19:41,920 --> 00:19:45,760
Speaker 3: I'm getting flashbacks to Glen Gary Glenn Ross. Didn't companies

398
00:19:46,040 --> 00:19:49,280
Speaker 3: used to buy lead lists as well?

399
00:19:48,680 --> 00:19:49,840
Speaker 2: They still do?

400
00:19:50,000 --> 00:19:52,840
Speaker 3: Yeah, yeah, right, So I imagine if you can basically build

401
00:19:52,880 --> 00:19:56,040
Speaker 3: your own lead generator, you would save some money as well.

402
00:19:56,600 --> 00:19:59,720
Speaker 4: That's exactly right. But the interesting thing is just this

403
00:20:00,000 --> 00:20:03,880
Speaker 4: feed moves faster. There's suddenly signals of intents. Maybe there's

404
00:20:03,920 --> 00:20:06,840
Speaker 4: an IP that you can back into. This is a

405
00:20:06,880 --> 00:20:09,720
Speaker 4: company has gone to your site five times today. Maybe

406
00:20:09,760 --> 00:20:13,159
Speaker 4: that's a low value list, it's the C list, it's

407
00:20:13,200 --> 00:20:15,439
Speaker 4: not the A grade list. But with kind of a

408
00:20:15,480 --> 00:20:19,080
Speaker 4: modern stack that's sifting across these signals, you can get

409
00:20:19,119 --> 00:20:21,600
Speaker 4: more interesting and so I think there's things like that.

410
00:20:21,720 --> 00:20:25,120
Speaker 4: The other big thing that has just transformed the job

411
00:20:25,200 --> 00:20:27,680
Speaker 4: is there's just so much more noise than there is

412
00:20:27,800 --> 00:20:29,359
Speaker 4: signal out there in the world. So if you're from

413
00:20:29,440 --> 00:20:31,919
Speaker 4: manager and you're trying to help a twenty two year

414
00:20:31,960 --> 00:20:35,040
Speaker 4: old new in their career get better at sales, it's

415
00:20:35,119 --> 00:20:37,960
Speaker 4: just too many hours of calls to listen to across

416
00:20:38,000 --> 00:20:40,720
Speaker 4: your whole team to do that. The large language model

417
00:20:40,800 --> 00:20:44,359
Speaker 4: has no problem listening to a thousand years of calls

418
00:20:44,359 --> 00:20:46,840
Speaker 4: in a single day, more than any human can. And

419
00:20:46,880 --> 00:20:49,400
Speaker 4: so suddenly, I actually think one of the more interesting

420
00:20:49,440 --> 00:20:51,960
Speaker 4: stories and lessons I learned about this actually came from

421
00:20:52,240 --> 00:20:55,159
Speaker 4: one hundred plus year old credit card company, where I

422
00:20:55,240 --> 00:20:57,200
Speaker 4: was first skeptical when I heard the story. The executive

423
00:20:57,240 --> 00:21:00,399
Speaker 4: was explaining how they stopped checking net promoter scores, and

424
00:21:00,440 --> 00:21:02,520
Speaker 4: I'm like, wow, they must have gotten that bad. But

425
00:21:02,560 --> 00:21:05,440
Speaker 4: the truth was something more interesting promoter So a net

426
00:21:05,440 --> 00:21:08,560
Speaker 4: promoter score is this process of often you sample. So

427
00:21:08,640 --> 00:21:11,639
Speaker 4: let's say there's ten thousand people who call in with

428
00:21:11,720 --> 00:21:13,879
Speaker 4: customer support, and you asked them at the end of

429
00:21:13,880 --> 00:21:17,400
Speaker 4: the call from zeroitech, how happy are you with the service?

430
00:21:17,920 --> 00:21:20,760
Speaker 4: And often in these surveys you ask a small sample

431
00:21:20,800 --> 00:21:22,840
Speaker 4: and you can see an aggregate how good are your agents?

432
00:21:23,680 --> 00:21:26,520
Speaker 4: The response was much more interesting. What they started doing

433
00:21:26,600 --> 00:21:29,440
Speaker 4: was applying a large language model to listen to all

434
00:21:29,520 --> 00:21:33,119
Speaker 4: calls simultaneously, and they could apply sentiment analysis from the

435
00:21:33,119 --> 00:21:35,960
Speaker 4: tone of a voice how happy were customers or not.

436
00:21:36,520 --> 00:21:40,480
Speaker 4: And what the algorithm started doing was routing calls from

437
00:21:40,520 --> 00:21:44,280
Speaker 4: customers increasingly to people who did a great job that

438
00:21:44,320 --> 00:21:45,440
Speaker 4: made customers happy.

439
00:21:45,920 --> 00:21:47,640
Speaker 3: It's always the case if you're good at your job,

440
00:21:47,720 --> 00:21:49,280
Speaker 3: you get more, that's the reward.

441
00:21:51,200 --> 00:21:53,760
Speaker 4: But it's a real way suddenly, like you can actually

442
00:21:53,800 --> 00:21:56,920
Speaker 4: tell from every customer how are they doing, and actually

443
00:21:57,359 --> 00:21:59,440
Speaker 4: get more output per unit of input. It's a real

444
00:21:59,440 --> 00:22:00,800
Speaker 4: ease case. I don't know how you do a few

445
00:22:00,880 --> 00:22:01,520
Speaker 4: years ago.

446
00:22:01,880 --> 00:22:05,560
Speaker 3: And I'm sorry, you were talking about where you're deploying

447
00:22:05,640 --> 00:22:08,920
Speaker 3: AI and you had engineering and sales, and we didn't

448
00:22:08,960 --> 00:22:10,080
Speaker 3: get to the third one.

449
00:22:10,359 --> 00:22:12,880
Speaker 4: Oh sure, yeah, we'll see if it works. Maybe most

450
00:22:12,880 --> 00:22:16,040
Speaker 4: avant garde would actually be kind of in growth and marketing.

451
00:22:16,040 --> 00:22:18,080
Speaker 4: Of course, there's customer support in other areas that that

452
00:22:18,119 --> 00:22:21,119
Speaker 4: people have talked about. But I actually think there's a

453
00:22:21,160 --> 00:22:24,439
Speaker 4: real case that marketing and growth is becoming a technical function.

454
00:22:24,600 --> 00:22:26,600
Speaker 4: People thought that art would be one of the last

455
00:22:26,640 --> 00:22:31,000
Speaker 4: use cases, and suddenly you can create images, videos, interesting things,

456
00:22:31,040 --> 00:22:33,600
Speaker 4: And you know, now in a world where you can

457
00:22:33,600 --> 00:22:35,080
Speaker 4: go and start to do this, you can see based

458
00:22:35,080 --> 00:22:37,280
Speaker 4: on intent on conversion, can you start to combine the

459
00:22:37,280 --> 00:22:41,960
Speaker 4: mathematical function and what really works with people to creating beautiful, striking,

460
00:22:42,000 --> 00:22:45,280
Speaker 4: interesting images on demand and so the jury is still

461
00:22:45,280 --> 00:22:47,480
Speaker 4: most out with that one, but it's been real I

462
00:22:47,480 --> 00:22:49,480
Speaker 4: think early tests are promising. And I think that the

463
00:22:49,480 --> 00:22:52,080
Speaker 4: early example of this that to me in some ways

464
00:22:52,119 --> 00:22:55,159
Speaker 4: kind of inspired this was maybe twofold one Amazon. They

465
00:22:55,160 --> 00:22:58,119
Speaker 4: had kind of the editorial and the personalization team. It

466
00:22:58,240 --> 00:23:00,480
Speaker 4: used to be that folks at Amazon actually wrote and

467
00:23:00,560 --> 00:23:03,200
Speaker 4: recommended you bought you know, here's a newsletter of all

468
00:23:03,200 --> 00:23:04,720
Speaker 4: the things that we have, and it was like the

469
00:23:04,760 --> 00:23:07,800
Speaker 4: series Robucks catalog. Eventually it became you bought this, you

470
00:23:07,880 --> 00:23:10,440
Speaker 4: might be interested in that, and eventually things went out,

471
00:23:10,480 --> 00:23:12,159
Speaker 4: And so you may be able to do that not

472
00:23:12,240 --> 00:23:14,879
Speaker 4: just with clustering of items, but actually understanding people and

473
00:23:14,920 --> 00:23:17,840
Speaker 4: making beautiful things. And you know, I think of inspirations

474
00:23:17,840 --> 00:23:20,280
Speaker 4: in New York like Andy Warhol his factory, where every

475
00:23:20,359 --> 00:23:23,320
Speaker 4: day they'd make something new, things were incredibly striking and

476
00:23:23,359 --> 00:23:26,520
Speaker 4: had this method of new levels of art thinking about

477
00:23:26,760 --> 00:23:30,760
Speaker 4: being incredibly generative. And I think that every industry in

478
00:23:30,840 --> 00:23:34,360
Speaker 4: some sense can be improved and augmented and made more

479
00:23:34,400 --> 00:23:37,919
Speaker 4: interesting actually through kind of the just the leverage that

480
00:23:37,960 --> 00:23:38,840
Speaker 4: these tools can bring.

481
00:23:39,480 --> 00:23:41,960
Speaker 2: I'm going to be cynical for a second, and I'm

482
00:23:41,960 --> 00:23:44,080
Speaker 2: going to press you a little bit further on the

483
00:23:44,160 --> 00:23:47,440
Speaker 2: question of AI and sales, because I certainly take your

484
00:23:47,520 --> 00:23:51,359
Speaker 2: point about searching for signal what companies might be in

485
00:23:51,359 --> 00:23:54,639
Speaker 2: a position suddenly where they're looking to upgrade their expense

486
00:23:54,720 --> 00:23:57,480
Speaker 2: management platform, or what might be the person who best

487
00:23:57,480 --> 00:24:01,359
Speaker 2: to call some of that, I imagine is stuff that someone

488
00:24:01,480 --> 00:24:04,600
Speaker 2: was selling via Salesforce for a while and calling it

489
00:24:04,720 --> 00:24:09,200
Speaker 2: five years ago machine learning or ten years ago machine learning. Obviously,

490
00:24:09,480 --> 00:24:12,560
Speaker 2: what is technically AI is one of these things that

491
00:24:12,840 --> 00:24:15,560
Speaker 2: people argue over to get a better multiple, et cetera.

492
00:24:16,119 --> 00:24:18,560
Speaker 2: But for the purposes of what many people in the

493
00:24:18,560 --> 00:24:20,840
Speaker 2: stock market are excited about a lot of it's the

494
00:24:20,920 --> 00:24:24,000
Speaker 2: sort of you know, the post twenty twenty two generative

495
00:24:24,040 --> 00:24:27,240
Speaker 2: AI that somehow comes in the lab and was inferred

496
00:24:27,320 --> 00:24:30,360
Speaker 2: on an in video chip or something like that. Say

497
00:24:30,400 --> 00:24:33,439
Speaker 2: more about sales and what are what other things that

498
00:24:33,480 --> 00:24:36,240
Speaker 2: you can do in sales today in twenty twenty five

499
00:24:36,680 --> 00:24:38,880
Speaker 2: that you could not have done in twenty twenty one.

500
00:24:39,600 --> 00:24:42,760
Speaker 4: I think that probably the most exciting area has to

501
00:24:42,760 --> 00:24:45,000
Speaker 4: do with reasoning. When I think about a lot of

502
00:24:45,080 --> 00:24:49,359
Speaker 4: classic machine learning, which is still deeply important, it's often

503
00:24:49,560 --> 00:24:53,280
Speaker 4: around correlation of certain variables and prediction in a very

504
00:24:53,359 --> 00:24:56,159
Speaker 4: narrow sense. And so maybe the first phase is of

505
00:24:56,200 --> 00:24:59,119
Speaker 4: machine learning is prediction of what comes next? You know,

506
00:24:59,200 --> 00:25:01,400
Speaker 4: my frame around a lot of this is now there's

507
00:25:01,440 --> 00:25:04,679
Speaker 4: generation based on what connects? What do you create? And

508
00:25:04,760 --> 00:25:06,880
Speaker 4: I think the most interesting when you think about kind

509
00:25:06,880 --> 00:25:11,000
Speaker 4: of the one models, three and the new reasoning models,

510
00:25:11,000 --> 00:25:13,639
Speaker 4: where it's thinking and can think multiple steps ahead. Some

511
00:25:13,720 --> 00:25:16,080
Speaker 4: of the techniques that made Alpha go so good at

512
00:25:16,080 --> 00:25:18,360
Speaker 4: what it's doing, it's those types of work. And so

513
00:25:18,600 --> 00:25:21,280
Speaker 4: you're exactly right. Some of this is actually just good

514
00:25:21,400 --> 00:25:24,760
Speaker 4: data infrastructure and prediction. But when you start to fuse

515
00:25:24,840 --> 00:25:27,440
Speaker 4: that with based off of these signals, what maybe should

516
00:25:27,480 --> 00:25:30,280
Speaker 4: we write based off of the context of the calls

517
00:25:30,680 --> 00:25:33,000
Speaker 4: and the usage of this data, how do we follow

518
00:25:33,040 --> 00:25:36,600
Speaker 4: up and orchestrate it across multiple steps. That's where I

519
00:25:36,640 --> 00:25:39,040
Speaker 4: think generative AI starts to get really interesting in these

520
00:25:39,080 --> 00:25:43,120
Speaker 4: boring use cases like expense management and saving people time

521
00:25:43,160 --> 00:25:45,159
Speaker 4: and money, where it starts to get very useful.

522
00:26:00,200 --> 00:26:03,080
Speaker 3: Can you talk a little bit more about what data

523
00:26:03,160 --> 00:26:07,280
Speaker 3: exactly you're scraping to infer those specific signals, like what

524
00:26:07,320 --> 00:26:08,760
Speaker 3: do you have access to and what do you find

525
00:26:08,800 --> 00:26:09,439
Speaker 3: most useful?

526
00:26:09,720 --> 00:26:12,479
Speaker 4: Yeah, so, first some of this is just let's take

527
00:26:12,640 --> 00:26:14,719
Speaker 4: a use case of Let's say an account manager. They

528
00:26:14,800 --> 00:26:18,679
Speaker 4: might be overseeing hundreds of individual accounts at ramp and

529
00:26:18,760 --> 00:26:20,199
Speaker 4: their goal is to get back on the phone and

530
00:26:20,240 --> 00:26:22,159
Speaker 4: make sure people are getting value out of it. We

531
00:26:22,200 --> 00:26:24,000
Speaker 4: want to save your business time and money. We want

532
00:26:24,000 --> 00:26:26,160
Speaker 4: to made automate accounting. But is it set up that way?

533
00:26:26,400 --> 00:26:28,239
Speaker 4: Are you seeing these terms of use cases? And so

534
00:26:28,280 --> 00:26:30,320
Speaker 4: some of this is going to be internal use cases

535
00:26:30,720 --> 00:26:32,760
Speaker 4: based off of spend that you wanted to bring over

536
00:26:32,880 --> 00:26:36,960
Speaker 4: versus which you actually spent. How is that going? Log

537
00:26:37,040 --> 00:26:40,280
Speaker 4: data large language models can remember all calls, all notes,

538
00:26:40,600 --> 00:26:42,520
Speaker 4: what people kind of committed to you by, what dates

539
00:26:42,520 --> 00:26:44,680
Speaker 4: and do that, and so when an account manager gets

540
00:26:44,720 --> 00:26:47,000
Speaker 4: on the phone, they can go and have all the

541
00:26:47,080 --> 00:26:49,520
Speaker 4: right contact in front of them that they didn't need

542
00:26:49,560 --> 00:26:51,719
Speaker 4: to go and spend hours the night before. Kind of creating,

543
00:26:51,720 --> 00:26:53,520
Speaker 4: but it's pulled up and pulled in terms of what's

544
00:26:53,560 --> 00:26:56,400
Speaker 4: most useful. You next may have interactive data can pull

545
00:26:56,440 --> 00:26:59,440
Speaker 4: through based on the website itself where their flows where

546
00:26:59,440 --> 00:27:01,760
Speaker 4: it appeared well just got stuck and confused. Maybe they

547
00:27:01,760 --> 00:27:04,480
Speaker 4: wanted to go close the books and it seemed like

548
00:27:04,800 --> 00:27:07,600
Speaker 4: too many steps and they kind of paused, we can

549
00:27:07,640 --> 00:27:07,960
Speaker 4: kind of.

550
00:27:08,080 --> 00:27:11,520
Speaker 3: So you're tracking like actual physical movements on the website

551
00:27:11,600 --> 00:27:12,840
Speaker 3: through beacons, I assume.

552
00:27:13,200 --> 00:27:15,600
Speaker 4: Yeah, Yeah, so you can do things like that coupled

553
00:27:15,600 --> 00:27:18,920
Speaker 4: with even external data as well. Sometimes a company could

554
00:27:18,960 --> 00:27:22,440
Speaker 4: be doing really well announce the fundraise, you know, need

555
00:27:22,480 --> 00:27:25,000
Speaker 4: to expand and make sure hey we've seen this good news.

556
00:27:25,040 --> 00:27:27,119
Speaker 4: We want to make sure that we're expanding with you.

557
00:27:27,560 --> 00:27:30,800
Speaker 4: And so I think that any one individual point a

558
00:27:30,880 --> 00:27:33,800
Speaker 4: person can do. But often the fullness of what does

559
00:27:33,840 --> 00:27:36,240
Speaker 4: it take to really understand and be a partner with

560
00:27:36,320 --> 00:27:39,680
Speaker 4: how are one to thousands of employees at a single

561
00:27:39,680 --> 00:27:42,280
Speaker 4: individual customer, how are they actually doing so we can

562
00:27:42,280 --> 00:27:44,639
Speaker 4: be a more useful partner is often where this comes together.

563
00:27:44,760 --> 00:27:45,560
Speaker 4: Does that make sense?

564
00:27:45,600 --> 00:27:48,400
Speaker 3: It does, although I have maybe this is a weird question,

565
00:27:48,760 --> 00:27:54,120
Speaker 3: but like how does the system actually generate its suggestions?

566
00:27:54,200 --> 00:27:57,920
Speaker 3: So if I'm a salesperson and something like the system

567
00:27:58,000 --> 00:28:01,399
Speaker 3: spots a lead of some sort and it thinks you

568
00:28:01,440 --> 00:28:04,159
Speaker 3: should get in touch with this person because of whatever reason,

569
00:28:04,520 --> 00:28:06,159
Speaker 3: how does that message actually get to me?

570
00:28:06,680 --> 00:28:09,680
Speaker 4: Yeah, And I think that's exactly in some sense probably

571
00:28:09,720 --> 00:28:12,080
Speaker 4: the million dollar question of which SaaS companies are going

572
00:28:12,160 --> 00:28:14,000
Speaker 4: to do really well and I think even the story

573
00:28:14,040 --> 00:28:17,600
Speaker 4: of technology now is you have different people with different aspects.

574
00:28:17,640 --> 00:28:19,800
Speaker 4: Some are in the browser, some have this sales tool,

575
00:28:19,920 --> 00:28:22,840
Speaker 4: some have this data tool, some are a data warehouse,

576
00:28:22,880 --> 00:28:25,479
Speaker 4: some are a training tool. Where does it show up?

577
00:28:25,480 --> 00:28:28,119
Speaker 4: And I can tell you we leave it ultimately in

578
00:28:28,200 --> 00:28:31,040
Speaker 4: the hands of our sales team and the growth team

579
00:28:31,040 --> 00:28:33,639
Speaker 4: building for them. Of whatever works, you are free to

580
00:28:33,720 --> 00:28:36,800
Speaker 4: pick given the period of change, but for them often,

581
00:28:37,200 --> 00:28:38,880
Speaker 4: like I'll tell you, one of the more useful tools

582
00:28:38,880 --> 00:28:40,800
Speaker 4: for that use case for account managers is a company

583
00:28:40,840 --> 00:28:44,520
Speaker 4: called Rocks. I guess like rox dot com. They're less

584
00:28:44,520 --> 00:28:47,560
Speaker 4: than a year old company, but effectively are polling data

585
00:28:47,560 --> 00:28:52,720
Speaker 4: from Salesforce usage data, internal level data analytics, and appends

586
00:28:52,800 --> 00:28:57,560
Speaker 4: notes for an account manager's calendar prior to meeting. Here's

587
00:28:57,600 --> 00:28:59,520
Speaker 4: the core things to know, here's lengths and if you

588
00:28:59,520 --> 00:29:00,880
Speaker 4: want to pull, so you can get things out a

589
00:29:00,920 --> 00:29:03,560
Speaker 4: glance in your calendar, and you can actually pull from

590
00:29:03,600 --> 00:29:07,040
Speaker 4: the website itself. Here's more data to go see it.

591
00:29:07,120 --> 00:29:09,000
Speaker 4: So it's they're trying to become a bit of a

592
00:29:09,040 --> 00:29:11,960
Speaker 4: mission command for sales. Of course, Salesforce is trying to

593
00:29:11,960 --> 00:29:15,360
Speaker 4: do these things too. To in engineering, there's tools like

594
00:29:15,400 --> 00:29:19,200
Speaker 4: cursor and Devin have different bets. Cursors like get hub

595
00:29:19,280 --> 00:29:22,560
Speaker 4: Coat pilots sort of won the love of many developers.

596
00:29:22,560 --> 00:29:25,080
Speaker 4: Where as you're coding, it's like a better auto complete

597
00:29:25,280 --> 00:29:27,320
Speaker 4: that you just say I want to build this app

598
00:29:27,320 --> 00:29:29,240
Speaker 4: and it can go ahead. It can audit lines of

599
00:29:29,280 --> 00:29:32,480
Speaker 4: codes and knows your repo to. There's stranger bets like

600
00:29:32,560 --> 00:29:35,760
Speaker 4: Devin and cognition, where the form factor of that is

601
00:29:36,160 --> 00:29:39,080
Speaker 4: it is meant to be a digital AI engineer and

602
00:29:39,120 --> 00:29:41,360
Speaker 4: you can tell Devin, I want to build this app.

603
00:29:41,640 --> 00:29:44,280
Speaker 4: Go research these websites make it look like this. Come

604
00:29:44,320 --> 00:29:47,920
Speaker 4: back to me when you have questions I get from

605
00:29:47,960 --> 00:29:49,040
Speaker 4: my app. Try it.

606
00:29:49,280 --> 00:29:51,320
Speaker 2: I just like I was trying to someone. I forget

607
00:29:51,360 --> 00:29:53,120
Speaker 2: who it was. Maybe there's even someone from one of

608
00:29:53,160 --> 00:29:56,000
Speaker 2: these companies. Never talk about like different approaches to some

609
00:29:56,080 --> 00:29:59,440
Speaker 2: of themself, including ones where like the AI basically controlled

610
00:29:59,440 --> 00:30:02,840
Speaker 2: the mouse, yeah, controlled the cursor, yes, and clicked on

611
00:30:02,920 --> 00:30:06,240
Speaker 2: websites and write websites. So we think of API calls,

612
00:30:06,360 --> 00:30:09,360
Speaker 2: but there's a different model where it's just like you're

613
00:30:09,400 --> 00:30:11,040
Speaker 2: scanning the website like a human.

614
00:30:11,200 --> 00:30:13,840
Speaker 4: Is right, You are exactly right. And this is one

615
00:30:13,840 --> 00:30:16,520
Speaker 4: of the strangest and most interesting things about the time

616
00:30:16,560 --> 00:30:19,120
Speaker 4: that we live in, and that computers can kind of think,

617
00:30:19,200 --> 00:30:21,400
Speaker 4: they can kind of see, they can kind of hear

618
00:30:21,480 --> 00:30:24,520
Speaker 4: and process different levels of data. And I think that

619
00:30:24,600 --> 00:30:28,480
Speaker 4: the general's story of computing is increasing levels of abstraction.

620
00:30:28,920 --> 00:30:32,000
Speaker 4: Back eighty years ago people were writing machine codes, it

621
00:30:32,040 --> 00:30:35,120
Speaker 4: was one zero binary, and over time it went to

622
00:30:35,160 --> 00:30:40,240
Speaker 4: see the Python which were increasingly higher order compromises between

623
00:30:40,600 --> 00:30:41,960
Speaker 4: language U and I speech.

624
00:30:41,880 --> 00:30:45,000
Speaker 2: Separate from binary like layers and layers of a binary.

625
00:30:45,080 --> 00:30:45,360
Speaker 1: Yeah.

626
00:30:45,400 --> 00:30:47,880
Speaker 4: And there are many people who would make the argument,

627
00:30:48,200 --> 00:30:51,320
Speaker 4: I think convincingly so that the next programming language, in

628
00:30:51,360 --> 00:30:55,200
Speaker 4: fact is English language. Yeah, and what you see? And

629
00:30:55,280 --> 00:30:58,640
Speaker 4: so I even think of now, what many people talk

630
00:30:58,640 --> 00:31:01,560
Speaker 4: about one of these coming battle rounds is actually what

631
00:31:01,640 --> 00:31:04,520
Speaker 4: are you seeing on your computer? And the best interface

632
00:31:04,600 --> 00:31:07,560
Speaker 4: is not chat, but actually it's just a large language

633
00:31:07,560 --> 00:31:10,400
Speaker 4: model that sees everything you're doing on your computer and

634
00:31:10,480 --> 00:31:12,600
Speaker 4: can predict what's next. Who knows it's nice.

635
00:31:12,640 --> 00:31:15,320
Speaker 2: Well, when I keep getting error messages in my Google collab,

636
00:31:15,360 --> 00:31:17,880
Speaker 2: I just take screenshots and then I upload them to

637
00:31:17,960 --> 00:31:20,240
Speaker 2: chat GBT, and I say, what's this error message mean?

638
00:31:20,400 --> 00:31:20,960
Speaker 3: Yeah?

639
00:31:21,000 --> 00:31:23,160
Speaker 2: And then I teld me one thing I've wondered about

640
00:31:23,280 --> 00:31:25,520
Speaker 2: and you know, people are very anxious about what are

641
00:31:25,560 --> 00:31:28,120
Speaker 2: the jobs of the future and all these things and

642
00:31:28,160 --> 00:31:30,920
Speaker 2: what professions are going to get disrupted away, and you know,

643
00:31:30,960 --> 00:31:33,719
Speaker 2: we people could speculate on this forever. But one thing

644
00:31:33,760 --> 00:31:36,680
Speaker 2: I've like wondered about is like there are certain jobs

645
00:31:36,960 --> 00:31:39,640
Speaker 2: where to be good at them, you have to have

646
00:31:39,720 --> 00:31:42,880
Speaker 2: probably like sacrificed years in your life and like not

647
00:31:43,040 --> 00:31:46,160
Speaker 2: gone to parties and not head friends and you know

648
00:31:46,320 --> 00:31:51,360
Speaker 2: not because like the true artists, right, No, the technical

649
00:31:51,440 --> 00:31:54,080
Speaker 2: skilled required just so years and years and years, et cetera.

650
00:31:54,320 --> 00:31:55,960
Speaker 2: And so you had to be the type of person

651
00:31:55,960 --> 00:31:57,960
Speaker 2: that was willing to sacrifice a lot to get good

652
00:31:58,040 --> 00:32:00,800
Speaker 2: at them or to build up that technical skill right

653
00:32:01,120 --> 00:32:04,360
Speaker 2: those hours. And I'm wondering, like if AI is going

654
00:32:04,400 --> 00:32:08,200
Speaker 2: to sort of cut into the jobs, We're a major

655
00:32:08,280 --> 00:32:10,600
Speaker 2: part of getting good at it was this sort of

656
00:32:10,640 --> 00:32:14,800
Speaker 2: being willing to like sacrifice being a normal person because

657
00:32:15,040 --> 00:32:18,080
Speaker 2: the AI could just do thousands, doesn't need to sacrifice anything.

658
00:32:18,120 --> 00:32:21,680
Speaker 2: It's just a computer, and that it would benefit the people, like,

659
00:32:22,800 --> 00:32:24,920
Speaker 2: you know, the low rounded people who I think want

660
00:32:24,960 --> 00:32:27,080
Speaker 2: to have. Yeah, you bring some MiQ to work, but

661
00:32:27,120 --> 00:32:27,880
Speaker 2: you also bring some.

662
00:32:28,320 --> 00:32:29,440
Speaker 3: Like emotional idea.

663
00:32:29,520 --> 00:32:31,360
Speaker 2: Yeah, exactly. This is what I'm wondering about.

664
00:32:32,160 --> 00:32:35,360
Speaker 4: I think this is exactly the right line of questions,

665
00:32:35,360 --> 00:32:36,600
Speaker 4: and I think is going to be a real one

666
00:32:36,640 --> 00:32:39,200
Speaker 4: to confront And I think in some sense, I think

667
00:32:39,240 --> 00:32:41,600
Speaker 4: even probably listeners of this podcast fall into this group

668
00:32:41,680 --> 00:32:44,440
Speaker 4: of like very curious people and people who know how

669
00:32:44,440 --> 00:32:48,640
Speaker 4: to ask interesting questions, keep going down it and create things.

670
00:32:48,720 --> 00:32:51,040
Speaker 4: I think we'll do very well in the future. But

671
00:32:51,240 --> 00:32:55,200
Speaker 4: you yeah, it's all gonna be okay, don't worry now.

672
00:32:55,240 --> 00:32:56,960
Speaker 4: But I think things are going to change a lot,

673
00:32:57,200 --> 00:32:59,600
Speaker 4: Like even in thinking about coming here today, like I

674
00:32:59,880 --> 00:33:02,560
Speaker 4: was curious at the turn of the century. In nineteen hundred,

675
00:33:02,960 --> 00:33:06,200
Speaker 4: I think in the United States, forty percent of all

676
00:33:06,320 --> 00:33:10,280
Speaker 4: jobs were in farming. You know, today less than one

677
00:33:10,280 --> 00:33:12,600
Speaker 4: percent of all jobs are in farming. And I think

678
00:33:12,640 --> 00:33:15,480
Speaker 4: things are okay currently, but I think that the nature

679
00:33:15,480 --> 00:33:17,120
Speaker 4: of job is going to change, probably in ways that

680
00:33:17,120 --> 00:33:19,400
Speaker 4: will be very hard for people, just as in nineteen

681
00:33:19,480 --> 00:33:22,240
Speaker 4: hundred to now to predict probably the same for us

682
00:33:22,280 --> 00:33:24,440
Speaker 4: and what it may look like in fifty one hundred years.

683
00:33:24,800 --> 00:33:30,760
Speaker 3: How are you or your clients handling privacy and legal

684
00:33:30,880 --> 00:33:34,760
Speaker 3: slash copyright concerns with some of these platforms.

685
00:33:34,440 --> 00:33:36,880
Speaker 4: Yeah, Well, first of all, like we have a bit

686
00:33:36,920 --> 00:33:40,080
Speaker 4: of a simpler time on at least copyright and that

687
00:33:40,120 --> 00:33:42,160
Speaker 4: we're not in the business of generation of how do

688
00:33:42,200 --> 00:33:44,560
Speaker 4: you go and create new art and images? And I

689
00:33:44,560 --> 00:33:48,480
Speaker 4: think those are real and present questions, especially for content generation.

690
00:33:48,600 --> 00:33:51,160
Speaker 4: A lot of what we're doing today is you've made

691
00:33:51,160 --> 00:33:54,840
Speaker 4: this card transaction, You've texted back thirty seconds later, here's

692
00:33:54,840 --> 00:33:57,120
Speaker 4: a photo of the receipt, and then we can match. Okay,

693
00:33:57,160 --> 00:33:59,400
Speaker 4: here's what the memo should be, Here's what the accounting

694
00:33:59,400 --> 00:34:01,520
Speaker 4: category should be, and you're done. And so a lot

695
00:34:01,520 --> 00:34:04,920
Speaker 4: of this is process automation and workflow automation, and there's

696
00:34:04,960 --> 00:34:07,520
Speaker 4: an interesting value we can bring and probably the most

697
00:34:07,520 --> 00:34:09,360
Speaker 4: interesting part of our business we need to think a

698
00:34:09,360 --> 00:34:13,799
Speaker 4: lot about this is on this area of price intelligence. Now,

699
00:34:13,840 --> 00:34:16,520
Speaker 4: as a consumer, you can go on to Zillo and

700
00:34:16,640 --> 00:34:19,560
Speaker 4: know here's what your home maybe is worth. You could

701
00:34:19,600 --> 00:34:21,520
Speaker 4: go on true car and get a sense of what

702
00:34:21,520 --> 00:34:23,400
Speaker 4: should you pay for this car based on lots of

703
00:34:23,440 --> 00:34:26,120
Speaker 4: aggregate and auontomized data. And for businesses it's very useful

704
00:34:26,160 --> 00:34:28,360
Speaker 4: to know what is this business paying down the street

705
00:34:28,719 --> 00:34:32,759
Speaker 4: for this supply or for this salesforce license, what are

706
00:34:32,800 --> 00:34:35,120
Speaker 4: people paying and so in that part of our business,

707
00:34:35,200 --> 00:34:36,520
Speaker 4: we actually do want to be able to go to

708
00:34:36,560 --> 00:34:40,160
Speaker 4: customers and say, you know, here's where you're pricing compares

709
00:34:40,200 --> 00:34:42,040
Speaker 4: to the rest of the market, and we want to

710
00:34:42,040 --> 00:34:44,200
Speaker 4: help you negotiate for lower prices. And we think it's

711
00:34:44,239 --> 00:34:46,160
Speaker 4: really good the way you get there. And I think

712
00:34:46,160 --> 00:34:50,160
Speaker 4: probably the core of our strategy around this is aggregated

713
00:34:50,280 --> 00:34:53,200
Speaker 4: anonymized is really the part that needs to come or

714
00:34:53,520 --> 00:34:56,880
Speaker 4: if it's individual data that really sits on ramp and

715
00:34:57,000 --> 00:34:59,520
Speaker 4: is for the purpose of providing service to you. And

716
00:34:59,560 --> 00:35:02,160
Speaker 4: if we share things broadly, it's give get. If you

717
00:35:02,200 --> 00:35:04,920
Speaker 4: want to see pricing data, you need to share in it.

718
00:35:05,000 --> 00:35:09,400
Speaker 4: But we need to have enough data to effectively anonymize

719
00:35:09,480 --> 00:35:11,000
Speaker 4: like what's coming out there. And so we think it's

720
00:35:11,080 --> 00:35:13,879
Speaker 4: very useful for finance teams for business owners to help

721
00:35:13,920 --> 00:35:15,720
Speaker 4: them pay less. It's part of what helps in average

722
00:35:15,719 --> 00:35:18,560
Speaker 4: customer cut expenses I about five percent per year, but

723
00:35:18,880 --> 00:35:21,360
Speaker 4: maybe less good news for people who try to discriminate

724
00:35:21,400 --> 00:35:22,680
Speaker 4: against our customers.

725
00:35:23,200 --> 00:35:26,800
Speaker 2: We are in an age of cracking down on spending

726
00:35:26,880 --> 00:35:30,400
Speaker 2: and cracking down on wasteful spending in DC for example,

727
00:35:30,920 --> 00:35:33,440
Speaker 2: and people are talking about doge et cetera. And we're

728
00:35:33,440 --> 00:35:35,640
Speaker 2: going to really move the dial on spending. It's not

729
00:35:35,680 --> 00:35:37,480
Speaker 2: going to be wasteful spending. It's going to be We're

730
00:35:37,520 --> 00:35:40,239
Speaker 2: going to have to like actually have different priorities. Nonetheless,

731
00:35:40,480 --> 00:35:43,279
Speaker 2: cracking down on waste seems good from the perspective of

732
00:35:43,320 --> 00:35:48,440
Speaker 2: an expense management platform. What are some fingerprints of wasteful

733
00:35:48,480 --> 00:35:51,040
Speaker 2: spending that you see? Can you see into companies and

734
00:35:51,160 --> 00:35:54,120
Speaker 2: like see the fingerprints of waste? Do you have any

735
00:35:54,160 --> 00:35:56,759
Speaker 2: advice or things that you look for when if you

736
00:35:56,840 --> 00:35:59,000
Speaker 2: wanted to hunt waste in spending.

737
00:35:59,280 --> 00:36:02,000
Speaker 4: I'm happy that people are thinking about this in a

738
00:36:02,040 --> 00:36:04,040
Speaker 4: real way because often there's an obsession of how do

739
00:36:04,080 --> 00:36:07,200
Speaker 4: you spend more? But if you want to make a change,

740
00:36:07,280 --> 00:36:08,560
Speaker 4: it actually comes.

741
00:36:08,360 --> 00:36:11,439
Speaker 2: Other than putting the entire government on rant. So other

742
00:36:11,480 --> 00:36:14,600
Speaker 2: than that, I assume you support that, But what's the

743
00:36:14,600 --> 00:36:15,360
Speaker 2: next thing you have to do?

744
00:36:15,600 --> 00:36:18,680
Speaker 4: Look, maybe one of the founding fathers, Ben Franklin, I think,

745
00:36:18,760 --> 00:36:21,120
Speaker 4: was known for saying a penny saved is a penny earned.

746
00:36:21,520 --> 00:36:24,640
Speaker 4: And if you look at reasonably efficient organizations, like an

747
00:36:24,680 --> 00:36:27,640
Speaker 4: average American company, they have a profit margin of about

748
00:36:27,640 --> 00:36:30,400
Speaker 4: eight and a half percent, So mathematically, a penny saved

749
00:36:30,440 --> 00:36:34,120
Speaker 4: is actually twelve earned is the same. And I think

750
00:36:34,160 --> 00:36:37,200
Speaker 4: about organizations like the government, which I have a lot

751
00:36:37,200 --> 00:36:40,040
Speaker 4: of empathy. They have a lot more complex constituencies and

752
00:36:40,120 --> 00:36:44,040
Speaker 4: needs than a profit making entity. But the exercise of

753
00:36:44,080 --> 00:36:47,080
Speaker 4: cutting costs has not been taken seriously for a very

754
00:36:47,160 --> 00:36:50,520
Speaker 4: long time, and it's led to very different behaviors. If

755
00:36:50,520 --> 00:36:53,520
Speaker 4: you want to sell to the government today, often with

756
00:36:53,560 --> 00:36:58,000
Speaker 4: the selection criteria is can you last through a one

757
00:36:58,040 --> 00:37:02,040
Speaker 4: to two year request for and RFP and process, which

758
00:37:02,080 --> 00:37:05,400
Speaker 4: is very different than how most companies and people select

759
00:37:05,400 --> 00:37:08,280
Speaker 4: for things. What has the most value, what's the lowest cost?

760
00:37:08,600 --> 00:37:11,200
Speaker 4: What can I try and see if it's done next.

761
00:37:11,280 --> 00:37:14,920
Speaker 4: The other very counterintuitive thing that's interesting about the government is,

762
00:37:15,440 --> 00:37:16,799
Speaker 4: you know, you would think that as one of the

763
00:37:16,960 --> 00:37:19,919
Speaker 4: largest buyers of anything in the world, we would get

764
00:37:19,920 --> 00:37:23,200
Speaker 4: that discount for all taxpayers if you're buying a million

765
00:37:23,239 --> 00:37:25,839
Speaker 4: licenses for a piece of software volume license. But it's

766
00:37:25,840 --> 00:37:29,160
Speaker 4: not that way. The typical way that government buys is

767
00:37:29,200 --> 00:37:32,920
Speaker 4: to pay sticker price and to not have discounts. And

768
00:37:33,000 --> 00:37:36,680
Speaker 4: so you see not only are different agencies paying very

769
00:37:36,719 --> 00:37:39,400
Speaker 4: different prices for the same sets of goods, but you

770
00:37:39,440 --> 00:37:42,080
Speaker 4: don't see normal common sense things of hey, we should

771
00:37:42,080 --> 00:37:45,480
Speaker 4: have a group discount for people in buying to last

772
00:37:45,880 --> 00:37:50,040
Speaker 4: a lot of the tools. Because procurement cycles are fifteen years,

773
00:37:50,360 --> 00:37:53,359
Speaker 4: you can't break contracts. You're paying full rates. So these

774
00:37:53,360 --> 00:37:54,960
Speaker 4: companies are going to sue you if you try to

775
00:37:55,000 --> 00:37:58,200
Speaker 4: go and do this. You start seeing some crazy things

776
00:37:58,520 --> 00:38:01,680
Speaker 4: in terms of the actual tools cells. The spend management

777
00:38:01,800 --> 00:38:05,480
Speaker 4: architecture that the government is using was primarily selected in

778
00:38:05,520 --> 00:38:08,680
Speaker 4: the early two thousands, and so whereas the private market

779
00:38:08,719 --> 00:38:11,319
Speaker 4: today can tap a card, their expense report has done

780
00:38:11,320 --> 00:38:14,000
Speaker 4: for them, their books are kept for them. The result

781
00:38:14,200 --> 00:38:16,239
Speaker 4: is that you have an incredible amount of waste of

782
00:38:16,239 --> 00:38:19,640
Speaker 4: people's time, of really hard working people in some cases

783
00:38:19,680 --> 00:38:22,359
Speaker 4: actually spending most of their time trudging through the bureaucracy

784
00:38:22,400 --> 00:38:24,759
Speaker 4: of old tools that don't work to each other. To

785
00:38:24,800 --> 00:38:27,239
Speaker 4: The most shocking I would say, is like you don't

786
00:38:27,239 --> 00:38:29,680
Speaker 4: have to look hard. There's several friends of different agencies

787
00:38:29,719 --> 00:38:32,160
Speaker 4: who in trying to learn and understand some of these

788
00:38:32,200 --> 00:38:34,480
Speaker 4: DOSEE efforts. A shocking thing I learned is that at

789
00:38:34,520 --> 00:38:37,879
Speaker 4: multiple agencies, four hours per night email is just shut down.

790
00:38:38,040 --> 00:38:41,319
Speaker 4: You can't send, you can't receive. It's crazy, right.

791
00:38:41,760 --> 00:38:44,680
Speaker 2: I've seen that government websites have like a time.

792
00:38:45,000 --> 00:38:50,120
Speaker 4: That's exactly right because they're using old private servers contracts

793
00:38:50,120 --> 00:38:52,279
Speaker 4: from thirty years ago. And so if you want to

794
00:38:52,320 --> 00:38:55,600
Speaker 4: talk about you know, you know, true efficiency, it's nowhere

795
00:38:55,600 --> 00:38:57,719
Speaker 4: close to the leading edge. And well, I think you're

796
00:38:57,719 --> 00:39:00,440
Speaker 4: exactly right. If you want to really make a dent budget,

797
00:39:00,480 --> 00:39:02,520
Speaker 4: you have to talk entitlements, you have to talk debt

798
00:39:02,560 --> 00:39:04,759
Speaker 4: service and what that's going to look like. But if

799
00:39:04,800 --> 00:39:07,320
Speaker 4: you want to go to this next level, great tools

800
00:39:07,680 --> 00:39:11,360
Speaker 4: that prevents wasting of time, that auto meet auditing of records,

801
00:39:11,360 --> 00:39:13,520
Speaker 4: so you don't have a Department of Defense that's failing

802
00:39:13,560 --> 00:39:16,120
Speaker 4: seven audits in a row and can't track where spending

803
00:39:16,200 --> 00:39:18,399
Speaker 4: is going. But actually it's all digital, it's all tied

804
00:39:18,440 --> 00:39:20,319
Speaker 4: to it. I think you need to get serious about

805
00:39:20,360 --> 00:39:23,320
Speaker 4: allowing people to pick best tools for the problems.

806
00:39:23,680 --> 00:39:26,560
Speaker 2: Eric Lyman, thank you so much for coming on on lots.

807
00:39:26,560 --> 00:39:27,760
Speaker 2: I'm really glad we made this happen.

808
00:39:27,880 --> 00:39:28,879
Speaker 4: I really appreciate it.

809
00:39:41,239 --> 00:39:44,360
Speaker 2: Tracy, I really liked that episode. There's there's all kinds

810
00:39:44,400 --> 00:39:47,240
Speaker 2: of AI tools I need to now dive into because

811
00:39:47,280 --> 00:39:49,200
Speaker 2: like now I'm going to be the person who's subscribed

812
00:39:49,200 --> 00:39:51,920
Speaker 2: to nine different things and that's a little bit Yeah,

813
00:39:51,960 --> 00:39:54,359
Speaker 2: I know, like, what am I subscribed? I'm like, I'm

814
00:39:54,400 --> 00:39:57,480
Speaker 2: subscribe to at least like three probably more right now,

815
00:39:57,480 --> 00:39:59,279
Speaker 2: I probably you know now that I got to like

816
00:39:59,320 --> 00:40:01,840
Speaker 2: dive into these specialized coding tools. But no, that was

817
00:40:01,880 --> 00:40:02,359
Speaker 2: really fun.

818
00:40:02,880 --> 00:40:03,120
Speaker 1: Uh.

819
00:40:03,120 --> 00:40:06,280
Speaker 3: Have I told you the story before about my coding

820
00:40:06,320 --> 00:40:10,080
Speaker 3: class in high school seymore So. Actually it was just

821
00:40:10,160 --> 00:40:13,280
Speaker 3: a basic like it class, but as part of the class,

822
00:40:13,320 --> 00:40:17,360
Speaker 3: we had to program our own little application. And this

823
00:40:17,560 --> 00:40:20,759
Speaker 3: was like in the early two thousands, so it was

824
00:40:20,800 --> 00:40:24,759
Speaker 3: all very rudimentary. But our teacher commissioned us to do this,

825
00:40:25,000 --> 00:40:28,360
Speaker 3: and I built a fortune cookie program where you like

826
00:40:28,680 --> 00:40:31,080
Speaker 3: clicked a button, well the button looked like a cookie

827
00:40:31,160 --> 00:40:33,640
Speaker 3: and it gave you your fortune blah blah blah blah.

828
00:40:33,680 --> 00:40:36,080
Speaker 3: And at the end of the assignment, everyone turns in

829
00:40:36,640 --> 00:40:39,760
Speaker 3: their program to the teacher and he made everyone sign

830
00:40:39,840 --> 00:40:45,000
Speaker 3: a contract giving away the rights, the licensing rights to him,

831
00:40:45,480 --> 00:40:47,319
Speaker 3: and he said he did it to teach us all

832
00:40:47,360 --> 00:40:51,440
Speaker 3: a lesson about copyright and how your work is rarely

833
00:40:51,600 --> 00:40:54,319
Speaker 3: your own if you're working for a big corporation, which

834
00:40:54,360 --> 00:40:57,279
Speaker 3: is why I asked that copyright question because I still

835
00:40:57,280 --> 00:40:59,080
Speaker 3: think about that to this day. So it was a

836
00:40:59,080 --> 00:41:01,840
Speaker 3: good lesson that's extremely funny.

837
00:41:02,080 --> 00:41:03,520
Speaker 2: But no, there was a bunch in there that I

838
00:41:03,560 --> 00:41:05,680
Speaker 2: thought was interesting. So one, you know, the idea that

839
00:41:05,719 --> 00:41:07,799
Speaker 2: like you can just like track like what percentage of

840
00:41:07,840 --> 00:41:09,960
Speaker 2: people like pay for something one month then also the

841
00:41:10,000 --> 00:41:12,320
Speaker 2: next month. Also this idea, and like we got to

842
00:41:12,440 --> 00:41:14,719
Speaker 2: keep like coming back to it, the lack of lock

843
00:41:14,800 --> 00:41:17,879
Speaker 2: in for some of these models. Right, We're really used

844
00:41:17,920 --> 00:41:21,000
Speaker 2: to there just being one winner in search, one winner

845
00:41:21,040 --> 00:41:24,839
Speaker 2: in social networking, one winner in e commerce so forth,

846
00:41:24,920 --> 00:41:28,000
Speaker 2: one winner in photo sharing. So it's really interesting to

847
00:41:28,040 --> 00:41:30,759
Speaker 2: think about like all this money going to models like

848
00:41:30,840 --> 00:41:32,839
Speaker 2: A where it's like really easy to move from one

849
00:41:32,880 --> 00:41:35,680
Speaker 2: to the other, a simple project B where there's open

850
00:41:35,719 --> 00:41:38,840
Speaker 2: source competitors that maybe are just as good. That seems

851
00:41:38,840 --> 00:41:39,960
Speaker 2: like a pretty big deal right there.

852
00:41:40,040 --> 00:41:42,399
Speaker 3: Yeah, And I guess the big question is will there

853
00:41:42,520 --> 00:41:45,319
Speaker 3: eventually be some sort of winner that turns out to

854
00:41:45,360 --> 00:41:48,000
Speaker 3: be better at it than everyone else or is there

855
00:41:48,000 --> 00:41:51,480
Speaker 3: going to be space for that sort of specialized either

856
00:41:51,800 --> 00:41:55,080
Speaker 3: you know, specialized use case or the specialized like actual

857
00:41:55,160 --> 00:41:57,080
Speaker 3: interface for different jobs.

858
00:41:57,200 --> 00:41:58,240
Speaker 4: Yeah, I think that's interesting.

859
00:41:58,480 --> 00:42:02,200
Speaker 3: We talked a little bit about importance of interfaces totally. Anyway,

860
00:42:02,239 --> 00:42:04,200
Speaker 3: shall we leave it there, let's leave it there. This

861
00:42:04,239 --> 00:42:07,319
Speaker 3: has been another episode of the All Thoughts podcast. I'm

862
00:42:07,360 --> 00:42:09,920
Speaker 3: Tracy Alloway. You can follow me at Tracy Alloway.

863
00:42:10,080 --> 00:42:12,920
Speaker 2: And I'm Joe Wisenthal. You can follow me at the Stalwart.

864
00:42:13,160 --> 00:42:17,080
Speaker 2: Follow our guest Eric Glyman, He's at e Gleman. Follow

865
00:42:17,120 --> 00:42:20,560
Speaker 2: our producers Carman, Rodriguez at Kerman Arman, dash Ol Bennett

866
00:42:20,560 --> 00:42:23,879
Speaker 2: at Dashbot, and Keil Brooks at Kelbrooks. From our odd

867
00:42:23,920 --> 00:42:26,439
Speaker 2: loots content, go to Bloomberg dot com slash odd lots,

868
00:42:26,440 --> 00:42:28,759
Speaker 2: where have transcripts a blog in the newsletter, and you

869
00:42:28,800 --> 00:42:30,880
Speaker 2: can chat about all of these topics twenty four to

870
00:42:30,920 --> 00:42:35,160
Speaker 2: seven in our discord discord dot gg slash odd lots.

871
00:42:35,480 --> 00:42:37,920
Speaker 3: And if you enjoy All Thoughts, if you like it

872
00:42:37,960 --> 00:42:40,840
Speaker 3: when we talk about how much companies are actually spending

873
00:42:40,880 --> 00:42:44,040
Speaker 3: on AI, then please leave us a positive review on

874
00:42:44,080 --> 00:42:47,400
Speaker 3: your favorite podcast platform. And remember, if you are a

875
00:42:47,440 --> 00:42:50,640
Speaker 3: Bloomberg subscriber, you can listen to all of our episodes

876
00:42:50,760 --> 00:42:53,239
Speaker 3: absolutely ad free. All you need to do is find

877
00:42:53,280 --> 00:42:57,319
Speaker 3: the Bloomberg channel on Apple Podcasts and follow the instructions there.

878
00:42:57,640 --> 00:42:58,440
Speaker 3: Thanks for listening

879
00:43:03,840 --> 00:43:03,880
Speaker 4: In