1
00:00:00,100 --> 00:00:01,966
welcome to money is freedom

2
00:00:01,966 --> 00:00:02,866
a podcast

3
00:00:02,866 --> 00:00:06,066
exploring how finance and freedom connect in our lives

4
00:00:07,033 --> 00:00:08,833
created from thoughtful research

5
00:00:08,833 --> 00:00:11,300
and narrated by Notebook lmaai

6
00:00:11,400 --> 00:00:13,266
this series brings you clear

7
00:00:13,266 --> 00:00:15,966
meaningful insights into finance and beyond

8
00:00:16,066 --> 00:00:17,833
in Episode 45

9
00:00:17,833 --> 00:00:20,700
we're exploring what working means for AI apps

10
00:00:20,700 --> 00:00:22,833
based on insights from top VC

11
00:00:22,833 --> 00:00:25,866
firm and Dreisen Horowitz from Silicon Valley

12
00:00:26,233 --> 00:00:29,300
the generative AI era has ushered in a period where

13
00:00:29,300 --> 00:00:31,400
startups are growing faster than ever

14
00:00:31,400 --> 00:00:33,200
often with fewer resources

15
00:00:33,433 --> 00:00:35,800
we're seeing a significant shift in benchmarks

16
00:00:35,800 --> 00:00:37,466
for the average AI company

17
00:00:38,000 --> 00:00:40,600
what was once considered best in class growth

18
00:00:40,600 --> 00:00:43,233
like $1 million in ARR in the first year

19
00:00:43,233 --> 00:00:44,666
for enterprise startups

20
00:00:44,700 --> 00:00:46,900
is now considered on the lower end

21
00:00:47,466 --> 00:00:48,366
if you're interested

22
00:00:48,366 --> 00:00:50,900
in the cutting edge of tech and startup growth

23
00:00:51,033 --> 00:00:52,666
tune in to understand why

24
00:00:52,666 --> 00:00:54,566
there's never been a better time to build

25
00:00:54,566 --> 00:00:55,600
an application layer

26
00:00:55,600 --> 00:00:57,966
software company let's dive in

27
00:00:58,000 --> 00:00:59,633
alright let's dive in

28
00:00:59,633 --> 00:01:00,533
there's this uh

29
00:01:00,766 --> 00:01:02,800
this question that feels really central right now

30
00:01:02,800 --> 00:01:03,500
doesn't it

31
00:01:03,500 --> 00:01:06,666
what does working actually mean for a startup today

32
00:01:06,666 --> 00:01:07,800
especially you know

33
00:01:07,800 --> 00:01:11,033
with AI everywhere how much has the game really changed

34
00:01:11,033 --> 00:01:11,666
oh definitely

35
00:01:11,666 --> 00:01:13,100
it's more than just a feeling

36
00:01:13,166 --> 00:01:13,866
the um

37
00:01:13,866 --> 00:01:16,466
the traditional signals what we used to look for

38
00:01:16,466 --> 00:01:18,933
they seem to be shifting and pretty dramatically

39
00:01:19,100 --> 00:01:21,300
the old rules might just not cut it anymore

40
00:01:21,300 --> 00:01:23,166
yeah and we've actually got some

41
00:01:23,200 --> 00:01:25,033
really interesting data to look at here

42
00:01:25,033 --> 00:01:27,833
we're digging into this piece from Andreesen Horowitz

43
00:01:27,833 --> 00:01:28,100
you know a

44
00:01:28,100 --> 00:01:31,900
16 is it's based on what they've seen across like

45
00:01:31,900 --> 00:01:34,300
hundreds of companies over the last year and a half

46
00:01:34,300 --> 00:01:34,800
that's right

47
00:01:34,800 --> 00:01:38,066
it's called what working means in the era of AI apps

48
00:01:38,066 --> 00:01:39,966
uh huh so it's grounded in

49
00:01:39,966 --> 00:01:41,833
you know real world observation

50
00:01:41,833 --> 00:01:43,600
not just theory exactly

51
00:01:43,866 --> 00:01:45,800
so our mission for this deep dive anyway

52
00:01:45,800 --> 00:01:47,666
is to unpack what they found

53
00:01:47,666 --> 00:01:49,333
what are the new benchmarks

54
00:01:49,366 --> 00:01:51,166
especially for those app layer companies

55
00:01:51,166 --> 00:01:53,366
the ones building the stuff you actually use

56
00:01:53,400 --> 00:01:55,666
and what does it mean for you right

57
00:01:55,666 --> 00:01:56,666
whether you're a founder

58
00:01:56,666 --> 00:01:58,300
or just trying to keep up with all this

59
00:01:58,300 --> 00:02:01,166
this pace and it's pretty surprising stuff

60
00:02:01,166 --> 00:02:03,000
it really is prepare yourself

61
00:02:03,166 --> 00:02:06,066
because the numbers aren't just like a little faster

62
00:02:06,066 --> 00:02:09,200
they suggest growth is accelerating significantly

63
00:02:09,266 --> 00:02:11,100
and maybe not where you'd expect okay

64
00:02:11,100 --> 00:02:12,400
so maybe let's set the scene first

65
00:02:12,400 --> 00:02:15,600
like before this whole AI wave really hit

66
00:02:15,800 --> 00:02:17,300
there was a pretty common benchmark

67
00:02:17,300 --> 00:02:19,600
wasn't there for what you know

68
00:02:19,600 --> 00:02:21,866
amazing looked like for an enterprise startup

69
00:02:21,866 --> 00:02:25,133
in its first year first 12 months after launch

70
00:02:25,266 --> 00:02:26,933
or starting to charge money

71
00:02:27,066 --> 00:02:28,033
yeah absolutely

72
00:02:28,033 --> 00:02:30,233
that magic number was often around

73
00:02:30,233 --> 00:02:32,466
$1 million in annual recurring revenue

74
00:02:32,500 --> 00:02:35,100
$1 Mar if you hit that in year one

75
00:02:35,100 --> 00:02:38,066
you were basically considered top tier best in class

76
00:02:38,066 --> 00:02:40,466
it showed you really had something in the b to B world

77
00:02:40,466 --> 00:02:41,966
right a strong signal

78
00:02:41,966 --> 00:02:43,800
yeah but for consumer companies

79
00:02:44,600 --> 00:02:45,600
totally different story

80
00:02:45,600 --> 00:02:47,100
revenue was like way down the road

81
00:02:47,100 --> 00:02:48,266
maybe years later the whole

82
00:02:48,266 --> 00:02:50,266
focus was just getting users

83
00:02:50,300 --> 00:02:52,400
millions of them tens of millions sometimes

84
00:02:52,500 --> 00:02:53,633
then you figure out how to make money

85
00:02:53,633 --> 00:02:55,800
usually ads or something so that was the setup

86
00:02:55,800 --> 00:02:57,866
$1 AR for top enterprise users

87
00:02:57,866 --> 00:02:59,500
first for consumer okay

88
00:02:59,500 --> 00:03:02,900
so now we've got this a 16 is data last 18 months

89
00:03:02,900 --> 00:03:04,666
hundreds of companies hmm

90
00:03:04,666 --> 00:03:05,833
what did they actually find

91
00:03:05,833 --> 00:03:06,966
what's changed well

92
00:03:06,966 --> 00:03:08,600
the picture is uh yeah

93
00:03:08,600 --> 00:03:11,100
very different looking at the companies they know well

94
00:03:11,100 --> 00:03:13,233
the numbers have just jumped right

95
00:03:13,233 --> 00:03:15,166
and this is where it gets really compelling

96
00:03:15,166 --> 00:03:17,833
the median enterprise company in their data

97
00:03:17,833 --> 00:03:21,166
it's not hitting 1 million dollar ARR in year one

98
00:03:21,166 --> 00:03:24,500
it's reaching over 2 million dollar ARR

99
00:03:24,833 --> 00:03:28,266
and get this they're typically raising their Series a

100
00:03:28,300 --> 00:03:30,966
just nine months after starting to monetize

101
00:03:30,966 --> 00:03:32,666
just pause on that the median

102
00:03:32,666 --> 00:03:35,500
so like the typical company in their sample

103
00:03:35,600 --> 00:03:36,966
is doing more than double

104
00:03:36,966 --> 00:03:38,366
what we used to call best in class

105
00:03:38,366 --> 00:03:40,200
just a few years back that's well

106
00:03:40,200 --> 00:03:41,500
that's a huge acceleration right

107
00:03:41,500 --> 00:03:43,766
and then the consumer side is even Wilder

108
00:03:43,933 --> 00:03:46,066
the median consumer AI companies in their data

109
00:03:46,066 --> 00:03:48,033
they're actually doing better on early revenue

110
00:03:48,033 --> 00:03:50,600
they're hitting something like $4.2 million in AR

111
00:03:50,600 --> 00:03:53,433
in the first year and raising Series a even faster

112
00:03:53,433 --> 00:03:55,900
maybe eight months in so that journey

113
00:03:55,900 --> 00:03:58,233
that zero to $1 million AR

114
00:03:58,233 --> 00:03:59,900
which used to be the gold standard

115
00:04:00,166 --> 00:04:01,400
it's now kind of well

116
00:04:01,400 --> 00:04:03,066
it's at the lower end for these AI

117
00:04:03,066 --> 00:04:05,700
companies that are getting traction in year one okay

118
00:04:05,700 --> 00:04:09,200
okay the speed is just undeniable

119
00:04:09,200 --> 00:04:11,200
faster money faster funding rounds

120
00:04:11,200 --> 00:04:15,166
but why what is it about AI specifically

121
00:04:15,400 --> 00:04:17,766
that lets companies accelerate like this

122
00:04:17,766 --> 00:04:19,166
yeah that's the key question

123
00:04:19,166 --> 00:04:21,333
is it probably a few things working together

124
00:04:21,566 --> 00:04:23,600
um first

125
00:04:23,866 --> 00:04:24,300
AI

126
00:04:24,300 --> 00:04:26,900
can let you iterate on the product incredibly quickly

127
00:04:27,200 --> 00:04:28,900
you know improving the core function

128
00:04:28,900 --> 00:04:30,400
might just mean training a new model

129
00:04:30,400 --> 00:04:31,466
or plugging in a better one

130
00:04:31,466 --> 00:04:33,100
that can be way faster than rewriting

131
00:04:33,100 --> 00:04:34,366
big chunks of traditional code

132
00:04:34,366 --> 00:04:35,666
right faster building

133
00:04:35,733 --> 00:04:37,433
and second these AI

134
00:04:37,433 --> 00:04:39,700
models can often deliver value almost immediately

135
00:04:39,700 --> 00:04:41,233
like right out of the box

136
00:04:41,233 --> 00:04:43,533
that initial hook for users or businesses

137
00:04:43,600 --> 00:04:45,800
it can happen faster less setup

138
00:04:45,800 --> 00:04:48,500
maybe less configuration than older software needed

139
00:04:48,500 --> 00:04:50,233
okay faster value delivery too

140
00:04:50,233 --> 00:04:51,266
and then there's just you know

141
00:04:51,266 --> 00:04:52,600
massive market pull right now

142
00:04:52,600 --> 00:04:54,866
people and companies are actively looking for AI tools

143
00:04:54,866 --> 00:04:57,300
the demand is just huge so faster building

144
00:04:57,300 --> 00:05:00,033
faster value hot market makes sense

145
00:05:00,033 --> 00:05:01,866
and the article hints that this speed

146
00:05:01,866 --> 00:05:04,966
this velocity it's becoming a competitive edge itself

147
00:05:04,966 --> 00:05:07,666
like a mote how does speed work as a mote here

148
00:05:07,666 --> 00:05:10,166
well it's more than just being first right

149
00:05:10,200 --> 00:05:12,200
speed lets you iterate on your models

150
00:05:12,200 --> 00:05:14,466
your features faster than others

151
00:05:14,633 --> 00:05:16,966
and if your AI learns from user data

152
00:05:17,500 --> 00:05:17,966
well

153
00:05:17,966 --> 00:05:22,066
moving fast means you potentially get more data sooner

154
00:05:22,200 --> 00:05:24,266
and that creates this cycle right

155
00:05:24,300 --> 00:05:25,566
better data better models

156
00:05:25,566 --> 00:05:26,966
better product more users

157
00:05:26,966 --> 00:05:28,266
more data ah

158
00:05:28,266 --> 00:05:29,100
the flywheel effect

159
00:05:29,100 --> 00:05:31,000
so the gap widens because you're learning

160
00:05:31,000 --> 00:05:32,333
faster exactly

161
00:05:32,600 --> 00:05:34,166
it can create a lead that slower

162
00:05:34,166 --> 00:05:36,033
companies just find really hard to close okay

163
00:05:36,033 --> 00:05:37,666
that makes sense speed to launch

164
00:05:37,666 --> 00:05:39,166
but also speed to improve

165
00:05:39,166 --> 00:05:42,200
so faster revenue faster rounds speed as a mode

166
00:05:42,433 --> 00:05:44,033
that sounds like the early game plan

167
00:05:44,033 --> 00:05:46,766
but um the article also throws in a pretty

168
00:05:46,766 --> 00:05:47,833
important dose of caution

169
00:05:47,833 --> 00:05:49,900
doesn't it it's not all about that topline growth

170
00:05:49,900 --> 00:05:51,000
especially later on oh

171
00:05:51,000 --> 00:05:53,400
absolutely crucial point that initial speed

172
00:05:53,400 --> 00:05:55,066
that velocity that gets you noticed

173
00:05:55,066 --> 00:05:56,766
gets you that fast series a

174
00:05:56,800 --> 00:05:58,700
but when investors look at later rounds

175
00:05:58,700 --> 00:06:00,400
series B series C

176
00:06:00,666 --> 00:06:02,133
they're gonna dig into the fundamentals

177
00:06:02,166 --> 00:06:04,700
the traditional software metrics okay

178
00:06:04,700 --> 00:06:06,800
so things like usage

179
00:06:06,900 --> 00:06:08,433
how much are people actually using it

180
00:06:08,433 --> 00:06:12,900
engagement and uh critically retention

181
00:06:13,233 --> 00:06:14,233
are they sticking around

182
00:06:14,233 --> 00:06:16,000
you can't just buy that initial revenue spike

183
00:06:16,000 --> 00:06:18,500
and expect it to last precisely yeah

184
00:06:18,500 --> 00:06:20,500
if that early surge is just

185
00:06:20,500 --> 00:06:22,566
you know burning cash unsustainably

186
00:06:22,566 --> 00:06:24,433
or if users aren't really engaged

187
00:06:24,433 --> 00:06:25,966
or sticking around long term

188
00:06:26,066 --> 00:06:27,800
it won't pass the test later

189
00:06:27,900 --> 00:06:29,766
the speed proves you can get started

190
00:06:29,833 --> 00:06:31,433
but retention proves you built something

191
00:06:31,433 --> 00:06:33,100
people actually value over time

192
00:06:33,100 --> 00:06:34,266
something sticky hmm

193
00:06:34,600 --> 00:06:37,166
and how does AI itself affect those metrics

194
00:06:37,166 --> 00:06:38,200
does having an AI

195
00:06:38,200 --> 00:06:41,666
feature automatically make retention better or worse

196
00:06:41,666 --> 00:06:43,500
and how do you even measure engagement

197
00:06:43,500 --> 00:06:45,466
with some of these generative AI tools

198
00:06:45,566 --> 00:06:47,966
you might use them intensely for a short burst

199
00:06:47,966 --> 00:06:49,633
then not for a while yeah

200
00:06:49,633 --> 00:06:50,366
that's definitely something

201
00:06:50,366 --> 00:06:51,466
the whole ecosystem is still

202
00:06:51,466 --> 00:06:54,833
figuring out but what the data seems to hint at is

203
00:06:54,833 --> 00:06:56,666
for the users who do convert to paying

204
00:06:56,666 --> 00:06:58,166
for generative AI apps

205
00:06:58,433 --> 00:07:01,000
their retention often looks pretty similar to older

206
00:07:01,000 --> 00:07:02,866
pre AI consumer apps oh interesting

207
00:07:02,866 --> 00:07:04,200
so once they pay they stick

208
00:07:04,200 --> 00:07:05,466
it seems like it yeah

209
00:07:05,466 --> 00:07:06,100
the challenge

210
00:07:06,100 --> 00:07:08,133
might be getting them to pay in the first place

211
00:07:08,433 --> 00:07:10,733
that free to paid conversion could be trickier

212
00:07:10,833 --> 00:07:13,300
but if they do convert they seem quite sticky

213
00:07:13,766 --> 00:07:14,833
which you know

214
00:07:14,833 --> 00:07:17,700
pushes companies to really focus on proving that value

215
00:07:17,833 --> 00:07:20,266
clearly and consistently to the paying users

216
00:07:20,266 --> 00:07:22,066
okay let's circle back to that point

217
00:07:22,066 --> 00:07:23,233
that really jumped out earlier

218
00:07:23,233 --> 00:07:26,700
the consumer AI companies making serious money

219
00:07:26,700 --> 00:07:30,600
really fast that just completely flips the old script

220
00:07:30,600 --> 00:07:32,966
doesn't it build users for years

221
00:07:32,966 --> 00:07:34,966
then worry about money totally clips it

222
00:07:34,966 --> 00:07:35,633
yeah yeah

223
00:07:35,633 --> 00:07:38,633
the data shows the median revenue for B to C AI

224
00:07:38,633 --> 00:07:40,166
is actually higher than B to B

225
00:07:40,166 --> 00:07:42,133
early on based on their sample

226
00:07:42,233 --> 00:07:44,233
and the way they grow can look different too

227
00:07:44,233 --> 00:07:45,500
how so different shape

228
00:07:45,500 --> 00:07:46,266
well think about it

229
00:07:46,266 --> 00:07:48,900
a lot of these consumer AI companies raise big money

230
00:07:48,900 --> 00:07:51,666
specifically to train their own models right

231
00:07:51,666 --> 00:07:53,300
their own large AI models

232
00:07:53,433 --> 00:07:56,233
and what they often see is a huge jump in revenue

233
00:07:56,233 --> 00:07:57,766
right after they release a new

234
00:07:57,766 --> 00:07:58,900
better model ah

235
00:07:58,900 --> 00:08:00,466
so it's not a smooth curve

236
00:08:00,466 --> 00:08:04,033
more like steps big jump when the model gets better

237
00:08:04,033 --> 00:08:04,666
kind of yeah

238
00:08:04,666 --> 00:08:08,033
more like step function growth tied to those big R&D

239
00:08:08,033 --> 00:08:10,633
moments you get a spike maybe things level off a bit

240
00:08:10,633 --> 00:08:13,066
then another spike with the next big model improvement

241
00:08:13,366 --> 00:08:15,833
and like we said maybe converting free users is tough

242
00:08:15,833 --> 00:08:18,900
but those paying users seem to stick around

243
00:08:19,033 --> 00:08:21,266
okay so the big picture take

244
00:08:21,433 --> 00:08:21,800
away from this

245
00:08:21,800 --> 00:08:25,066
a Sixteen's data seems pretty loud and clear

246
00:08:25,766 --> 00:08:28,133
startups are working faster

247
00:08:28,133 --> 00:08:31,133
much faster businesses consumers

248
00:08:31,600 --> 00:08:34,633
they seem genuinely willing to pay for new AI

249
00:08:34,633 --> 00:08:36,333
powered stuff right from the start

250
00:08:36,333 --> 00:08:38,033
yeah and based on this

251
00:08:38,033 --> 00:08:39,633
it feels like a really dynamic time

252
00:08:39,633 --> 00:08:41,333
to be building an application layer

253
00:08:41,433 --> 00:08:42,466
software company

254
00:08:42,466 --> 00:08:45,366
the potential for just really rapid traction

255
00:08:45,366 --> 00:08:47,766
significant early revenue seems

256
00:08:47,833 --> 00:08:49,500
kind of unprecedented which is well

257
00:08:49,500 --> 00:08:51,333
it's exciting if you're a founder with a great idea

258
00:08:51,333 --> 00:08:53,133
but also probably pretty daunting

259
00:08:53,233 --> 00:08:55,500
because the bar for showing that early traction

260
00:08:55,500 --> 00:08:57,133
it just moved way way higher

261
00:08:57,133 --> 00:08:57,166
yeah

262
00:08:57,166 --> 00:08:59,466
and the time you have to prove it before that Series A

263
00:08:59,466 --> 00:09:01,866
it seems shorter speed is definitely table stakes now

264
00:09:01,866 --> 00:09:03,266
yeah but like we talked about

265
00:09:03,266 --> 00:09:05,100
it's not the only thing not for the long haul

266
00:09:05,100 --> 00:09:07,466
you absolutely need that underlying product strength

267
00:09:07,466 --> 00:09:10,133
that stickiness to back up the early speed right

268
00:09:10,133 --> 00:09:11,666
so just to quickly recap

269
00:09:11,666 --> 00:09:13,900
what we pulled from this deep dive into the A16

270
00:09:13,900 --> 00:09:14,566
z's numbers

271
00:09:14,566 --> 00:09:18,433
that old $1 ARR year one benchmark kind of blown away

272
00:09:18,433 --> 00:09:21,433
median AI companies are hitting over $2 for enterprise

273
00:09:21,566 --> 00:09:24,766
and an amazing four dollars and two cents for consumer

274
00:09:24,900 --> 00:09:28,566
fundraising like Series a is happening way faster 8

275
00:09:28,566 --> 00:09:29,633
9 months maybe yeah

276
00:09:29,633 --> 00:09:32,433
and consumer AI is a real revenue story now

277
00:09:32,433 --> 00:09:35,066
right from the beginning changes that whole sector

278
00:09:35,066 --> 00:09:35,666
but crucially

279
00:09:35,666 --> 00:09:38,433
while that speed and early cash get you in the game

280
00:09:38,433 --> 00:09:39,833
maybe get you funded initially

281
00:09:39,900 --> 00:09:41,266
the traditional metrics usage

282
00:09:41,266 --> 00:09:43,100
engagement and especially retention

283
00:09:43,100 --> 00:09:45,666
they are absolutely vital for later rounds

284
00:09:45,700 --> 00:09:47,666
for actually building something that lasts

285
00:09:47,666 --> 00:09:50,100
speed gets you the meeting quality keeps you going

286
00:09:50,100 --> 00:09:51,800
it really underlines it doesn't it

287
00:09:51,833 --> 00:09:53,733
the AI era isn't just faster

288
00:09:53,733 --> 00:09:55,833
it's actually redefining what early success

289
00:09:55,833 --> 00:09:57,866
even looks like so thinking about that

290
00:09:57,866 --> 00:10:00,933
this incredible speed these new high benchmarks

291
00:10:00,933 --> 00:10:03,466
this potential gap between the fast and the slow

292
00:10:03,500 --> 00:10:04,533
it makes you wonder

293
00:10:04,533 --> 00:10:06,566
what does all this velocity really mean

294
00:10:06,566 --> 00:10:07,733
for the whole startup world

295
00:10:07,733 --> 00:10:09,566
beyond just getting funded faster

296
00:10:09,666 --> 00:10:11,033
how does speed as a mode

297
00:10:11,033 --> 00:10:12,533
really change the fundamental game

298
00:10:12,533 --> 00:10:14,033
not just building but competing

299
00:10:14,033 --> 00:10:16,666
surviving in this AI age something to think about