Fireside chat with Tomasz Tunguz:AI Disruption of GTM Motions

1
00:00:14.440 –> 00:00:25.739
Julia Nimchinski: Welcome back to the AI led Growth Summit Day 3. Today we dive deeper into AI’s destructive role in the future of Sas. B. 2 B marketing partner led community led

2
00:00:25.850 –> 00:00:35.619
Julia Nimchinski: and many more emerging Gtm strategies. Make sure you’re following our conversation in the Hsc. Slack, and it’s the events channel

3
00:00:35.780 –> 00:00:39.329
Julia Nimchinski: and also check out all the sponsors in the right side of the screen.

4
00:00:39.810 –> 00:00:47.529
Julia Nimchinski: Let’s kick things off. I think our community was the most excited about this event, so the whole conference was worth it.

5
00:00:47.720 –> 00:00:52.189
Julia Nimchinski: Welcome to the show. Doug Lenderson to Austin Goose. You need no introduction.

6
00:00:53.160 –> 00:00:54.240
Julia Nimchinski: How are you doing.

7
00:00:54.850 –> 00:00:58.579
Doug Landis: Rad, I’m great excited for this conversation.

8
00:00:58.750 –> 00:00:59.960
Tomasz Tunguz: Thrilled to be here.

9
00:01:00.440 –> 00:01:01.759
Doug Landis: Lot to unpack.

10
00:01:02.800 –> 00:01:03.590
Julia Nimchinski: Let’s do it.

11
00:01:03.860 –> 00:01:18.539
Doug Landis: For sure awesome, since, since you know, Tomas, you’re my guest here, I’ll just lead, with a little bit of background, information that some people may not necessarily know about you. I didn’t actually know until I started doing a little digging. Thank you. Chat Gpt.

12
00:01:18.660 –> 00:01:40.079
Doug Landis: Early twenties. You’re in DC. You read Jack Kerouac’s book, and that gives you the inspiration that you need to go to California. Was, that is that a little bit part of it? Google was clearly kind of on your on your roadmap here, because you, as an Ml. Expert or Ml. Engineer, that was the place where you wanted to land. But talk. How did you? Who gave you the Jack Kerouac book? That’s what.

13
00:01:40.080 –> 00:01:55.089
Tomasz Tunguz: Yeah. Great question. Yeah. I don’t remember who gave it to me, but I remember reading on the road, and it blew my mind that wrote it in a single sitting, and you can actually buy the book on a single scroll, which is like a single piece of paper now, and I.

14
00:01:55.090 –> 00:01:55.940
Doug Landis: That’s cool.

15
00:01:55.940 –> 00:02:01.319
Tomasz Tunguz: Yeah, a friend of mine I had a cousin who had an internship in San Francisco, and I came to visit him, and

16
00:02:01.721 –> 00:02:06.500
Tomasz Tunguz: anyway, fell in love with California. I came out to Google. And there’s this one moment where

17
00:02:06.770 –> 00:02:19.774
Tomasz Tunguz: interviewing for entry level customer support job, and we went into this cafe was called no name cafe at the time, and there was this bowl of raspberries there, and the recruiter said I could have as many as I wanted, and I knew I wanted to work there.

18
00:02:20.030 –> 00:02:23.899
Doug Landis: That’s amazing. You know, it’s funny. I was actually at Google at the same time, I.

19
00:02:23.900 –> 00:02:24.910
Tomasz Tunguz: No way. Is that true?

20
00:02:25.240 –> 00:02:28.659
Doug Landis: Absolutely 100%. The No name cafe was the best

21
00:02:28.810 –> 00:02:42.509
Doug Landis: they did. They did the greatest thing, like they with the whole, with the food and the food and the gym, and the whole all the experiences on campus you never wanted to leave. I’d get there at like 6 in the morning, do my workouts, have breakfast and and work until the evening, and then play soccer.

22
00:02:43.090 –> 00:02:44.130
Doug Landis: It was amazing.

23
00:02:44.130 –> 00:02:46.309
Tomasz Tunguz: It was an amazing amazing time. Amazing.

24
00:02:46.310 –> 00:02:46.640
Doug Landis: Yeah.

25
00:02:46.640 –> 00:02:47.050
Tomasz Tunguz: Gonna be there.

26
00:02:47.050 –> 00:03:14.060
Doug Landis: Genius, genius. You know what’s interesting, you know, prepping for this conversation one of the things that I’ve look. I followed you for a long, long time as somebody who’s also been in Vc. And in the startup world. For a long time you have been on the precipice in particular, of like this. AI wave this AI revolution that we’re in the middle of right now. And in fact, at 1 point in time you even said that AI is poised to disrupt these entrenched workflows that these major companies, these big incumbents like Salesforce, Zendesk, Servicenow, workday, etc.

27
00:03:14.060 –> 00:03:20.560
Doug Landis: are all built upon, and and I guess I want to kind of start there is like, how do you think about how

28
00:03:20.560 –> 00:03:45.229
Doug Landis: you know you’ve got this massive influx of these new AI 1st startups if you will, competing against these massive incumbents, who own the distribution channels and who seem to be bolting AI on top of their existing technology, talk to me about how you think about the difference between maybe an AI 1st startup versus a big incumbent, and how you think they could potentially disrupt these big players.

29
00:03:45.290 –> 00:03:53.391
Tomasz Tunguz: Great question. So I think the mobile app store. When when jobs launched the the iphone, we had a bunch of like

30
00:03:54.170 –> 00:04:06.789
Tomasz Tunguz: apps that were basically websites in boxes. And we called those mobile apps. And then you had this like Uber moment, where for the 1st time use location, it completely changed everything because location was really built in. I think we’re kind of in that.

31
00:04:07.380 –> 00:04:22.109
Tomasz Tunguz: We are in that phase of mobile apps that are websites right now, and this is what all the bolt ons are, and for some period of time, and it will depend on different industries whether it’s an accounting or legal or engineering. Those will

32
00:04:22.600 –> 00:04:27.189
Tomasz Tunguz: work, and then at some point it will all break, and I think the point it all breaks is when

33
00:04:27.751 –> 00:04:35.710
Tomasz Tunguz: the workflows themselves change right? You’ve had, like 2020 year, 25 years of A and Sdr. Motion that has been

34
00:04:36.060 –> 00:04:37.450
Tomasz Tunguz: perfected

35
00:04:37.946 –> 00:05:06.410
Tomasz Tunguz: over time and honed and whittled. And now, all of a sudden, an A can manage. Many more leads. The. The. A lot of the work that a sales development rep will do is now fully, is fully automated, like the research you just talked about understanding individual changes, listening to Youtube videos or or podcasts or whatever it is. And so as those workflows change. Eventually the the systems that grew around those workflows no longer apply, and people will start to feel a significant amount of friction.

36
00:05:07.190 –> 00:05:27.250
Doug Landis: Yeah, the change management. I feel like involved in adopting AI technology in an existing organization. It feels a little daunting to be honest. The thing that’s really fascinating is this notion of agentic AI, and just for everybody who’s here, you probably already know what that is. It’s AI that can adapt and make

37
00:05:27.250 –> 00:05:48.810
Doug Landis: decisions independently, very dynamically. I’m curious. How do you think it’s going to dethrone Sas? And do you think it’s going to dethrone sas completely? Are we just moving into an AI dominated world, or will Sas still have a place? I’m curious. Your thoughts on. We’ll call it the next 5 years, if you will, between AI agentic, AI and Sas.

38
00:05:48.810 –> 00:06:00.290
Tomasz Tunguz: Yeah, I think they’re the same thing. I think they basically they fuse you can’t. You won’t be a software company without AI. Everyone will expect it. I think the the point you made Doug is is the

39
00:06:00.620 –> 00:06:13.480
Tomasz Tunguz: is the opportunity and the challenge in the puzzle, which is, there’s a lot of change management. The workflows themselves like I I mean, the workflows are changing so fast that it’s hard for anyone to keep up.

40
00:06:13.480 –> 00:06:14.610
Doug Landis: It’s crazy.

41
00:06:14.610 –> 00:06:24.780
Tomasz Tunguz: Yeah, I’m just signing a Powerpoint presentation. I’ve been trained for whatever. If I grew up in the Mckinsey world I was trained in a very particular way of making a presentation. And now.

42
00:06:24.780 –> 00:06:25.610
Doug Landis: Not anymore.

43
00:06:25.610 –> 00:06:32.910
Tomasz Tunguz: Not anymore. I go to Cloud, and I dictate a whole bunch of stuff and then say, create 20 slides. And here’s a format, and there’s the images that I want right. And

44
00:06:33.030 –> 00:06:37.180
Tomasz Tunguz: so what we see with a lot of these agent companies is the

45
00:06:37.490 –> 00:06:47.629
Tomasz Tunguz: customer success. People are consultants. They’re recruited from Bain and Mckinsey, because it’s all about that change management and helping people understand how you use these tools in a novel way.

46
00:06:48.180 –> 00:07:04.739
Doug Landis: Yeah, yeah, look to be honest, I think if you just think about historically, over the last 100 years, every time there’s been new technological advancement like whether it was the telephone or the iphone, or whatever the Palm pilot right every time there’s been new technology introduced. There’s the one, the one thing that hasn’t

47
00:07:04.840 –> 00:07:26.179
Doug Landis: that they’re the one constant, I guess, is change, right change, management and helping to drive change across organizations both internally and externally, is still the one hurdle that we all have to navigate. I think we can do it faster and easier now with technology. I’m curious. Are there areas of a business or industry, or sector, where you think kind of agentic. AI is not really a great fit.

48
00:07:27.980 –> 00:07:35.149
Tomasz Tunguz: I think. Excuse me, the places where humans still need to build trust with each other. It doesn’t work very well.

49
00:07:35.930 –> 00:07:45.210
Tomasz Tunguz: so I don’t. I think it will be a long time before there’s a fully autonomous account executive, especially for considered purchases.

50
00:07:45.390 –> 00:07:46.070
Doug Landis: Yeah.

51
00:07:46.510 –> 00:07:50.289
Tomasz Tunguz: That’s not to say that. I mean, I don’t know. This is all heuristics. But, like

52
00:07:50.600 –> 00:08:15.695
Tomasz Tunguz: 2 decades ago, I would say, most software buyers were maybe 15% of the way through the journey in buying buying a software before they reached an Ae. And now I think they’re probably like 70% of the way educated themselves. But they’re still when someone buys a an Erp system almost betting their career on that Erp. And they, it’s very human to be able to want the call to Doug.

53
00:08:16.030 –> 00:08:20.260
Tomasz Tunguz: yeah, meet you, man like this is busted. My boss is on my case, and.

54
00:08:20.260 –> 00:08:20.640
Doug Landis: Yeah.

55
00:08:20.660 –> 00:08:28.219
Tomasz Tunguz: So I think that those areas where there’s a lot of human trust that’s necessary. So at the Ae level.

56
00:08:29.140 –> 00:08:29.710
Tomasz Tunguz: beggar.

57
00:08:29.710 –> 00:08:30.260
Doug Landis: Agree with that.

58
00:08:30.470 –> 00:08:32.110
Tomasz Tunguz: You agree with that? Yeah, but otherwise.

59
00:08:32.110 –> 00:08:32.970
Doug Landis: Totally agree with that.

60
00:08:32.970 –> 00:08:34.759
Tomasz Tunguz: Yeah. Go. Please, keep. Continue.

61
00:08:34.760 –> 00:08:59.760
Doug Landis: Well, no, I think well, here’s I mean, just think about it this way, even if they’re 70 80% of the way through the buying process. I’m like, look you make buying decisions. We all make buying decisions on a pretty regular basis, even though I’m further along in the process. There’s still that 2030% left that I need to have a meaningful conversation with somebody, and it needs to feel different with all of this information and intelligence at your fingertips. I fundamentally believe that intelligence is becoming ubiquitous. It’s like it’s there, it’s available.

62
00:08:59.760 –> 00:09:05.239
Doug Landis: The answers are everywhere. So now my expectation is actually different for you as a seller.

63
00:09:05.240 –> 00:09:10.240
Doug Landis: My expectation is like, you’re gonna come at me with a very, very thoughtful

64
00:09:10.240 –> 00:09:21.140
Doug Landis: question or conversation, rather than it’s just a list of kind of what I would call your your typical, you know, discovery questions like, No, don’t. I don’t want to have that conversation anymore.

65
00:09:21.150 –> 00:09:25.889
Doug Landis: It just feels like that’s shifted quite a quite a bit, or, as it’s continuing to shift.

66
00:09:26.960 –> 00:09:36.483
Tomasz Tunguz: Yeah, awesome. Yeah, this is a thanks, Julia. This is a chart. So there was a an engineer who published this. Yeah, he published.

67
00:09:36.990 –> 00:09:52.420
Tomasz Tunguz: the blue curve here, or the blue gray curve, and his gosh! I forget his name. But let me pull it up while we’re talking about this. But the idea was he had this thought experiment, which is at which levels of seniority does AI impact the most

68
00:09:52.670 –> 00:09:54.350
Tomasz Tunguz: and the least. And so.

69
00:09:54.350 –> 00:09:54.800
Doug Landis: Oh!

70
00:09:54.800 –> 00:10:07.510
Tomasz Tunguz: He? He started with this blue curve. And the idea is like, Okay, whatever if you’re and he did it for software engineering. So let’s take the curve. If you’re an entry level software engineer, your ability as entry level software engineer to ship a lot of code is is much faster

71
00:10:07.750 –> 00:10:36.370
Tomasz Tunguz: and you will make fewer mistakes, and you’ll get the benefit of the median right. If we could assume, like a large language model is the average of all the code in the world, and you’re starting as an entry level software engineer. You’re probably below average. But the Llm. Gets you. The AI gets you to a place where you’re in the middle. And so that’s a big benefit. Now, if you’re in middle management, let’s say, and you’re very familiar with the system. You’ve been working in a company for 5 years. His contention was well, you’re probably equally fast, both writing code with AI versus not with AI, because you understand

72
00:10:36.590 –> 00:10:37.770
Tomasz Tunguz: the code base.

73
00:10:37.970 –> 00:10:40.910
Tomasz Tunguz: And so maybe there’s like a modest benefit there

74
00:10:41.470 –> 00:11:01.624
Tomasz Tunguz: and then for the Vp of Edge or the chief architects, there’s a huge advantage again, which is, you can go through different architectures. You can understand. You can go back and forth and test different things and experiment really quickly. And so that’s that blue curve where you have positive for juniors and positive for the most senior and and middle

75
00:11:03.203 –> 00:11:06.772
Doug Landis: Middle impact management is at risk is what you’re saying.

76
00:11:07.090 –> 00:11:08.570
Tomasz Tunguz: Yeah. So there’s a broader.

77
00:11:08.570 –> 00:11:08.960
Doug Landis: True.

78
00:11:08.960 –> 00:11:26.249
Tomasz Tunguz: No, no, it’s true. It is true. I think. I was having this conversation with 2 different teams yesterday, which is, if you imagine a company very large company, you have the senior leadership. And then you have the doers. And in between, there’s this huge communication organization that’s basically multiplexing across everybody. Yeah.

79
00:11:26.640 –> 00:11:40.289
Tomasz Tunguz: And so what happens in a world where all the unstructured information of all of our conversations and our actions on at work are then assimilated right? So maybe we’ll have more doers. And and I think this raises a broader point, which is.

80
00:11:40.490 –> 00:12:00.418
Tomasz Tunguz: you know, talking about like the go to market of software companies. We have all these ratios like, here’s an A, here’s a Aesdr ratio, right? And we 15 years. We’ve been like, okay, here’s the science. Here’s the machine, and here’s the arr per csm, all of a sudden that’s completely changed. How many engineers per salesperson. And so that, I think, is something where,

81
00:12:00.820 –> 00:12:03.720
Tomasz Tunguz: the architecture of organizations will fundamentally change.

82
00:12:03.720 –> 00:12:14.700
Doug Landis: Yeah, yeah. Well, I mean, I think you’ve even said it that you think there’s a billion dollar software company that can be built with one person. I’m I’m working on a company right now. We’re about to launch. I believe we can get to a hundred 1 million in revenue with a hundred people.

83
00:12:14.700 –> 00:12:38.679
Doug Landis: just the the way in which you can use technology to drive efficiency today, like. And I want to talk about this part of this part of what’s happening right now is, there’s going to be a massive shift, as we think about hiring in an organization right? Because AI is going to replace some human tasks. So we’re not going to need as many Sdrs. Maybe we don’t need as many Ams because the technology is actually onboarding people so much better that we don’t need as many Ams right, or Csms as an example.

84
00:12:38.680 –> 00:12:42.489
Doug Landis: Maybe, if it’s even augmenting new roles. So how should companies rethink

85
00:12:42.490 –> 00:12:49.950
Doug Landis: hiring versus upskilling and kind of talent. Acquisition in this AI 1st world, especially on the go to market side.

86
00:12:50.270 –> 00:13:00.399
Tomasz Tunguz: Yeah, great question. So I was chatting with one of our founders. I was asking him this question, and he was basically saying, you have to. You have to interview for prompting skills.

87
00:13:01.380 –> 00:13:03.980
Tomasz Tunguz: which is totally different. Right? That’s like.

88
00:13:03.980 –> 00:13:04.889
Doug Landis: No new.

89
00:13:04.890 –> 00:13:07.889
Tomasz Tunguz: It’s so new. It’s like, okay, here is a task.

90
00:13:08.220 –> 00:13:12.269
Tomasz Tunguz: I mean, here is, yeah, I mean, yeah, here is a task

91
00:13:13.074 –> 00:13:19.145
Tomasz Tunguz: using the AI to solve it. And what you really want to see, particularly in

92
00:13:19.840 –> 00:13:24.499
Tomasz Tunguz: the more junior roles is, you want to see an ability to create a huge amount of leverage from the AI.

93
00:13:24.930 –> 00:13:25.940
Doug Landis: Yeah.

94
00:13:25.940 –> 00:13:28.780
Tomasz Tunguz: And you know, yeah, I’m so.

95
00:13:28.780 –> 00:13:42.409
Doug Landis: The reason why I’m freaking out. I’m thinking about the like. The Sdr manager who’s like, okay, I’m gonna go hire some Sdrs right now let’s see how they can use technology to actually, you know, drive leverage. So instead of having 8, I only need 4,

96
00:13:42.920 –> 00:13:43.880
Doug Landis: it’s it’s so.

97
00:13:43.880 –> 00:14:07.010
Tomasz Tunguz: Exactly right, that’s what. So I think that’s what you want to see that sophistication, I mean, like, it’s kind of like. 20 years ago, when people were really good at Googling right? But but this is the productivity is like 100 x right compared to that. And it’s knowing how to use these machines, and Cloud has its own idiosyncrasies and chatgpt open AI. Its system has its own idiosyncrasies like, how can you chain them together.

98
00:14:07.140 –> 00:14:08.350
Tomasz Tunguz: put different systems together.

99
00:14:08.350 –> 00:14:08.850
Doug Landis: Oh.

100
00:14:08.860 –> 00:14:36.850
Doug Landis: so I guess it raises a slightly different question, which is, how does this, then, have impact on a company’s culture? I feel like when you showed that graph, one of the things that immediately jumped out is, if my expectations for the junior folks are so much greater. Well, I mean if if I believe that their output is so much greater than my expectation is likely a lot greater for the junior folks, likely even for the senior folks because you have so much more leverage. So if you start missing numbers, if you start kind of missing plan, whether it’s hiring, etc.

101
00:14:36.850 –> 00:14:48.600
Doug Landis: maybe does that. I mean, maybe the impact feels greater because it’s like, look, we gave you a lot of leverage. We gave you a lot of runway, and you’re still not able to do that. How does all this change a company’s culture.

102
00:14:49.730 –> 00:15:01.419
Tomasz Tunguz: I think. Well, I I hope it improves the culture. I think you get to a place where the teams, the expectations of high performance are much broader, I think. As an so for a manager.

103
00:15:01.650 –> 00:15:13.620
Tomasz Tunguz: your expectation of the team is, you can do a whole lot more than you could. Right? I mean, you just you were just talking about 1 million in arr per employee, like best in class is like, I don’t know 2, 300 K and Arr, so basically tripling and productivity, maybe more.

104
00:15:13.870 –> 00:15:29.760
Tomasz Tunguz: So you’re expecting a whole lot more, I think, as an individual contributor doer, because you’re doing so much more. You’re kind of building an index of all the different workflows. And so if you’re strong in one area, you might be weak in another, you should be able to actually manage that pretty effectively.

105
00:15:30.230 –> 00:15:35.443
Tomasz Tunguz: and the teams themselves will be smaller. I think. You know, this might be a little bit

106
00:15:36.320 –> 00:15:53.599
Tomasz Tunguz: idealistic, but all of us remember when we were working in a small team, and we loved working in a small team because we could move really fast, and things were simple. And now I don’t think it’s crazy to say, like, you can achieve a hundred 1 million in arr on a really small team. And so maybe we can extend that sort of

107
00:15:54.110 –> 00:16:00.769
Tomasz Tunguz: that that nostalgic period, or the 2 pizza box period much, much longer. It’ll be a lot more fun.

108
00:16:01.120 –> 00:16:02.290
Doug Landis: Pizza box.

109
00:16:03.065 –> 00:16:03.580
Doug Landis: Yeah.

110
00:16:03.580 –> 00:16:05.279
Tomasz Tunguz: No, that’s a visa thing right. He has no meat.

111
00:16:05.280 –> 00:16:06.080
Doug Landis: Yeah. Totally.

112
00:16:06.080 –> 00:16:06.960
Tomasz Tunguz: Yeah, yeah.

113
00:16:07.200 –> 00:16:16.859
Doug Landis: That’s exactly so. So let’s let’s talk about pricing a little bit here, because this is a really important lever, I guess, in the world, especially in the world, to go to market when it comes to pricing and packaging.

114
00:16:16.860 –> 00:16:39.250
Doug Landis: It feels like, you know, there’s so many, and everyone has opinion about which was the right way. Is it outcome based? Is it? Usage is a hybrid, is it? You know it feels like the one thing that I think we’re forgetting about is the fact that, like the AI inference costs are variable and they’re unpredictable. And they’re usually built in right. There’s some element of variability in terms of the cost.

115
00:16:39.250 –> 00:16:47.299
Doug Landis: and in many cases, if you’re just a wrapper on top of a model, then you’ve got to. You’ve got to add in that cost. What like, how? How are?

116
00:16:47.300 –> 00:17:05.710
Doug Landis: How should companies be thinking about? Especially all these go to market companies that are launching right now? That are, whether it’s an Aisdr or AI Rfp vehicle or demo automation whatever. How should they be thinking about pricing with these AI driven features in a kind of a scalable way.

117
00:17:05.859 –> 00:17:35.559
Tomasz Tunguz: Yeah, I well, you’re in an optimal place. I think your your margins increase with time. So all you have to do is raise enough actual capital to to to wait. I mean, you look at the deep, seek moment. You’ve seen a thousand X reduction in inference cost in the last 2 years. I think you’ll probably see another 1,000 next year just on the algorithmic side. Then the hardware itself is improving. Google just said that they’ve improved efficiency. 5 x on their hardware. So you’re looking at like, I don’t know, like a million X reduction in inference cost. And so let’s say, overall inference volumes increase by a trillion.

118
00:17:35.829 –> 00:17:42.119
Tomasz Tunguz: you know, the overall market will increase, but your margins will improve it. You don’t have to do anything. Yeah. So

119
00:17:42.180 –> 00:17:45.810
Tomasz Tunguz: as long as you enough capital to be able to sustain. I think you’ll be in a good place.

120
00:17:45.810 –> 00:17:59.649
Doug Landis: So, okay, that leads me to my next question. As a Vc things are shifting right? So arguably, companies may not necessarily need to raise as much capital. Or how about this? I’ve been talking to my co-founder. I’m like, okay, so we may not need to raise past Series B,

121
00:17:59.920 –> 00:18:10.619
Doug Landis: if we can build this company. If if we’re capital efficient, if we’re cash, flow positive from the get, go if we are hyper focused on efficiency while we’re still scaling and growing.

122
00:18:10.770 –> 00:18:15.420
Doug Landis: How should we think about raising capital beyond maybe seed or series? A. How does that?

123
00:18:15.420 –> 00:18:15.850
Doug Landis: Yeah.

124
00:18:15.850 –> 00:18:16.769
Doug Landis: Change your world?

125
00:18:16.770 –> 00:18:33.240
Tomasz Tunguz: I, so I don’t think it changes it very much. And so when I started in venture, there was a slide on the end of every deck that showed how much cash the company would burn, and then they would add 20%. And that was the round size in like 2012 that stopped being the case, it was. The round sizes were no longer a function of burn.

126
00:18:33.240 –> 00:18:37.550
Doug Landis: Crazy. Right? Here’s the cash you’re gonna burn. Here’s 20%.

127
00:18:37.550 –> 00:18:55.329
Tomasz Tunguz: Yeah, that’s you know. Wait, that’s what it was. And and starting in like 2012, that was no longer. It was supply demand. It’s like you want to raise roughly this amount, what is the market willing to give us? And the bigger the auction you create, the more that you can raise and lower your cost of capital. And so, for the most, you know, for the fastest growing companies.

128
00:18:55.460 –> 00:19:24.920
Tomasz Tunguz: I it’s I think it’s been at least 10 years since there was any sort of bottoms up rationale for raising a certain amount of money. It’s like we want to raise from a certain kind of investor, a certain size, balance sheet for. And so it’s the same dynamic. Other reasons you might raise capital produce secondary liquidity for the employees, or finance and acquisition, or whatever it is. So I don’t. I don’t know if it it I think maybe put it a different way. The fraction of companies that could Bootstrap, if they desired, are higher than they have been in the past

129
00:19:24.920 –> 00:19:25.540
Tomasz Tunguz: because of some.

130
00:19:25.540 –> 00:19:25.880
Doug Landis: Interesting.

131
00:19:25.880 –> 00:19:26.550
Tomasz Tunguz: Since he gave.

132
00:19:26.550 –> 00:19:30.799
Tomasz Tunguz: Yeah, yeah. So maybe they can get a little further along before they go out and start to raise.

133
00:19:30.800 –> 00:19:31.190
Doug Landis: Which gives

134
00:19:31.190 –> 00:19:44.869
Doug Landis: more leverage in terms of valuation. Look, valuation is kind of somewhat finger in the air. How much people are willing to pay. If you can get further along based on bootstrapping, then that kind of gives you a bit more leverage, and and your your Vc. Conversations super.

135
00:19:44.870 –> 00:19:45.840
Tomasz Tunguz: Exactly right.

136
00:19:46.847 –> 00:19:52.459
Doug Landis: Going back to kind of the customer interactions in regard to AI.

137
00:19:52.940 –> 00:20:00.919
Doug Landis: Do you think AI is going to fully own a customer interaction. And if so, like? If so, what markets are those best for

138
00:20:01.040 –> 00:20:08.510
Doug Landis: versus where you know? Where’s the handoff in in bigger markets? Or are we always gonna have a hybrid model. Just curious.

139
00:20:08.510 –> 00:20:10.318
Tomasz Tunguz: Yeah. Yeah. So I think,

140
00:20:11.560 –> 00:20:22.788
Tomasz Tunguz: okay, couple of different. I think almost all the support is will be fully automated. The Nps coming out of AI support is much higher than humans, because their patience, the patience of the robots is infinite. Patience of the human is not.

141
00:20:23.050 –> 00:20:23.650
Doug Landis: Yeah.

142
00:20:23.650 –> 00:20:50.929
Tomasz Tunguz: And then I think, you see, Sdr, it used to be, you need to justify an Aesdr model. You probably needed an Acv of 20 k. Maybe 15 K. If you were super efficient. I think you’ll see. Sdr push. I think one of the major opportunities is to push to lower acd products, prosumer products, even consumer products where you have somebody educating the buyer. And now, all of a sudden, because the marginal cost is 0, you have an opportunity. And then, like we said before, I think getting to like 90% education of the buyer.

143
00:20:50.930 –> 00:20:59.829
Tomasz Tunguz: few questions about technical integration and then really having a relationship. Maybe a dinner. That part won’t be automated. But but a lot of these interactions will be.

144
00:21:00.920 –> 00:21:02.159
Doug Landis: Yeah, yeah.

145
00:21:02.470 –> 00:21:26.940
Doug Landis: that makes so much sense, especially on the Sdr side. Well, especially when you’ve got companies out there, you know. Like one, mind, for example, that’s like, basically can do all the selling, the low end selling for you. So it’s like, Hmm, okay, if they can handle like low end. Commercial transactions that was always one of the biggest challenges. Right? You’re like, well, what about the smaller deals? Sub 10 k deals like who handles that your Sdrs may not be ready for it. But I tell you what, your Smbas don’t really want to

146
00:21:26.990 –> 00:21:32.890
Doug Landis: talk to a customer, that’s, you know, 10 K. Right as an example. So I think there’s a great opportunity or market there for it

147
00:21:32.950 –> 00:21:34.780
Doug Landis: for technology to help.

148
00:21:34.910 –> 00:21:43.679
Tomasz Tunguz: Right, and you know it’s the the difference between unassisted and assisted conversion rates is 4 x, and so at some point there’s break even right. There’s an Roi on investment.

149
00:21:44.480 –> 00:21:46.720
Doug Landis: Wow! There’s a Stat. Julia

150
00:21:46.830 –> 00:22:11.720
Doug Landis: that needs to be shared broadly. That’s crazy. Let’s talk about data for a second here, because one of the things that I know that a lot of companies really struggled with, with all the models coming out was the security of having these open models. Whether it’s using chat, Gpt cloud, etc, within their organization. And and you’ve got these AI 1st companies that are largely built on top of these models. How is that impacting companies decisions on being able to leverage new technology.

151
00:22:12.260 –> 00:22:14.240
Doug Landis: How have you? How have you seen that? I’ve got

152
00:22:14.850 –> 00:22:23.089
Doug Landis: where companies are like? No, we can’t. We can only use this part of it because this is touching, you know, Chat Gpt, and that’s not allowed according to our policy.

153
00:22:23.300 –> 00:22:42.479
Tomasz Tunguz: Yeah, I think. So 12 months ago, biggest blocker to AI adoption, I think, was security today. It’s not some we see it less. There are obviously sensitive groups. There are large enterprises who want to run it on cloud prem, whatever in their Vpc. And then there’s a Federal government. Certain people, I intelligence community needs to run it. Air gapped.

154
00:22:42.560 –> 00:23:05.630
Tomasz Tunguz: What we find is most people in the enterprise typically start with a model that’s off the shelf, and they will either decide like I’ll buy an Api or I’ll go to an inference as a service provider, and then over time they they whittle them those large language models down to these constellations of small models, because there’s about a thousand X reduction in cost.

155
00:23:05.630 –> 00:23:06.060
Doug Landis: Hmm, so.

156
00:23:06.060 –> 00:23:19.650
Tomasz Tunguz: So it’s the the limiting factor is no longer security. It’s all in the enterprise. It’s now about Roi making sure that the margins, the contribution margins or the the burn of the individual team is improving pretty significantly. At least, that’s what we hear in market.

157
00:23:20.440 –> 00:23:24.729
Doug Landis: That’s so smart. It’s by the way, it’s only been 12 months.

158
00:23:25.113 –> 00:23:25.880
Tomasz Tunguz: Thanks. Everyone.

159
00:23:25.880 –> 00:23:29.909
Doug Landis: We’re like. It’s changing. Things are changing so fast. It’s crazy.

160
00:23:29.910 –> 00:23:38.450
Tomasz Tunguz: It’s hard to keep, I mean, I think I was chatting with a Vp of engineering. And he was. I was asking like, Hey, you know what’s your AI strategy? And he said, it’s really hard, because.

161
00:23:39.070 –> 00:23:45.909
Tomasz Tunguz: the models themselves change really fast, and whenever the models change 20 to 40% of the prompts break.

162
00:23:46.850 –> 00:24:05.969
Tomasz Tunguz: And so imagine like, you release a product. And 20% of the functionality breaks. And the hard part is like, you definitely want to move to the next model. Right? But you have this like, you have this release cycle where you know the AI companies that are selling these models obviously want to move you to the next model and the next model, because their pricing power is the strongest. There.

163
00:24:05.970 –> 00:24:06.410
Doug Landis: Right.

164
00:24:06.410 –> 00:24:29.180
Tomasz Tunguz: Difference in pricing. And so but you have. So there’s this dynamic where an enterprise is like, I just want to freeze the model. And so I wonder if at some point some of these like open source models, or some of these inference companies, will have to say, like this model, will exist at this level of Sla for 24 months, and we will not change it. The challenge is the data, like the web obviously is updating all the time and.

165
00:24:29.180 –> 00:24:29.690
Doug Landis: Right.

166
00:24:29.690 –> 00:24:34.579
Tomasz Tunguz: And so the underlying data changes. But this is a this is a point of friction for a lot of enterprise.

167
00:24:34.580 –> 00:24:35.230
Doug Landis: Yeah.

168
00:24:35.390 –> 00:24:56.809
Doug Landis: Well, there’s a massive cost associated with that as well. Right? So if I’m constantly like, if you tell me, I need to lock down this model for the next 24 months. Now, I got to basically manage and maintain that. Well, meanwhile, we’re also continuing to invest, to stay ahead of the competition that’s in the market. It’s just it’s it’s kind of it’s kind of mind boggling. But it feels like that’s the new version of tech debt, right?

169
00:24:56.810 –> 00:24:59.790
Tomasz Tunguz: Great, great framing great. That’s exactly right. That’s exactly.

170
00:24:59.790 –> 00:25:17.180
Doug Landis: It’s so good. It’s a totally. It’s just a different form of tech debt. Okay, one kind of last question here for you and Julie. I don’t know if we have other questions coming in, but if you had to bet on one Major AI driven, go to market innovation that’s going to dominate the next 5 years. What would that be?

171
00:25:17.450 –> 00:25:18.270
Tomasz Tunguz: Training.

172
00:25:18.450 –> 00:25:25.079
Tomasz Tunguz: I I think, the training of salespeople will be the training and the evaluation of salespeople will be fully automated.

173
00:25:26.210 –> 00:25:54.690
Doug Landis: The training. It’s interesting. You just. I’m glad you added that second piece, because it’s the evaluation of salespeople like, you know, you come into a team. You take over a team, and the very 1st thing you do is you have to assess my, the butts and seats right? Who do I have, and just trying to determine if they’re if they’re kind of if they’re any good, or if they were just lucky in terms of how they hit their numbers or like, what am I working with is actually really hard. It’s trying to assess, like their skills, their knowledge, their ability to actually go execute. I think that the evaluation piece is equally as important as the training piece.

174
00:25:55.300 –> 00:25:59.715
Tomasz Tunguz: Yeah, I think. I mean, there was this paper that was published on

175
00:26:00.170 –> 00:26:09.439
Tomasz Tunguz: creativity of AI versus humans and humans created AI. Creativity is better than human creativity. I think if you had, yeah, if you.

176
00:26:09.440 –> 00:26:11.030
Doug Landis: So brutal.

177
00:26:11.030 –> 00:26:14.489
Tomasz Tunguz: Well, it’s just because, like, think about like the.

178
00:26:14.950 –> 00:26:34.419
Tomasz Tunguz: I think the analogy is like, you’re playing a chess computer. The chess computers, knowledge of all the history of chess or the computer’s knowledge of all the blog posts that are relevant, or all the videos just much larger than a human can store in their brain. And so they can pick from a bunch of different areas. Then it’s more creative. And so I think there’s an analogy to be made here, which is

179
00:26:35.162 –> 00:26:38.510
Tomasz Tunguz: A sales manager has a certain amount of experience

180
00:26:38.760 –> 00:26:44.440
Tomasz Tunguz: with different kinds of products and different kinds of go to markets, but, like an AI sales evaluator, has

181
00:26:45.260 –> 00:26:55.519
Tomasz Tunguz: universal knowledge across every kind of technique and and different kinds of sales calls across millions of different conversations, or maybe more. And so you’ll probably end up with a better coach. That yeah.

182
00:26:56.120 –> 00:27:14.820
Doug Landis: What kind of so what kind of training is it more like prompt training like how to use and how to leverage technology to give you leverage in your role? Or is it how to have advanced conversations? Given the fact that AI has given you so much intelligence and information like what kind of training? Where is that sweet spot in the training space? I’m asking this question for all my friends who run their own training companies.

183
00:27:14.820 –> 00:27:26.060
Tomasz Tunguz: Yeah, no, I think I mean, I don’t. Have you seen the latest Gemini ad? It’s a i want to say it’s like a 60 year old, dad, and he’s in his kitchen, and he’s practicing his job. Interviews with Gemini live.

184
00:27:26.060 –> 00:27:26.510
Doug Landis: Yeah.

185
00:27:26.680 –> 00:27:41.140
Tomasz Tunguz: And so I think about if I’m a salesperson. Okay, I can pitch. I mean, how many, if I come in like, I have one or 2 opportunities to pitch in front of my manager, and maybe he or she will listen to a handful of my calls before

186
00:27:41.280 –> 00:27:45.760
Tomasz Tunguz: they sent me. But imagine I could spend my whole 1st week

187
00:27:46.700 –> 00:27:49.860
Tomasz Tunguz: I mean I could have 50 different pitches right.

188
00:27:49.860 –> 00:27:51.130
Doug Landis: Yeah, yeah.

189
00:27:51.130 –> 00:27:51.739
Tomasz Tunguz: I can.

190
00:27:51.740 –> 00:27:53.110
Doug Landis: Onboarding woo.

191
00:27:53.110 –> 00:28:14.420
Tomasz Tunguz: Yeah, extremely accurate and precise feedback. And I think there’s there’s both a benefit to the individual in that their rate of improvement will have a completely different slope. And then there’s a product marketing, positioning benefit, which is, you can really enforce a single message through the go to market org in a way that was just not possible. Before

192
00:28:15.130 –> 00:28:16.219
Tomasz Tunguz: right, you can

193
00:28:16.220 –> 00:28:20.530
Tomasz Tunguz: kind of like enforce compliance, and then we can debate whether or not that’s good or bad.

194
00:28:20.710 –> 00:28:21.540
Doug Landis: Yeah, right, yeah.

195
00:28:21.778 –> 00:28:27.259
Tomasz Tunguz: Because maybe it leads to less experimentation. But, like you can have a single voice, a single voice in market in a way.

196
00:28:27.260 –> 00:28:32.149
Doug Landis: Yeah, totally. That’s that’s that’s assuming that the messaging coming from marketing is actually on point.

197
00:28:32.860 –> 00:28:40.190
Doug Landis: But we don’t want to have that conversation about the the challenges between sales and marketing, alignment and integration. That’s not what we’re here to talk about.

198
00:28:40.620 –> 00:28:46.129
Tomasz Tunguz: I think. Are we, Julia? Do we have any other questions, or are we at at time here.

199
00:28:46.660 –> 00:28:57.559
Julia Nimchinski: Far at time hate to do this. We have a lot of questions about Google search and your take on advertising. I don’t know. Advertising within Chat Gpt, and all sorts of

200
00:28:57.720 –> 00:29:01.400
Julia Nimchinski: AI products. What are your thoughts? Tomaz and Doug.

201
00:29:04.540 –> 00:29:10.610
Doug Landis: I, what’s the question? Is it like how to use advertising? Chat? Gpt. To create ads, to run house.

202
00:29:10.610 –> 00:29:11.150
Doug Landis: search.

203
00:29:11.150 –> 00:29:11.670
Julia Nimchinski: Heavily.

204
00:29:11.670 –> 00:29:12.040
Doug Landis: O.

205
00:29:12.040 –> 00:29:15.020
Julia Nimchinski: Yeah, how will AI disrupt SEO, basically.

206
00:29:17.770 –> 00:29:22.690
Tomasz Tunguz: We’ve been. So we’ve been spending a lot of time here. I think. I think

207
00:29:23.940 –> 00:29:38.480
Tomasz Tunguz: the so on the SEO market. It’s less clear on the Sem market search engine marketing. There was a data point coming out of Google’s most recent earnings call where they said the monetization on AI searches is equivalent to classic searches which blew my mind. I didn’t think that would be the case.

208
00:29:38.640 –> 00:29:58.750
Tomasz Tunguz: and so, you know I have. I had sort of wondered whether Google would come out of the AI wave with just as much search traffic, but an inferior ads business, because the cost of inference was higher. Maybe the click through rates were less. But it turns out that they aren’t. They’ll have just as strong of a business there on the AI SEO front.

209
00:29:59.080 –> 00:30:18.450
Tomasz Tunguz: 50% of traffic today on the Internet is for robots, or it’s robots scanning. It’ll probably be 95% in 3 years. And so I don’t know if you and I will surf web pages the way that we do today, we’ll probably have an agent that goes and gathers information. So the Internet itself will change pretty meaningfully. I think it will shift to be much more text based

210
00:30:19.180 –> 00:30:24.849
Tomasz Tunguz: can understand text better. You kind of argue with the multimodal models. But

211
00:30:25.265 –> 00:30:31.174
Tomasz Tunguz: it’s just a lot less expensive to understand a text based web page for a computer than it is for

212
00:30:31.680 –> 00:30:32.700
Tomasz Tunguz: for human.

213
00:30:33.320 –> 00:30:33.910
Tomasz Tunguz: And then the.

214
00:30:33.910 –> 00:30:42.359
Doug Landis: Then what are humans, gonna then what are humans gonna do on the Internet? Besides, use social media and post pictures of their dogs and and political rants.

215
00:30:42.360 –> 00:30:55.149
Tomasz Tunguz: It’ll just be a different. I mean, you’ll listen. I think you’ll listen to the news right like you might listen to a radio broadcast about what’s happening inside of your company on the way to work. So I think you’ll spend as much time with a computer. But you won’t spend as much time in front of a screen.

216
00:30:55.320 –> 00:31:02.166
Tomasz Tunguz: Yeah, that’s which would by, although, which would be great cause. Then maybe you start paying attention to what the world around you a little bit more.

217
00:31:03.200 –> 00:31:08.480
Doug Landis: Versus being glued to your screens, and you know how many people you know fall off of cliffs because they’re not paying attention.

218
00:31:08.720 –> 00:31:10.469
Doug Landis: Maybe we’ll save some lives.

219
00:31:10.965 –> 00:31:16.460
Doug Landis: Man, this has been such a fun conversation. Thank you, Julia, for setting this up, Tomas. Your Rad.

220
00:31:16.770 –> 00:31:19.210
Tomasz Tunguz: No, Doug, this is a pleasure. Thanks, Julia.

221
00:31:19.430 –> 00:31:20.270
Julia Nimchinski: Thank you so much.