2025-11-21_ai_team_sprint_planning_for_next_week_3b541548

Meeting Title: AI Team Sprint Planning for Next Week Date: 2025-11-21 Meeting participants: Gabriel Lam, Uttam Kumaran

---

WEBVTT

1 00:02:02.110 ⇒ 00:02:03.520 Uttam Kumaran: Hello!

2 00:02:05.210 ⇒ 00:02:06.270 Gabriel Lam: Hey, how’s it going?

3 00:02:06.270 ⇒ 00:02:07.400 Uttam Kumaran: Hey, good.

4 00:02:08.600 ⇒ 00:02:09.340 Gabriel Lam: Sorry.

5 00:02:09.340 ⇒ 00:02:15.349 Uttam Kumaran: Yeah, I just need to… I need to join the meetings, get everybody on rails, and then get out. That’s, like, my job for the most part.

6 00:02:15.350 ⇒ 00:02:16.810 Gabriel Lam: You’re all good.

7 00:02:17.290 ⇒ 00:02:21.260 Gabriel Lam: I was just like, you know, I know you’re busy, if we need to reschedule, we can reschedule.

8 00:02:21.260 ⇒ 00:02:22.160 Uttam Kumaran: No, no, no, I won’t

9 00:02:22.500 ⇒ 00:02:27.569 Uttam Kumaran: You’re… you’re who I wanna spend… as… like, I wanna… I wish life was flipped, like, I actually want

10 00:02:28.540 ⇒ 00:02:31.870 Uttam Kumaran: 90% of my time on the stuff that you’re helping with.

11 00:02:32.010 ⇒ 00:02:38.209 Uttam Kumaran: And it’s such a shame, but dude, it’s like, it’s tough because we’re not, like, a product company, so we don’t have, like, R&D budget.

12 00:02:38.310 ⇒ 00:02:41.420 Gabriel Lam: necessarily. In fact, like, a company like ours should have no…

13 00:02:41.420 ⇒ 00:02:45.909 Uttam Kumaran: we should not be working on stuff like we’re working on. It’s mainly because I have, like, a compulsion.

14 00:02:45.910 ⇒ 00:02:46.470 Gabriel Lam: Hmm.

15 00:02:46.470 ⇒ 00:02:47.330 Uttam Kumaran: This type of stuff.

16 00:02:48.720 ⇒ 00:02:55.770 Uttam Kumaran: But, like, we should not be dedicating budget to these. Not because there’s no ROI, it’s actually, like, not common. So…

17 00:02:55.770 ⇒ 00:02:56.200 Gabriel Lam: Right.

18 00:02:56.200 ⇒ 00:03:01.190 Uttam Kumaran: For me, it’s like, I’m doing my best to see, like, okay, what can I push strategically without

19 00:03:01.610 ⇒ 00:03:15.749 Uttam Kumaran: being so involved, but then also, like, make sure that it’s supported. I mean, I think before you joined, it was very hard for the team, like, we had good weeks, we had tough weeks, mainly because of… of me, but I think it’s… it’s getting a lot better now. So yeah, I just want to, like…

20 00:03:16.150 ⇒ 00:03:27.359 Uttam Kumaran: maybe spend time with you, and what I’m actually gonna… let me just grab my, my notebook one sec, it’s just in the other room, and then we can just sort of give you, like, kind of the ideas I’m thinking about.

21 00:03:28.100 ⇒ 00:03:31.020 Uttam Kumaran: You’ll have this transcript to work with, but yeah, we can…

22 00:03:31.590 ⇒ 00:03:34.210 Uttam Kumaran: Just, like, talk broadly about the platform. Oh, one sec.

23 00:03:34.210 ⇒ 00:03:35.230 Gabriel Lam: For sure.

24 00:03:59.020 ⇒ 00:04:06.200 Uttam Kumaran: Okay, so… Couple things that… I mean, you now, I think, have a good sense of, like.

25 00:04:06.870 ⇒ 00:04:19.250 Uttam Kumaran: And then… and again, like, you let me know if I’m just going too broad, but maybe I just want to kind of, like, give you where I think the platform could go, and then we can get into any direction, but, like, this is a conversation I wanted to have, like, a few weeks ago.

26 00:04:19.660 ⇒ 00:04:20.390 Uttam Kumaran: But…

27 00:04:20.870 ⇒ 00:04:29.169 Uttam Kumaran: probably when you joined would have seemed like, what is this guy talking about? Now that you see the core components of the platform, I think you’ll see how all of these are…

28 00:04:29.500 ⇒ 00:04:34.700 Uttam Kumaran: are possible. Basically,

29 00:04:37.080 ⇒ 00:04:45.040 Uttam Kumaran: there’s two components to the platform. There’s one are these, like, use case-specific tools or workflows.

30 00:04:45.220 ⇒ 00:04:56.730 Uttam Kumaran: The second piece is actually building the platform in a very composable and in a way where you can move the LEGO pieces in a way to support the next use case.

31 00:04:56.860 ⇒ 00:05:05.050 Uttam Kumaran: But not only that, is you can move the LEGO pieces in a way and develop in a way where anyone can build parts of the application.

32 00:05:05.240 ⇒ 00:05:19.550 Uttam Kumaran: Right? And so you have a two-fold challenge as the product leader here. One is you’re not only supporting building new features, but I actually want anybody in the company to be able to build end-to-end features.

33 00:05:19.930 ⇒ 00:05:26.930 Uttam Kumaran: Like, I don’t… and what does that require? That requires that people don’t need… need to be able to use a tool like Cursor.

34 00:05:27.040 ⇒ 00:05:33.850 Uttam Kumaran: to plan… use a tool like Cursor or Magic Patterns to… Plan a new feature design.

35 00:05:34.030 ⇒ 00:05:39.850 Uttam Kumaran: And also, maybe take that design and get it implemented halfway, without…

36 00:05:40.070 ⇒ 00:05:42.439 Uttam Kumaran: Having to have, like, tons of roadblocks.

37 00:05:42.530 ⇒ 00:05:57.549 Uttam Kumaran: So that’s… that is truly what will unlock the speed of development. It’s actually, like, removing the entire team from the ideation and core development. For example, if Rico wakes up one day and is like, I’m frustrated with

38 00:05:57.630 ⇒ 00:06:04.210 Uttam Kumaran: how the prompt works for the linear tickets. Just like you did, he should be able to go make that change.

39 00:06:04.210 ⇒ 00:06:04.870 Gabriel Lam: Hmm.

40 00:06:04.870 ⇒ 00:06:10.140 Uttam Kumaran: That is not very common, where always the engineers or the product person’s the gatekeep.

41 00:06:10.330 ⇒ 00:06:15.300 Uttam Kumaran: not… maybe not because they want to, but because they have to. And so now that AI is here.

42 00:06:15.680 ⇒ 00:06:18.589 Uttam Kumaran: I actually want to develop the system in a way where

43 00:06:18.660 ⇒ 00:06:38.290 Uttam Kumaran: non-technical folks can build things on the platform. That is, like, what a true platform means, right? The AI team will still work on the probably the most complicated things, and the primitives, like, okay, we need a linear ticket API, okay, we need a Slack API, okay, we need a way to hold prompts somewhere, like laying views.

44 00:06:38.290 ⇒ 00:06:46.300 Uttam Kumaran: those were all things that are probably in their purview, but more and more, I want to… I want someone on the team to wake up frustrated.

45 00:06:46.630 ⇒ 00:06:51.719 Uttam Kumaran: say, I wish the platform had this, and actually be able to go do… build that thing, right?

46 00:06:52.270 ⇒ 00:07:07.639 Uttam Kumaran: I know there’s a lot of steps to get people to do that, but, like, that’s, like, what I want to see, because otherwise, we’re never gonna have the R&D budgets of a big firm, and so this is gonna continue… this is not gonna get… like, I’ve already pushed

47 00:07:08.120 ⇒ 00:07:11.849 Uttam Kumaran: To get stuff out as fast as it could probably go.

48 00:07:12.060 ⇒ 00:07:22.179 Uttam Kumaran: Like, I wouldn’t… I would not… I’m not really, like, confident that I can say, cool, we’re shipping things in a week, ship it in 3 days. Like, there are some physics that you can’t overcome.

49 00:07:22.470 ⇒ 00:07:31.809 Uttam Kumaran: So, I feel like we’re fast enough, but as you can see, I hope we’ll kind of walk through, there’s still, like, probably 50 more features that would be super tremendous for us, and so…

50 00:07:31.980 ⇒ 00:07:45.289 Uttam Kumaran: the only way to accomplish that is actually by letting other people be able to build, right? So that’s, like, maybe one piece, like, what do you think about that? And then I can talk about, like, other… kind of, like, some of the different areas we can go.

51 00:07:45.940 ⇒ 00:07:51.820 Gabriel Lam: I like that idea, I think… it… Alleviates a lot of the…

52 00:07:52.920 ⇒ 00:08:06.270 Gabriel Lam: I think it’s because they know the pain points best, right? So they know exactly what it is that they need, and they know exactly what it is they want out of it. And I think… I think there’s two sides. There’s… on the one hand, it’s like, okay, we have these…

53 00:08:06.460 ⇒ 00:08:12.769 Gabriel Lam: time sinks or bottlenecks that we experience on the day-to-day. On the other hand, I think there also is a…

54 00:08:12.930 ⇒ 00:08:14.980 Gabriel Lam: Training curve, to be like.

55 00:08:14.980 ⇒ 00:08:15.880 Uttam Kumaran: Yeah, yeah, yeah.

56 00:08:15.880 ⇒ 00:08:16.659 Gabriel Lam: This is…

57 00:08:16.830 ⇒ 00:08:21.449 Gabriel Lam: now that we have these tools, this is how you can get to where you need to go. And I think the…

58 00:08:21.590 ⇒ 00:08:24.840 Gabriel Lam: The client hub, or maybe the stand-up position is a good example where

59 00:08:25.110 ⇒ 00:08:44.239 Gabriel Lam: because it’s not the typical workflow, there is some getting used to the fact that, like, hey, we’re gonna start having this source of truth for you to look back to run your meetings, or you’re gonna start talking to an AI instead of, like, Hannah for these interviews. So I think there’s those two things. On the pain point side, I think…

60 00:08:44.470 ⇒ 00:08:48.430 Gabriel Lam: Even… Well, okay, I think there’s…

61 00:08:49.860 ⇒ 00:08:56.069 Gabriel Lam: getting them to be able to… getting them to be able to execute on ideas, I think, is going to be a lift.

62 00:08:56.300 ⇒ 00:08:59.019 Gabriel Lam: In the sense that… Like, what are the two.

63 00:08:59.020 ⇒ 00:09:00.369 Uttam Kumaran: Hearing impossible.

64 00:09:01.530 ⇒ 00:09:08.800 Uttam Kumaran: So, Sam, 2 months ago, was like, dude, there’s no way. Now, I’m not hearing Impossible, I’m hearing…

65 00:09:09.030 ⇒ 00:09:11.690 Uttam Kumaran: Damn, like, imagine the… imagine, like.

66 00:09:12.560 ⇒ 00:09:17.969 Uttam Kumaran: it… I could tell that the hesitancy is like, okay, we still have to teach them a lot, but, like.

67 00:09:18.560 ⇒ 00:09:20.240 Uttam Kumaran: Okay, I agree.

68 00:09:20.240 ⇒ 00:09:21.770 Gabriel Lam: Yeah. We do have to teach…

69 00:09:21.770 ⇒ 00:09:27.420 Uttam Kumaran: how to use magic patterns. We do have to teach some product building fundamental, like.

70 00:09:27.530 ⇒ 00:09:36.589 Uttam Kumaran: I want to build this, you need to give it some context on, like, where do I need data from, what does the UI look like? We will have to do some upskilling on, like.

71 00:09:36.860 ⇒ 00:09:46.020 Uttam Kumaran: product development fundamentals. Similarly, for the folks that go from design to cursor, there will have to be some upskilling on how to use that.

72 00:09:47.100 ⇒ 00:09:51.719 Uttam Kumaran: But I don’t… but, like, I’m telling… everybody in the company’s actually, like, damn smart. I feel like they’ll… they’ll…

73 00:09:52.420 ⇒ 00:10:00.769 Uttam Kumaran: if people really put their mind to it, I think they’ll figure it out. And especially, we have 6 or 7 other engineers on the team, like, on the team.

74 00:10:00.870 ⇒ 00:10:12.450 Uttam Kumaran: Those guys, like, I want them to start ripping features that they want as they work on things for clients. They’re like, damn, I wish I had a… a thing that helped me build out, like, this SOP.

75 00:10:12.630 ⇒ 00:10:18.200 Uttam Kumaran: Okay, like, I want… I want, in that meeting, in that moment, to be like, why don’t you go build it?

76 00:10:19.120 ⇒ 00:10:19.690 Gabriel Lam: Mmm.

77 00:10:19.690 ⇒ 00:10:22.020 Uttam Kumaran: And I want that to not be, like,

78 00:10:22.340 ⇒ 00:10:32.400 Uttam Kumaran: sarcastic comment. I want to be serious, like, hey, why don’t you go give it a shot? Like, you have an idea around building, like, a forecasting module for a client?

79 00:10:32.480 ⇒ 00:10:43.620 Uttam Kumaran: why don’t you write a PRD, get… send it internally, get it approved, go to Magic Patterns, go do that, get the design approved, and then go into Cursor and try to do a first version.

80 00:10:45.150 ⇒ 00:10:57.039 Uttam Kumaran: Right? Like, that’s, like, that’s sort of, like, how I… I think, realistically, some of these will happen. And also, a lot of the use cases aren’t gonna be actually for clients, it’s gonna be the internal people, which…

81 00:10:57.480 ⇒ 00:11:03.990 Uttam Kumaran: They will have a higher, threshold of… Technicality to overcome.

82 00:11:04.750 ⇒ 00:11:07.809 Uttam Kumaran: Books on design, operations, finance.

83 00:11:08.520 ⇒ 00:11:12.729 Uttam Kumaran: They are… but they are also doing so much manual stuff.

84 00:11:13.090 ⇒ 00:11:19.650 Uttam Kumaran: that I think if I can give them an understanding that, hey, you can actually develop some of these things on your own, they will go for it.

85 00:11:20.250 ⇒ 00:11:21.420 Uttam Kumaran: You know.

86 00:11:21.860 ⇒ 00:11:22.460 Gabriel Lam: Yeah.

87 00:11:22.460 ⇒ 00:11:30.950 Uttam Kumaran: like, a good… like a… and I’ll talk you through a couple of examples of where things that are, like… you… I’ll go through the list of things I have, you’ll be like, damn, that’s, like, 2 years of roadmap.

88 00:11:31.750 ⇒ 00:11:32.700 Gabriel Lam: And for.

89 00:11:32.700 ⇒ 00:11:33.300 Uttam Kumaran: For me, I’m like.

90 00:11:33.300 ⇒ 00:11:33.970 Gabriel Lam: No doubt, I love it.

91 00:11:33.970 ⇒ 00:11:39.469 Uttam Kumaran: What… what is the supply… like, what in the supply chain allows us to execute this in the next…

92 00:11:40.100 ⇒ 00:11:44.120 Uttam Kumaran: 3 months. Just, like, what I have to… what we have to work on is, like.

93 00:11:44.380 ⇒ 00:11:50.590 Uttam Kumaran: there’s some things on my list that I’m like, okay, these are too complicated, very novel ideas.

94 00:11:51.110 ⇒ 00:11:53.870 Uttam Kumaran: AI team should handle. There’s also some other things where I’m like.

95 00:11:54.230 ⇒ 00:12:01.269 Uttam Kumaran: Pretty sure if you got Ryan and Rico to sort of, like, get up to speed on how to develop, they could go build some of these, you know?

96 00:12:01.590 ⇒ 00:12:08.639 Gabriel Lam: Yeah. So… Yeah, I think even getting them to the PRD stage is…

97 00:12:08.640 ⇒ 00:12:09.290 Uttam Kumaran: Yeah.

98 00:12:09.900 ⇒ 00:12:13.879 Gabriel Lam: is pretty good, because I think it just gets… even when.

99 00:12:13.880 ⇒ 00:12:14.329 Uttam Kumaran: I was prepping.

100 00:12:14.330 ⇒ 00:12:19.719 Gabriel Lam: the PRD for the linear tickets, it was like, okay, where does it live? How do we want to get there?

101 00:12:20.500 ⇒ 00:12:26.950 Gabriel Lam: what might it even look like? And I think just even having people think of, like, okay, if there’s a problem that I’m facing, how do I want it fixed?

102 00:12:27.610 ⇒ 00:12:29.680 Gabriel Lam: Yes. And I think that part…

103 00:12:30.820 ⇒ 00:12:38.780 Gabriel Lam: I think, sorry, I think the product development part will come with time, where people are like, okay, I can use magic patterns for the UI, I can use cursor for all these things.

104 00:12:38.950 ⇒ 00:12:40.800 Gabriel Lam: And maybe, you know.

105 00:12:41.220 ⇒ 00:12:47.330 Gabriel Lam: the AI team can serve as, like, guidance or facilitating those things, but I think the…

106 00:12:48.310 ⇒ 00:12:55.849 Gabriel Lam: just knowing what they want to build, and being… and being like, hey, I have the agency to be like, I want something built.

107 00:12:56.500 ⇒ 00:13:02.210 Gabriel Lam: how can I do it? Can you help me? And then eventually, as I get used to it, I can imagine seeing, like, okay.

108 00:13:02.350 ⇒ 00:13:04.980 Gabriel Lam: If Rico or Ryan’s like, I want something built.

109 00:13:05.300 ⇒ 00:13:07.240 Gabriel Lam: I kind of know how to get there.

110 00:13:08.110 ⇒ 00:13:13.969 Gabriel Lam: I’ll do my best, and if I need help, I’ll ask, and then the next time, I might know a little more than I did last time.

111 00:13:14.090 ⇒ 00:13:18.120 Gabriel Lam: Yes. So I think there is a, sort of, like…

112 00:13:20.620 ⇒ 00:13:22.959 Gabriel Lam: Yeah, I think there is a sort of the…

113 00:13:23.530 ⇒ 00:13:28.550 Gabriel Lam: the training there. I think the other part would also be… Maybe…

114 00:13:29.290 ⇒ 00:13:30.880 Gabriel Lam: How do I… how do I say this?

115 00:13:35.730 ⇒ 00:13:38.800 Gabriel Lam: It’s like getting people to a point that…

116 00:13:40.250 ⇒ 00:13:49.109 Gabriel Lam: they have time and capacity to be like, okay, I know I have, you know, X number of tickets to handle today, or X number of follow-ups to do, but then…

117 00:13:49.600 ⇒ 00:13:58.539 Gabriel Lam: I know I have time, or I have, you know, an hour a day to dedicate to something I’m interested in, or something I want changed. Yes. Might be another part of that.

118 00:13:58.720 ⇒ 00:14:15.019 Uttam Kumaran: So that’s a cultural thing that, like, if you’re like, hey, maybe people don’t feel comfortable, like, they… or they don’t want to burn time on this, okay, like, I can budget out that people do have time for that. But this is also where, like, we can’t… there’s gonna be no other firm

119 00:14:15.040 ⇒ 00:14:21.119 Uttam Kumaran: That is, like, I want marketing people to ship product applications. There’s not another company that’s doing that.

120 00:14:21.340 ⇒ 00:14:28.059 Uttam Kumaran: Like, I can name you the 2 or 3 companies that I know are trying to do this, and, like, I take some inspiration from.

121 00:14:28.230 ⇒ 00:14:34.970 Uttam Kumaran: But, like… this is sort of a new world, right? So, that being said.

122 00:14:35.490 ⇒ 00:14:40.939 Uttam Kumaran: I’m… if I’m able to ship front-end and back-end code.

123 00:14:41.070 ⇒ 00:14:44.059 Uttam Kumaran: Without knowing anything, we’re not too far.

124 00:14:44.300 ⇒ 00:14:47.680 Uttam Kumaran: So for me, it’s like, how far, and…

125 00:14:48.370 ⇒ 00:14:52.900 Uttam Kumaran: Okay, like, that’s… that’s what our tax is to get over, so… I hear you on that.

126 00:14:54.280 ⇒ 00:14:54.820 Gabriel Lam: Yeah.

127 00:14:55.680 ⇒ 00:15:02.429 Gabriel Lam: Yeah, I think, on that note, there is, like, a general industry…

128 00:15:03.680 ⇒ 00:15:16.519 Gabriel Lam: hesitancy, like, I was talking to a few friends who are like, oh, you know, engineers who are like, AI does write a lot of slop, and I’m like, sure, like, you’re always gonna need someone to read through it, but I think it’s more of a velocity problem getting…

129 00:15:17.250 ⇒ 00:15:18.940 Gabriel Lam: Generally, for people.

130 00:15:19.840 ⇒ 00:15:27.870 Uttam Kumaran: AI does write a lot of slop, but those people are gonna sit on the sidelines and just opine about philosophy about AI.

131 00:15:28.030 ⇒ 00:15:38.399 Uttam Kumaran: I don’t care about any of that. I… we see with our own eye… you’re seeing with your own eyes that we are going end-to-end on new features in a week. This is not like…

132 00:15:38.550 ⇒ 00:15:39.720 Uttam Kumaran: Fairy tale.

133 00:15:39.930 ⇒ 00:15:46.420 Uttam Kumaran: And so, while the rest of the world sits and debates whether AI is worth it or not.

134 00:15:47.020 ⇒ 00:15:54.920 Uttam Kumaran: like, we will keep… we were gonna go forward, right? So, part of this is, like, you have to have a belief that it’s possible.

135 00:15:55.160 ⇒ 00:16:06.649 Uttam Kumaran: Second is, like, okay, but the slop thing turns into, how does slop get introduced in the system? Well, for the most part, it’s because the inputs suck. People are going to AI and saying, I want this.

136 00:16:06.750 ⇒ 00:16:10.989 Uttam Kumaran: and hitting enter. That is not valid, right? So what…

137 00:16:11.030 ⇒ 00:16:19.690 Uttam Kumaran: I’ve always preached that we have an input validation problem. Like, for example, if you’re like, cool, I want to design a system that allows anyone to write PRDs.

138 00:16:19.720 ⇒ 00:16:31.100 Uttam Kumaran: Your AI system should not allow people to create bad PRDs. It should keep pushing on them and say, hey, you have not given me enough information to write something worth publishing.

139 00:16:31.700 ⇒ 00:16:35.159 Uttam Kumaran: you… all you gave me two lines. It should be mean, right?

140 00:16:35.570 ⇒ 00:16:40.709 Uttam Kumaran: it should say that. It should not just go ahead and do it. And this is the problem, is like.

141 00:16:41.020 ⇒ 00:16:48.220 Uttam Kumaran: as part of a product growth motion, all of the tools, Gemini, ChatGPT, their UI is just one chat.

142 00:16:48.300 ⇒ 00:16:54.680 Uttam Kumaran: Right? Yeah. But what do you do? Of course you get… you can do two things. You can work like I do, where I give it a shit ton of info.

143 00:16:54.680 ⇒ 00:17:11.130 Uttam Kumaran: Or you can not, and you give it something, and then you get the output. Yes, like, people are… like, you’re gonna get what you put in. So for us, we don’t have to have that problem, because you can actually structure the exact inputs that you require, and at every step of the way, have AI judge itself and be like.

144 00:17:11.520 ⇒ 00:17:13.130 Uttam Kumaran: I still don’t have enough.

145 00:17:13.240 ⇒ 00:17:14.890 Gabriel Lam: You need to give me more.

146 00:17:15.000 ⇒ 00:17:17.849 Gabriel Lam: I… I don’t mind… this is where also, like.

147 00:17:18.099 ⇒ 00:17:20.749 Uttam Kumaran: Response times is not a KPI.

148 00:17:22.250 ⇒ 00:17:24.000 Uttam Kumaran: Quality is the KPI.

149 00:17:24.210 ⇒ 00:17:24.970 Uttam Kumaran: Right.

150 00:17:25.310 ⇒ 00:17:25.650 Gabriel Lam: Yeah.

151 00:17:25.650 ⇒ 00:17:31.439 Uttam Kumaran: Like, a great PRD is more important than getting a response back within 30 seconds.

152 00:17:31.560 ⇒ 00:17:44.299 Uttam Kumaran: So you should use the thinking models, you should restrict inputs, so, like, that’s the trade-off that I can tell you, is that I don’t care about response times being high, I only care about the PRDs being rich.

153 00:17:44.580 ⇒ 00:17:49.599 Uttam Kumaran: And being something that, at that point, can get reviewed, a few more revisions, and get out the door.

154 00:17:50.300 ⇒ 00:17:54.530 Uttam Kumaran: So that’s how… that’s how I would tackle that… that problem.

155 00:17:55.230 ⇒ 00:17:55.650 Gabriel Lam: Hmm.

156 00:17:55.650 ⇒ 00:18:05.099 Uttam Kumaran: human in the loop is important, and I would rather have more human in the loop to get a better outcome than less, and, like, a risk… and, like, not an accurate outcome.

157 00:18:05.100 ⇒ 00:18:05.600 Gabriel Lam: Yeah.

158 00:18:05.600 ⇒ 00:18:10.739 Uttam Kumaran: You know? So we’re doing that a couple ways, right? We have free text descriptions, we have a couple other things, but I think

159 00:18:10.930 ⇒ 00:18:14.919 Uttam Kumaran: A lot of AI, the nature is to just have it

160 00:18:15.360 ⇒ 00:18:25.649 Uttam Kumaran: go and do everything, versus at different ways, you should think about it as, like, what inputs do I need from the human at these different steps in order to ensure

161 00:18:25.860 ⇒ 00:18:28.010 Uttam Kumaran: at least an A-minus outcome.

162 00:18:28.130 ⇒ 00:18:35.310 Uttam Kumaran: Right? And don’t… and don’t… there are some use cases that are… that KPI is response time, right?

163 00:18:35.430 ⇒ 00:18:41.330 Uttam Kumaran: Chatting with a meeting, saying, hey, what happened in that meeting? You can’t have that take, like, 90 seconds, right?

164 00:18:41.330 ⇒ 00:18:41.810 Gabriel Lam: Yep.

165 00:18:41.810 ⇒ 00:18:45.920 Uttam Kumaran: But there are some use cases, like producing great linear tickets.

166 00:18:47.080 ⇒ 00:19:02.259 Uttam Kumaran: you can wait. I’m down to wait a few minutes and have that run in the background, it does not matter, because the cost of having a bad ticket go out, enter the flow, get to the Monday meeting next week, where nobody knows what it is, because the description sucks.

167 00:19:02.430 ⇒ 00:19:08.509 Uttam Kumaran: That is the, like, causes such a domino effect of crappiness.

168 00:19:08.840 ⇒ 00:19:16.669 Uttam Kumaran: That could have been resolved by another pass at AI, an AI looking at the ticket saying, this is not good enough, I should ask them for more feedback.

169 00:19:17.010 ⇒ 00:19:18.040 Uttam Kumaran: Right? So…

170 00:19:18.700 ⇒ 00:19:24.389 Uttam Kumaran: at the idea generation stage, like, so when you’re creating a linear ticket, when you’re creating a PRD,

171 00:19:25.420 ⇒ 00:19:28.770 Uttam Kumaran: Everything is gonna be downstream of the fidelity of that.

172 00:19:29.220 ⇒ 00:19:35.990 Uttam Kumaran: If that is bad, then everybody who plays telephone from that moment, it’s going… it is naturally gonna degrade.

173 00:19:36.410 ⇒ 00:19:48.709 Uttam Kumaran: Right? Even if it’s the best, it’s going to start to degrade, but it’s the pace, it’s like the half-life of it, right? Right. So, for example, if a ticket’s bad, we may end up talking about it 5 more times.

174 00:19:48.760 ⇒ 00:19:57.210 Uttam Kumaran: Because you talk about it, no one updates a ticket. You talk about it again tomorrow. What did we say yesterday? Talk about it again. Finally, someone updates a ticket.

175 00:19:57.270 ⇒ 00:20:02.980 Uttam Kumaran: But you can see, like, how much did that cost? Like, how much did the bad Fidelity ticket cost?

176 00:20:03.100 ⇒ 00:20:04.770 Uttam Kumaran: Could be hundreds of dollars in waste.

177 00:20:04.770 ⇒ 00:20:05.150 Gabriel Lam: the time.

178 00:20:05.560 ⇒ 00:20:07.580 Uttam Kumaran: Right. And so.

179 00:20:07.750 ⇒ 00:20:14.320 Uttam Kumaran: That’s why nailing it up top, and this is why product managers, everybody has a great… have these jobs, but usually.

180 00:20:15.020 ⇒ 00:20:20.819 Uttam Kumaran: Their job ends up being, like, herding cats, because

181 00:20:21.250 ⇒ 00:20:30.800 Uttam Kumaran: they didn’t spend the time up front on getting the ultimate requirements right. Instead, they’re dealing with the downstream ramifications of bad planning.

182 00:20:30.940 ⇒ 00:20:34.880 Gabriel Lam: Yeah. That’s what I feel like as a product, when I was a product manager and stuff.

183 00:20:35.700 ⇒ 00:20:50.029 Uttam Kumaran: that was never taught, because all you see in product managers are people that are like, every day, what do we do, what do we do? Ultimately, if the plan and the expectations are set clear, the engineers will self-form to that, right?

184 00:20:50.520 ⇒ 00:20:55.309 Uttam Kumaran: But if you… if you have a lot of ambiguity in the system, you’re gonna… you’re…

185 00:20:55.660 ⇒ 00:20:58.209 Uttam Kumaran: You’re gonna stray from the bullseye.

186 00:20:58.390 ⇒ 00:21:16.659 Uttam Kumaran: sometimes you still may hit it, but, like, you can’t… we may not be able to take credit for it, apart from just, like… you… and yeah, so that’s the thing, I think, as part of this whole equation, it’s, like, really considering more opportunities for human in the loop, taking more time to build things, to have things, like, iterate.

187 00:21:16.750 ⇒ 00:21:21.609 Uttam Kumaran: Using the thinking models, and, like, pushing back on the user and saying, you didn’t give me enough.

188 00:21:22.190 ⇒ 00:21:22.660 Gabriel Lam: Yeah.

189 00:21:22.660 ⇒ 00:21:24.690 Uttam Kumaran: ChatGPT will, like, never say that.

190 00:21:24.780 ⇒ 00:21:26.089 Gabriel Lam: I know it’s too much.

191 00:21:26.090 ⇒ 00:21:34.940 Uttam Kumaran: Some of my prompts, I have it in the prompt that says, you need to not immediately jump to the conclusion, you need to ask for feedback.

192 00:21:35.100 ⇒ 00:21:45.279 Uttam Kumaran: and get approval before going to this next phase. It’s in the prompts. Because some of the prompts I’ve developed are, like, really sophisticated, like, project management or SOW things, where

193 00:21:45.500 ⇒ 00:21:49.709 Uttam Kumaran: If I was to take my two-word thing, I’d need to have a conversation with somebody.

194 00:21:49.890 ⇒ 00:21:50.220 Gabriel Lam: Yeah.

195 00:21:50.220 ⇒ 00:21:56.570 Uttam Kumaran: Right? So it needs to take what I first say, be like, okay, great, you answered 2 out of, like, 7 questions, here’s the rest of them, answer them.

196 00:21:56.790 ⇒ 00:22:03.029 Uttam Kumaran: Okay, then, okay, yeah, but let’s go deeper on one thing, two things. Okay, now I have enough. Doing a true, like, discovery, right?

197 00:22:03.550 ⇒ 00:22:05.930 Uttam Kumaran: Some of the prompts are doing a good job.

198 00:22:06.530 ⇒ 00:22:13.469 Uttam Kumaran: But you could structure that into a product itself, something that, like, helps you generate the PRD. It’s very similar to the case study interviewer thing, you know?

199 00:22:13.470 ⇒ 00:22:14.030 Gabriel Lam: Yeah.

200 00:22:14.430 ⇒ 00:22:19.340 Uttam Kumaran: But what did the case study interviewer product itself unlock? You guys unlocked voice?

201 00:22:19.670 ⇒ 00:22:22.579 Uttam Kumaran: You know, you unlock this voice and chat thing.

202 00:22:22.930 ⇒ 00:22:28.310 Uttam Kumaran: You unlock some type of, like, thinking where it takes the inputs, thinks, does it output?

203 00:22:28.490 ⇒ 00:22:31.219 Uttam Kumaran: Those are primitives that you can now use across.

204 00:22:31.340 ⇒ 00:22:33.649 Gabriel Lam: Those are LEGO blocks to use elsewhere.

205 00:22:33.860 ⇒ 00:22:39.630 Uttam Kumaran: Right? So that’s why I wanted to work on that product, because I knew it would unlock voice, because that’s something new.

206 00:22:39.860 ⇒ 00:22:44.600 Uttam Kumaran: And it would unlock a few couple of other things about pre-text inputs and, like, linking.

207 00:22:44.740 ⇒ 00:22:47.899 Uttam Kumaran: But those are all now building blocks that you don’t have to develop again.

208 00:22:47.910 ⇒ 00:22:51.079 Gabriel Lam: Yeah. Right, so if your next feature wants voice, you now have.

209 00:22:51.400 ⇒ 00:22:54.220 Uttam Kumaran: The endpoints, and we have the knowledge on how to develop.

210 00:22:54.740 ⇒ 00:22:55.370 Gabriel Lam: Yeah.

211 00:22:56.620 ⇒ 00:23:01.409 Gabriel Lam: Yeah, I’m with you on that. I’m happy to dive into some ideas.

212 00:23:01.630 ⇒ 00:23:02.190 Uttam Kumaran: Yeah, yeah.

213 00:23:02.190 ⇒ 00:23:07.070 Gabriel Lam: Or you have a giant list, and we can just kind of hash out, you’re like, oh, that looks like something for…

214 00:23:07.270 ⇒ 00:23:14.890 Uttam Kumaran: So let me, yeah, let me, like… let me just do a few seconds on each, so…

215 00:23:16.980 ⇒ 00:23:23.370 Uttam Kumaran: Broadly, there are two types of people in the company. There are people that work on delivery, and there’s people that work on the business, right? So…

216 00:23:23.820 ⇒ 00:23:38.759 Uttam Kumaran: And all of, I think, our features, we need to have a lens by which we look at the feature, and I don’t know yet whether it’s gonna be, hey, you’re a PM, so when you log in, the platform is configured towards your project manager view, or you’re in sales.

217 00:23:38.900 ⇒ 00:23:43.929 Uttam Kumaran: Or you’re delivering, non-delivery. I’m not sure yet, but roughly, I want to start with the users.

218 00:23:44.380 ⇒ 00:23:45.590 Uttam Kumaran: the users.

219 00:23:45.740 ⇒ 00:23:53.579 Uttam Kumaran: are, like, really what this whole thing hinges on. And so I want to start with thinking, like, what do they need? So there’s these types of views.

220 00:23:54.020 ⇒ 00:23:57.989 Uttam Kumaran: a couple of, like, ideas that I had is, one.

221 00:23:58.180 ⇒ 00:24:09.700 Uttam Kumaran: For the delivery side, we need something around sprint reviews, Client meeting preparation, project reviews.

222 00:24:10.190 ⇒ 00:24:12.189 Uttam Kumaran: So these are weekly or monthly.

223 00:24:12.600 ⇒ 00:24:14.719 Uttam Kumaran: Something to help with grooming?

224 00:24:15.930 ⇒ 00:24:19.529 Uttam Kumaran: And then something to help with stand-up. And so we’ve done the stand-up thing.

225 00:24:20.410 ⇒ 00:24:32.669 Uttam Kumaran: to give you insights, a couple of those, grooming is like, hey, during the weeks, we create new tickets. During grooming is where you have to go back, be like, why did we create this ticket? Who needs it? You basically write all the descriptions in there.

226 00:24:32.670 ⇒ 00:24:40.770 Uttam Kumaran: But that’s all information that’s probably living in a mix of Zoom, Slack, whatever. So there has to be something to help with grooming, which is, like, what are the tickets in our backlog?

227 00:24:40.850 ⇒ 00:24:59.090 Uttam Kumaran: what should be planned for the next two sprints, and then what should be in those things, right? So, something around grooming, project reviews, these could be a mix of, help me create a project review deck, help me create a Gantt chart, help me create the narrative of, like, what I should talk through for the project review. This could be, hey.

228 00:24:59.390 ⇒ 00:25:04.669 Uttam Kumaran: project review isn’t happening, but I still want to send something, like an email or a Slack, write me that.

229 00:25:04.870 ⇒ 00:25:08.319 Uttam Kumaran: Client meeting. So this is just, like, before any client meeting.

230 00:25:08.710 ⇒ 00:25:13.879 Uttam Kumaran: you should be able to tell an agent, I’m having this client meeting, it’s on this topic.

231 00:25:14.040 ⇒ 00:25:18.620 Uttam Kumaran: Help me build an agenda, and as part of that agenda, pull out relevant

232 00:25:18.780 ⇒ 00:25:26.180 Uttam Kumaran: Thanks. For example, I have to go call a client on, they have a question about, like, a data model.

233 00:25:26.590 ⇒ 00:25:31.469 Uttam Kumaran: okay, like, someone on the team may have worked on it 3 weeks ago, I may have worked on that.

234 00:25:31.590 ⇒ 00:25:36.370 Uttam Kumaran: I also may have had questions about it that this is a great opportunity for me to ask those.

235 00:25:37.000 ⇒ 00:25:43.410 Uttam Kumaran: those are all, like, limited by, like, literally the RAM in your brain, which is so unfortunate, because

236 00:25:43.920 ⇒ 00:25:50.919 Uttam Kumaran: Again, like, we may go into that client meeting… the risk is, like, one, a lot of people are taking client meetings unprepared, no agenda.

237 00:25:51.440 ⇒ 00:25:53.030 Uttam Kumaran: That is already F.

238 00:25:53.220 ⇒ 00:25:58.250 Uttam Kumaran: So, every client meeting needs to have an agenda, and people need to be able to walk in prepared.

239 00:25:58.800 ⇒ 00:26:05.950 Uttam Kumaran: Right? So that’s the baseline. Second, after every client meeting, there has to be a set of actions that get taken.

240 00:26:06.490 ⇒ 00:26:19.210 Uttam Kumaran: I would like to see at least that there’s a… there’s a summary, like a follow-up, here’s what we talked about, here’s what you’re working on, we’re working on, and potentially even tickets created, right? And potentially even, like, there’s a couple of their actions.

241 00:26:19.730 ⇒ 00:26:31.199 Uttam Kumaran: all… most of the things in the platform are client meetings. And we’re also doing the… the second part of that, where it’s, like, create the tickets. But similarly to stand-up, we’re not doing the early… you can’t, like, pre-prepare.

242 00:26:31.460 ⇒ 00:26:32.350 Uttam Kumaran: Right?

243 00:26:33.010 ⇒ 00:26:43.789 Uttam Kumaran: So, that’s something. The other piece is, this is all stored in your calendar. Like, in my calendar, I have all of the client meetings I’m going to. They’re all named Brainforge X Client with a topic.

244 00:26:44.240 ⇒ 00:26:54.799 Uttam Kumaran: So, you could naturally suppose that if something had access to my calendar, you could see all the upcoming clients, and you could probably help me in a more proactive way prepare for those.

245 00:26:55.210 ⇒ 00:26:55.970 Uttam Kumaran: Right?

246 00:26:57.290 ⇒ 00:26:58.100 Uttam Kumaran: You know?

247 00:26:58.340 ⇒ 00:27:02.249 Gabriel Lam: So someone… so someone with access to a roadmap, they say, hey.

248 00:27:02.350 ⇒ 00:27:07.340 Uttam Kumaran: Stream in all my calendar events, identify the ones that are client meetings.

249 00:27:07.710 ⇒ 00:27:10.139 Uttam Kumaran: Add those to the platform as…

250 00:27:10.300 ⇒ 00:27:15.640 Uttam Kumaran: future client meetings, so I can go into that and work on preparation.

251 00:27:16.260 ⇒ 00:27:20.359 Uttam Kumaran: As an example of a product that came to… oh, sorry, I don’t want to do too much prescribing, but, like.

252 00:27:20.940 ⇒ 00:27:21.640 Gabriel Lam: As an example.

253 00:27:21.640 ⇒ 00:27:26.649 Uttam Kumaran: We could do, right? That’s logically, we could do with even a lot of the things we have today.

254 00:27:26.840 ⇒ 00:27:27.460 Gabriel Lam: Right.

255 00:27:28.400 ⇒ 00:27:32.539 Uttam Kumaran: And so you get from… you get to AI to move from mostly reactive.

256 00:27:32.650 ⇒ 00:27:39.639 Uttam Kumaran: hey, I want to add… I need help doing this, to, hey, I noticed you are going to do this, here’s what you need.

257 00:27:40.350 ⇒ 00:27:46.170 Uttam Kumaran: This is the real… like, level 3 automation in AI.

258 00:27:46.180 ⇒ 00:27:50.010 Gabriel Lam: Right, and I sort of think about AI very similar to how.

259 00:27:50.010 ⇒ 00:27:54.320 Uttam Kumaran: People think about, self-driving automation.

260 00:27:54.570 ⇒ 00:28:02.290 Uttam Kumaran: There are a couple different levels. I don’t know if you’ve seen… this is a thing I use as a metaphor all the time, so let me know if I’ve already…

261 00:28:02.480 ⇒ 00:28:10.479 Uttam Kumaran: Mentioned this, but… the National Highway Transportation Safety, they have, like, a self-driving, like, levels of automation.

262 00:28:10.780 ⇒ 00:28:15.480 Uttam Kumaran: And I like to use that as an example of…

263 00:28:15.650 ⇒ 00:28:18.680 Uttam Kumaran: Like, how we think about things.

264 00:28:18.830 ⇒ 00:28:23.199 Uttam Kumaran: So, I’m gonna just… I’ll share you this,

265 00:28:23.970 ⇒ 00:28:28.410 Uttam Kumaran: I’ll show you this image in… Zoom.

266 00:28:30.470 ⇒ 00:28:31.400 Gabriel Lam: Mmm.

267 00:28:31.890 ⇒ 00:28:33.530 Gabriel Lam: I have seen this before.

268 00:28:33.530 ⇒ 00:28:35.609 Uttam Kumaran: Yes. So if you’ve… if you’ve…

269 00:28:36.200 ⇒ 00:28:40.490 Uttam Kumaran: Paid attention at all to self-driving or whatever, but, like, this is…

270 00:28:40.720 ⇒ 00:28:45.520 Uttam Kumaran: this is, like, how it works. So, L0 is… is… Hucky-dory.

271 00:28:45.520 ⇒ 00:28:46.410 Gabriel Lam: Oh, yeah.

272 00:28:46.410 ⇒ 00:28:50.589 Uttam Kumaran: L1 is, like, lane assist, it’s, like, now, like, what…

273 00:28:50.820 ⇒ 00:28:56.470 Uttam Kumaran: You know, if you have, like, my car, they didn’t go fancy, so they just have, like, lane assist, or, like, you’re about to crash.

274 00:28:56.580 ⇒ 00:29:05.890 Uttam Kumaran: L2 is like, okay, it’s actually… maybe it’s like, you know how they do cruise control, like, adaptive cruise? That’s like L2, right? So it’s not… it’s not like…

275 00:29:06.050 ⇒ 00:29:11.020 Uttam Kumaran: Cruise where it just stays the same, it’s like, if the car comes within this distance, it starts to slow. L2.

276 00:29:11.340 ⇒ 00:29:14.430 Uttam Kumaran: L3 is, like.

277 00:29:14.610 ⇒ 00:29:16.220 Gabriel Lam: It’s like your Waymo kind of thing.

278 00:29:16.220 ⇒ 00:29:23.380 Uttam Kumaran: No, L3 is almost like the first version of, like, self-driving from Tesla, where it’s able to…

279 00:29:23.580 ⇒ 00:29:29.990 Uttam Kumaran: Full control over steering, but you still have to, like, be there, and you still have to touch it and things like that.

280 00:29:30.170 ⇒ 00:29:32.270 Uttam Kumaran: L4 is…

281 00:29:32.270 ⇒ 00:29:34.749 Gabriel Lam: Sorry, I don’t know if I cut off or you cut off. One second.

282 00:29:34.750 ⇒ 00:29:38.390 Uttam Kumaran: Sorry, sorry. L3 is, like… Oh.

283 00:29:38.570 ⇒ 00:29:39.519 Uttam Kumaran: Can you hear me?

284 00:29:41.390 ⇒ 00:29:43.000 Uttam Kumaran: I think it may be you.

285 00:29:44.460 ⇒ 00:29:45.609 Uttam Kumaran: Hello, hello?

286 00:29:47.410 ⇒ 00:29:48.509 Uttam Kumaran: Hello, hello, hello.

287 00:29:48.510 ⇒ 00:29:49.839 Gabriel Lam: Okay, I’m hearing you now, okay.

288 00:29:49.840 ⇒ 00:29:51.200 Uttam Kumaran: Okay, okay, okay.

289 00:29:52.320 ⇒ 00:29:55.290 Uttam Kumaran: L3… L3 is like Tesla.

290 00:29:55.540 ⇒ 00:29:56.890 Gabriel Lam: Okay. Whole self-driving.

291 00:29:56.890 ⇒ 00:29:58.470 Uttam Kumaran: So you kind of have to be there.

292 00:29:58.710 ⇒ 00:30:02.159 Uttam Kumaran: You could do L4, but they just don’t let you.

293 00:30:02.520 ⇒ 00:30:02.870 Gabriel Lam: Yeah.

294 00:30:02.870 ⇒ 00:30:05.990 Uttam Kumaran: Meaning they ask you to kind of keep your hand there, and things like that.

295 00:30:05.990 ⇒ 00:30:06.630 Gabriel Lam: for regulation.

296 00:30:06.630 ⇒ 00:30:10.929 Uttam Kumaran: Yeah, but as you can see in L4, they say passenger.

297 00:30:11.220 ⇒ 00:30:16.020 Uttam Kumaran: A passenger who would notice can take over driving. L5…

298 00:30:16.140 ⇒ 00:30:21.760 Uttam Kumaran: is, like… so L4 is, like, Waymo. L5 is, like.

299 00:30:21.900 ⇒ 00:30:26.260 Uttam Kumaran: those… the next generation Waymo, where they have two benches facing each other in the car.

300 00:30:26.730 ⇒ 00:30:27.679 Uttam Kumaran: No steering wheel.

301 00:30:28.060 ⇒ 00:30:28.620 Gabriel Lam: Yeah.

302 00:30:28.970 ⇒ 00:30:34.589 Uttam Kumaran: That’s like… so, I would say, in our company, we are between L1 and L2.

303 00:30:34.720 ⇒ 00:30:37.129 Uttam Kumaran: L1 is, like, is everyone using ChatGPT?

304 00:30:37.410 ⇒ 00:30:39.179 Uttam Kumaran: First up. General.

305 00:30:39.330 ⇒ 00:30:41.530 Uttam Kumaran: L2 is, like…

306 00:30:41.670 ⇒ 00:30:47.989 Uttam Kumaran: Okay, are they using ChatGPT, or are we using custom prompts? Are we using GPTs? Maybe we have Slack integrations, right?

307 00:30:48.140 ⇒ 00:31:00.519 Uttam Kumaran: L3 is, like, we’re starting to build our own platform, we’re calling external agents, we’re able to build things, but still, it’s… AI’s not taking a lot of actions on its own.

308 00:31:00.830 ⇒ 00:31:08.109 Uttam Kumaran: L4, L3 and L4 is when you get to… that are kind of described right now in the industry as ambient agents.

309 00:31:08.720 ⇒ 00:31:13.209 Uttam Kumaran: Ambient agents are things that are, like, taking actions on your behalf.

310 00:31:13.650 ⇒ 00:31:19.599 Uttam Kumaran: In a way that is much more on the proactive versus reactive, right?

311 00:31:19.770 ⇒ 00:31:38.040 Uttam Kumaran: And so you can see how, as this goes, it’s actually aligning more towards, like, what you expect an employee. You expect an employee to be working, like, let’s say you hire an employee at Brainforge, and you don’t see them in front of you, but they’re working… people are working on things, right? And things will come up, you’re able to assign them things.

312 00:31:38.050 ⇒ 00:31:43.329 Uttam Kumaran: So it’s actually, like, going that direction. You don’t expect all of your employees, great employees, to, like.

313 00:31:43.350 ⇒ 00:31:48.649 Uttam Kumaran: wait for you to say something to then give it to you, and then the moment you stop talking to them, they, like, can’t.

314 00:31:48.690 ⇒ 00:31:58.869 Uttam Kumaran: do anything. So, like, you can only get work done like Zoom. That’s, like, L2, L1, right? So now… so our AI is kind of still in that zone where it’s purely, like, based on our inputs.

315 00:31:59.040 ⇒ 00:32:01.600 Uttam Kumaran: Yes, there are now some things that we have on schedule.

316 00:32:02.140 ⇒ 00:32:17.979 Uttam Kumaran: So getting towards things that are more on schedule, but on schedule is kind of dumb, right? It’s like, just on a… every Monday, do this thing, right? Or anytime there is this, do this thing. It’s still kind of level 2, level 3. There’s no, like, thinking. There’s no,

317 00:32:18.480 ⇒ 00:32:26.379 Uttam Kumaran: I… is this important to Gabe? Yes, I should alert him about it. Like, there’s no, like, higher level. L4 and L3 is, like.

318 00:32:27.040 ⇒ 00:32:30.459 Uttam Kumaran: Hey, I know Gabe has, like, 2 client meetings coming up.

319 00:32:30.620 ⇒ 00:32:34.840 Uttam Kumaran: I should go ahead and create the client meetings, and I should go ahead and Slack him.

320 00:32:35.330 ⇒ 00:32:38.370 Uttam Kumaran: The links and, like, what to be prepared for.

321 00:32:38.980 ⇒ 00:32:45.620 Uttam Kumaran: Right? That’s now getting into true, like, oh my gosh, that’s like having an executive assistant, basically, right?

322 00:32:45.620 ⇒ 00:32:46.100 Gabriel Lam: Right.

323 00:32:46.100 ⇒ 00:32:55.219 Uttam Kumaran: that’s where, like, I kind of want some of these systems to head, but where… what I kind of want to show you this is all of our tools will start from the left and move to the right.

324 00:32:55.700 ⇒ 00:32:57.250 Gabriel Lam: Stand-up assistant.

325 00:32:57.250 ⇒ 00:33:04.499 Uttam Kumaran: will start as, like, an L2… the user has to go in and do that. But, we have stand-up every day.

326 00:33:05.060 ⇒ 00:33:07.240 Uttam Kumaran: The AI has access to the transcript.

327 00:33:07.570 ⇒ 00:33:13.090 Uttam Kumaran: okay, it can already start to prepare, take notes during, and send summaries, right? We’re, like.

328 00:33:13.260 ⇒ 00:33:16.320 Uttam Kumaran: one more notch to there, so…

329 00:33:16.870 ⇒ 00:33:20.079 Uttam Kumaran: And this is, again, only the way I think about things, so…

330 00:33:20.220 ⇒ 00:33:26.950 Uttam Kumaran: It may or may not be helpful, but I think of each of the areas of our platform as this leveling journey.

331 00:33:27.120 ⇒ 00:33:29.420 Uttam Kumaran: Right? Something starts with a prompt.

332 00:33:29.850 ⇒ 00:33:32.649 Uttam Kumaran: something then starts as, like, a custom GPT.

333 00:33:32.770 ⇒ 00:33:42.259 Uttam Kumaran: Then it’s like a meeting with the UI and features. Then it’s something that runs on a schedule. Then it’s something that proactively does some actions, right?

334 00:33:42.460 ⇒ 00:33:46.520 Uttam Kumaran: Each of the parts of our platform can kind of move in that direction at the time.

335 00:33:46.960 ⇒ 00:33:47.770 Gabriel Lam: Hmm.

336 00:33:50.070 ⇒ 00:33:52.270 Uttam Kumaran: Okay. Did I lose you? You kind of see where I’m.

337 00:33:52.510 ⇒ 00:33:54.409 Gabriel Lam: Yeah, yeah, yeah. Yeah.

338 00:33:54.890 ⇒ 00:33:55.260 Uttam Kumaran: Yeah.

339 00:33:55.260 ⇒ 00:33:55.990 Gabriel Lam: Okay.

340 00:33:56.410 ⇒ 00:34:00.769 Uttam Kumaran: And, like, and for example, self-driving cars are still just inputs and outputs.

341 00:34:00.770 ⇒ 00:34:01.600 Gabriel Lam: Right?

342 00:34:01.600 ⇒ 00:34:03.189 Uttam Kumaran: But what do they have? They have…

343 00:34:03.300 ⇒ 00:34:06.149 Uttam Kumaran: higher level planning. I need to get from here to here.

344 00:34:06.260 ⇒ 00:34:22.029 Uttam Kumaran: And there’s levels of, like, automation. There’s… there is the, hey, there’s a car in front of me. There’s also the, I’m going to Starbucks over here, and I need to constantly, like, remap things. Oh, there’s a… for example.

345 00:34:22.580 ⇒ 00:34:32.829 Uttam Kumaran: This road is closed. Remap. I got the new map, I know how to get there. Road is closed. So they’re doing, like, higher level order thinking at every time, in addition to just…

346 00:34:32.980 ⇒ 00:34:46.769 Uttam Kumaran: I’m about to hit a speed bump, there’s a stop sign. So this is where the sophistication of driving is actually underappreciated, because as humans, we have the map, we’ve seen the signs, we know the speed.

347 00:34:46.909 ⇒ 00:34:50.379 Uttam Kumaran: We are seeing the road in front of us, what’s happening.

348 00:34:50.690 ⇒ 00:34:54.730 Uttam Kumaran: We may also know the weather, we may also know that our road is closed, like…

349 00:34:54.889 ⇒ 00:35:03.619 Uttam Kumaran: we actually… driving is very… it’s a very complicated problem, so you… they have to go these, like, step by step by step. Start by just get to the end of the road.

350 00:35:03.830 ⇒ 00:35:05.250 Uttam Kumaran: Oh, can you do a lap?

351 00:35:05.460 ⇒ 00:35:09.269 Uttam Kumaran: And this is why Google is really advantageous here, because they have all the mapping data.

352 00:35:09.870 ⇒ 00:35:24.340 Uttam Kumaran: And then all their… when you use Google Maps, they send your car driving data to their servers, which informs everybody else’s Google Maps experience. And then what are they? Now the cars are using that, so it’s, like, a great data flywheel for them.

353 00:35:24.960 ⇒ 00:35:33.439 Uttam Kumaran: So this is, like, kind of where I think, similarly to our platform, like, we want to see the, like, the levelings up. They may take a long time, but…

354 00:35:34.220 ⇒ 00:35:41.420 Uttam Kumaran: when you’re building things, you should keep in mind not only that I’m building this feature, but am I building the primitive ways

355 00:35:41.540 ⇒ 00:35:43.460 Uttam Kumaran: For, like, proactive alerting.

356 00:35:43.990 ⇒ 00:35:48.129 Uttam Kumaran: Can that be reused across… Across other features.

357 00:35:48.460 ⇒ 00:35:56.119 Uttam Kumaran: voice. Can that be reused across features? I think that way, when you get to the nth feature, you’re kind of just plugging stuff in.

358 00:35:56.120 ⇒ 00:35:56.500 Gabriel Lam: Yeah.

359 00:35:56.500 ⇒ 00:36:09.170 Uttam Kumaran: it actually gets way… you’re… and what is it in, like, financial terms? Your COGS are going down. You’re getting economies of scale. You’re building, like, modular components that can be reused. That is, like, great.

360 00:36:09.570 ⇒ 00:36:11.250 Uttam Kumaran: engineering, right? Like.

361 00:36:11.400 ⇒ 00:36:20.570 Uttam Kumaran: So, that’s, like, when you push on the team on how they build things, that’s what I would push them on, is to… for them to make sure that they’re not building shit just one-off for, like, stand-up.

362 00:36:20.570 ⇒ 00:36:20.910 Gabriel Lam: I’m…

363 00:36:21.340 ⇒ 00:36:25.619 Uttam Kumaran: That they’re thinking about it, like, hey, tomorrow, if we want to go ship a feature that uses the same data.

364 00:36:26.000 ⇒ 00:36:31.340 Uttam Kumaran: Can… is this, like, purely configured for stand-ups, or, like, can we go do that?

365 00:36:31.780 ⇒ 00:36:32.350 Gabriel Lam: Mmm.

366 00:36:32.350 ⇒ 00:36:35.280 Uttam Kumaran: So that’s, like… so we have stuff around…

367 00:36:35.570 ⇒ 00:36:41.989 Uttam Kumaran: delivery, right? We also have stuff around delivery tasks, which are, like, the actual work.

368 00:36:42.870 ⇒ 00:36:46.610 Uttam Kumaran: So… These are, like, data analysis.

369 00:36:47.140 ⇒ 00:36:51.060 Uttam Kumaran: These are… Putting together architecture diagrams.

370 00:36:51.330 ⇒ 00:36:55.679 Uttam Kumaran: This is using cursor to ship SQL code.

371 00:36:56.140 ⇒ 00:37:07.489 Uttam Kumaran: This is hooking into the MCP servers for, like, Amplitude, Mixpanel, like, some of the tools we use. So this is, like, assisting our engineering delivery staff with their work.

372 00:37:07.910 ⇒ 00:37:08.710 Uttam Kumaran: Right?

373 00:37:09.610 ⇒ 00:37:13.009 Uttam Kumaran: So, what is L1 here? Everyone in the company is using cursor.

374 00:37:13.230 ⇒ 00:37:14.609 Gabriel Lam: Yeah. So that’s, like, L1.

375 00:37:14.610 ⇒ 00:37:20.690 Uttam Kumaran: So I just, I’m like, anytime you’re doing coding, use cursor. L2 is like, well, you’re doing this specific coding task.

376 00:37:20.930 ⇒ 00:37:30.310 Uttam Kumaran: you should use this specific workflow, whether that’s a cursor macro, whether that’s a couple… so there’s, like, some automation to be done there.

377 00:37:30.720 ⇒ 00:37:41.419 Uttam Kumaran: this is where, like, it will take a really good understanding of what the services we sell to each client are, and what are the LEGO blocks we’re building for them, and then looking at

378 00:37:41.660 ⇒ 00:37:45.249 Uttam Kumaran: what can I build? This is… this part is a little bit tough, because

379 00:37:45.840 ⇒ 00:37:58.109 Uttam Kumaran: it may or may not be things that can be done in the platform as we know it. This may be… this may be custom GPTs, this may be internal MCP servers, this may be just, like, better documentation for cursor.

380 00:37:59.480 ⇒ 00:38:05.100 Uttam Kumaran: But the objective here is, like, how do I take the time it takes for us to build X widget for Y client.

381 00:38:05.800 ⇒ 00:38:10.059 Uttam Kumaran: It may take us an hour. How do I accomplish that in 30 minutes?

382 00:38:10.190 ⇒ 00:38:11.240 Gabriel Lam: Right?

383 00:38:11.600 ⇒ 00:38:21.809 Uttam Kumaran: But the… it’s not gonna be… may not necessarily be, like, shipping a feature in the platform. It may be a set of other automations. There’s also things on the sales side, so…

384 00:38:22.390 ⇒ 00:38:36.939 Uttam Kumaran: And this is where, like, Ryan on the sales team has already done a lot of this, and he just has this on his laptop, but, like, helping to write a blog post, helping to write a case study, helping to write white papers, helping to create a new marketing campaign.

385 00:38:37.100 ⇒ 00:38:44.150 Uttam Kumaran: Helping to… This… another sort of meeting-related assistant is, like, preparation for the sales meeting.

386 00:38:45.170 ⇒ 00:38:49.380 Uttam Kumaran: You now know all of our leads that are in progress from HubSpot.

387 00:38:49.490 ⇒ 00:38:58.000 Uttam Kumaran: When we go to the sales meeting, we need to understand what phase of the sale in? How should we be prepared? What did we talk about last time? What are our goals? How do I know objections?

388 00:38:58.140 ⇒ 00:39:01.379 Uttam Kumaran: Everything around, like, enabling the seller.

389 00:39:01.550 ⇒ 00:39:04.000 Uttam Kumaran: to… to advance…

390 00:39:04.160 ⇒ 00:39:12.329 Uttam Kumaran: the lead as many positions as possible at every meeting. We are doing this completely raw and doing well.

391 00:39:13.130 ⇒ 00:39:26.519 Uttam Kumaran: So just imagine how much better is it if it could be if I was going to a meeting and I knew what I talked about last time, because I’m going in and Robin and I are just completely dancing. What is… yes, there’s, like… I’m glad that we are able to do that.

392 00:39:26.680 ⇒ 00:39:29.089 Uttam Kumaran: We cannot hire a salesperson at this point.

393 00:39:29.450 ⇒ 00:39:35.359 Uttam Kumaran: Because there’s so much dancing that we are doing, that there’s nothing that is sort of, like.

394 00:39:35.940 ⇒ 00:39:54.210 Uttam Kumaran: I can’t… it would be so hard to bring someone else into this process, right? And we will have to do that very soon. If we’re gonna move this business… if we’re gonna 10x this business, we’re not… the two of us can’t sell at all the business. In fact, we need to go sell the more complicated million-dollar deal.

395 00:39:54.370 ⇒ 00:39:55.599 Uttam Kumaran: Where it’s, like.

396 00:39:56.010 ⇒ 00:40:09.239 Uttam Kumaran: there’s not much AI that could go… that’s, like, real selling, and then these things, these are, like, 10, 20K deals, is where we need to bring on a seller. But for that seller, okay, what are the series of automations that helps that person get prepared?

397 00:40:09.570 ⇒ 00:40:21.499 Uttam Kumaran: Things like that. Think about what services to offer, what other clients have we worked with that’s similar to them? They mentioned these four or five objections. How do I come prepared to this next meeting to address those objections?

398 00:40:21.800 ⇒ 00:40:23.850 Uttam Kumaran: The nice thing about sales, though.

399 00:40:24.230 ⇒ 00:40:27.230 Uttam Kumaran: Is there’s a shit ton of sales frameworks.

400 00:40:27.340 ⇒ 00:40:34.269 Uttam Kumaran: You don’t have… you can actually really boil this down into, like, this stage, you need to do these actions, and…

401 00:40:34.350 ⇒ 00:40:53.400 Uttam Kumaran: at the biggest sales companies, they’re very prescribed, and the salesperson is just dancing within those rails, right? Right. Okay, here’s how to do objection handling. The worst salespeople are the people that are, like, they just read, like, for example, they just read, like, a book like this, like, The Challenger Sale, and they’re like.

402 00:40:53.740 ⇒ 00:41:03.360 Uttam Kumaran: oh, now it is time for me to address your objection. Less time, like, who will be the ultimate stakeholder for this engagement? And they’re at, like, that’s…

403 00:41:03.470 ⇒ 00:41:06.249 Uttam Kumaran: So this is where, like, people in sales, they end up, like.

404 00:41:06.730 ⇒ 00:41:09.030 Uttam Kumaran: Listening to a framework, and then, like.

405 00:41:09.740 ⇒ 00:41:15.100 Uttam Kumaran: Oh, hi, would you mind articulating your problem for me so then I can sign this opportunity? It’s like, dude.

406 00:41:15.250 ⇒ 00:41:23.239 Uttam Kumaran: I’m coming in there, and I’m saying, like, how’s the weather? Where are you from? Oh, I just visited there. Like, there… I actually…

407 00:41:23.600 ⇒ 00:41:27.299 Uttam Kumaran: I watched a show briefly called Industry on HBO, and one of the.

408 00:41:27.300 ⇒ 00:41:29.170 Gabriel Lam: Oh, I do know that show. Yep.

409 00:41:29.170 ⇒ 00:41:42.599 Uttam Kumaran: Yeah, I stopped watching it after the first season, I thought it was, like, I thought it was just too sexual, and I like… but I like the banking stuff. Yeah, yeah, yeah. But then I thought it got, like, too druggy, I don’t know, I just didn’t… I just wasn’t getting into it, but the one woman on the show, she has a,

410 00:41:42.960 ⇒ 00:41:47.880 Uttam Kumaran: a, hourglass. A minute hourglass.

411 00:41:48.100 ⇒ 00:41:54.990 Uttam Kumaran: On her desk. And she says when she gets on the call, you can’t talk about work until the hourglass is over.

412 00:41:55.410 ⇒ 00:42:00.079 Uttam Kumaran: And I always think about that, because in great client relationship.

413 00:42:00.580 ⇒ 00:42:06.790 Uttam Kumaran: You can’t do that. And so you… for me, I always think about that, so I spend the first few minutes, we don’t talk about work.

414 00:42:07.270 ⇒ 00:42:09.550 Uttam Kumaran: And so those are… but those are things that, like…

415 00:42:09.920 ⇒ 00:42:15.330 Uttam Kumaran: those are principles of how we sell that we can start to integrate into each of these flows, right? So…

416 00:42:15.530 ⇒ 00:42:20.699 Uttam Kumaran: There’s also something around sales follow-ups, And then, the other piece…

417 00:42:20.990 ⇒ 00:42:28.030 Uttam Kumaran: this is sort of probably, like, more workflows is, like, people submitting out-of-office requests. They should probably do that on the platform.

418 00:42:28.250 ⇒ 00:42:31.290 Uttam Kumaran: People want to see what clients are allocated to.

419 00:42:31.490 ⇒ 00:42:33.600 Uttam Kumaran: Right? That should probably be in a platform.

420 00:42:33.730 ⇒ 00:42:37.430 Uttam Kumaran: where people… Google Drive search sucks.

421 00:42:37.630 ⇒ 00:42:41.099 Uttam Kumaran: So hard to find stuff, even though we are pretty organized, still hard.

422 00:42:41.970 ⇒ 00:42:54.889 Uttam Kumaran: we should have some AI that helps people find things that should be in the platform. There should be a My Meeting section, right? My tasks, whatever. So there’s probably something around these, like, smaller things. The last piece is…

423 00:42:55.210 ⇒ 00:43:01.659 Uttam Kumaran: We should probably build agents that help manage our company.

424 00:43:01.980 ⇒ 00:43:07.980 Uttam Kumaran: And that… what I’m saying is, we should probably build a Brainforged finance analyst, sales analyst.

425 00:43:08.290 ⇒ 00:43:11.099 Uttam Kumaran: VM analyst and Ops Analyst Agent.

426 00:43:11.390 ⇒ 00:43:16.420 Uttam Kumaran: That can be used by our operation, by our leadership, to say.

427 00:43:16.600 ⇒ 00:43:21.409 Uttam Kumaran: Hey, what is our mar- what is our highest margin client, and what do we do differently?

428 00:43:22.770 ⇒ 00:43:27.019 Uttam Kumaran: Go… go spend 10 minutes and, like, give me a write-up.

429 00:43:28.130 ⇒ 00:43:28.690 Gabriel Lam: Yeah.

430 00:43:29.010 ⇒ 00:43:35.370 Uttam Kumaran: Right? So these are, like, proactive… these are… these are, like, much more, like, proactive things that…

431 00:43:35.600 ⇒ 00:43:38.500 Uttam Kumaran: like, we have to consider, and, like, this is, like, why…

432 00:43:38.680 ⇒ 00:43:44.569 Uttam Kumaran: for the operators in the company, this is also things that I want to… I want to have, like…

433 00:43:46.440 ⇒ 00:43:56.730 Uttam Kumaran: I want to develop these for our clients as well. Because our clients… this is where, like, truly, we don’t have the time or the operations staff to ask some of these really, like, high-level questions that I want to ask.

434 00:43:56.850 ⇒ 00:43:57.730 Uttam Kumaran: But, like.

435 00:43:57.890 ⇒ 00:44:03.620 Uttam Kumaran: Okay, but if AI agent has access to all our docs, all of our Clockify hours, all of our lived here tickets, all of the meetings.

436 00:44:04.330 ⇒ 00:44:08.770 Uttam Kumaran: I’m pretty sure it’ll give me a better answer than any other human in the company, you know?

437 00:44:09.070 ⇒ 00:44:09.670 Gabriel Lam: Right.

438 00:44:10.400 ⇒ 00:44:16.009 Uttam Kumaran: So… That’s sort of, like… what I’m thinking, so…

439 00:44:16.200 ⇒ 00:44:17.040 Gabriel Lam: Hmm.

440 00:44:17.040 ⇒ 00:44:20.580 Uttam Kumaran: That’s, like, that’s all I decided to write.

441 00:44:21.870 ⇒ 00:44:22.220 Uttam Kumaran: I think…

442 00:44:22.220 ⇒ 00:44:24.149 Gabriel Lam: Even there, there’s a lot.

443 00:44:24.150 ⇒ 00:44:27.310 Uttam Kumaran: Yeah, so as you can tell, there is a lot of roadmap even there.

444 00:44:27.310 ⇒ 00:44:28.050 Gabriel Lam: Yeah.

445 00:44:28.240 ⇒ 00:44:29.970 Gabriel Lam: But I do think it is…

446 00:44:30.950 ⇒ 00:44:35.289 Gabriel Lam: I agree with you in the sense that I think the speed will be…

447 00:44:35.740 ⇒ 00:44:39.410 Gabriel Lam: like, exponential to some degree, right? It’s just, like, there’s gonna be, like.

448 00:44:43.710 ⇒ 00:44:47.780 Gabriel Lam: For example, even though those Asians, There’s gonna be a,

449 00:44:48.050 ⇒ 00:44:52.510 Gabriel Lam: refining, a refining process and making sure the prompts are right, but I think once…

450 00:44:53.820 ⇒ 00:44:59.680 Gabriel Lam: we hit that threshold, I think things will really… accelerate.

451 00:44:59.860 ⇒ 00:45:00.590 Gabriel Lam: Once we’re there.

452 00:45:00.590 ⇒ 00:45:01.180 Uttam Kumaran: Okay.

453 00:45:01.340 ⇒ 00:45:08.700 Uttam Kumaran: Yeah, but you can tell some of these ideas, like, dude, I could… if… at this point, like, your job is to…

454 00:45:09.230 ⇒ 00:45:17.449 Uttam Kumaran: Like, I want you to focus on the toughest ones. The things where you’re like, dude, I don’t even know if this is, like, physically possible.

455 00:45:17.950 ⇒ 00:45:19.509 Uttam Kumaran: That’s, like, what I need.

456 00:45:19.630 ⇒ 00:45:22.779 Uttam Kumaran: like, the people in the company most enabled, but, like, for example.

457 00:45:22.780 ⇒ 00:45:23.240 Gabriel Lam: True.

458 00:45:23.240 ⇒ 00:45:24.710 Uttam Kumaran: The white paper writer.

459 00:45:24.860 ⇒ 00:45:31.110 Uttam Kumaran: That’s something that you should be able to hand… you should be able to hand to Hannah and say, like, here are the steps for you to create this product in the.

460 00:45:31.110 ⇒ 00:45:31.870 Gabriel Lam: Awesome.

461 00:45:33.200 ⇒ 00:45:33.800 Uttam Kumaran: Right.

462 00:45:34.420 ⇒ 00:45:39.459 Uttam Kumaran: But the engineering around that itself is something that only you probably can work on.

463 00:45:39.460 ⇒ 00:45:40.560 Gabriel Lam: Right, right.

464 00:45:40.560 ⇒ 00:45:42.310 Uttam Kumaran: There’s… I don’t think there’s anyone else

465 00:45:42.700 ⇒ 00:45:44.299 Uttam Kumaran: Yeah, it’s given to me or you?

466 00:45:44.520 ⇒ 00:45:45.650 Gabriel Lam: Right. So…

467 00:45:46.230 ⇒ 00:45:51.170 Uttam Kumaran: those are the two things to balance, and I don’t know whether it helps to maybe do…

468 00:45:51.320 ⇒ 00:45:57.360 Uttam Kumaran: one week here, one week here, one week here, one week here, or like, I don’t know, I’ll let you kind of think about how to organize around it.

469 00:45:57.690 ⇒ 00:46:03.700 Uttam Kumaran: I’m also more than happy to give priorities around, like, these ideas, but…

470 00:46:05.340 ⇒ 00:46:12.819 Uttam Kumaran: I don’t know, for me, my mind goes to, like, we have to build a platform and build on the platform at the same time.

471 00:46:13.000 ⇒ 00:46:13.620 Gabriel Lam: Right.

472 00:46:13.990 ⇒ 00:46:24.010 Uttam Kumaran: what I… what I’m curious about from your side is, given our current resources, What is possible?

473 00:46:24.970 ⇒ 00:46:31.839 Uttam Kumaran: That way, I can go ahead and think, okay, what is the revenue impact if we were to unlock

474 00:46:32.450 ⇒ 00:46:33.919 Uttam Kumaran: 3 or 4 of these.

475 00:46:34.220 ⇒ 00:46:38.959 Uttam Kumaran: What does Gabe need from a staffing perspective, or my time perspective?

476 00:46:39.150 ⇒ 00:46:40.800 Uttam Kumaran: So there’s some bucket of money.

477 00:46:40.870 ⇒ 00:46:42.949 Gabriel Lam: Yeah. And then I can go say.

478 00:46:43.340 ⇒ 00:46:45.410 Uttam Kumaran: Cool, here’s what I can do for you.

479 00:46:45.550 ⇒ 00:46:55.889 Uttam Kumaran: Right? So in that way, this is how I work… this is how I work for you. In that if you’re like, well, I need 40 more hours from somebody decent, I don’t need, like, a Sam, but I need, like, somebody.

480 00:46:56.500 ⇒ 00:46:59.630 Uttam Kumaran: And I need, like, 2 or 3 hours with you.

481 00:47:00.110 ⇒ 00:47:08.789 Uttam Kumaran: That’s a price. So then I go and I look at everything we’re selling, I’m like, cool, we just sold 3 deals, we’re gonna allocate that money, Gabe needs this.

482 00:47:09.670 ⇒ 00:47:11.829 Uttam Kumaran: Let me find what I can do.

483 00:47:13.980 ⇒ 00:47:20.830 Uttam Kumaran: So that’s ultimately where… just because… today’s constrains are not tomorrow’s constraints. Right. But…

484 00:47:21.190 ⇒ 00:47:28.170 Uttam Kumaran: We are moving to a model where we want people in the company to feel like they own a budget, and the budget is…

485 00:47:28.330 ⇒ 00:47:30.719 Uttam Kumaran: Is time and money.

486 00:47:30.980 ⇒ 00:47:39.729 Uttam Kumaran: Yeah. Like, and so, if you have an understanding of what you have access to now, it is kind of going up and down as we loop more people into clients.

487 00:47:40.010 ⇒ 00:47:52.389 Uttam Kumaran: for… it’s… what I’m looking for is that you’re able to say, hey, I know I have access to this, and we can get this done, but if you can get me a one full-time sort of person that just hustles on AI stuff.

488 00:47:52.920 ⇒ 00:47:58.400 Uttam Kumaran: And… I can… I’m… I’m fairly… and again, none of this is, like, do or die, like…

489 00:47:58.560 ⇒ 00:48:03.210 Uttam Kumaran: but you’re fairly confident that you can get XYZ done, That is, like.

490 00:48:03.540 ⇒ 00:48:08.340 Uttam Kumaran: for me, really motivates me to be like, okay, damn, I’m gonna find someone for Gabe.

491 00:48:08.470 ⇒ 00:48:16.809 Uttam Kumaran: And I’m gonna… instead of moving some of this budget… because we have 5 other competing areas to allocate budgets towards, but then I’ll say, okay.

492 00:48:17.060 ⇒ 00:48:19.199 Uttam Kumaran: It truly, at my level, will be like.

493 00:48:19.370 ⇒ 00:48:26.660 Uttam Kumaran: do we need to get another project manager, or can I allocate this budget to Gabe’s team? Or here? Yeah. And, like, and we will look at that

494 00:48:26.700 ⇒ 00:48:40.150 Uttam Kumaran: the way we look at that is, like, the pain and the opportunity. So if truly you’re like, I can deliver these three things, and if some of those unlocks actually just allows us to either get someone more junior or not have to hire somewhere else.

495 00:48:40.350 ⇒ 00:48:44.899 Uttam Kumaran: It is a no-brainer for me to give you the budget, right?

496 00:48:44.900 ⇒ 00:48:45.630 Gabriel Lam: Right.

497 00:48:45.630 ⇒ 00:48:54.460 Uttam Kumaran: Because if I’m gonna hire someone to do some of the things that you’re gonna automate, and that person has to be really senior, instead I can get a junior person with your tools, and they’re gonna operate at that level.

498 00:48:54.460 ⇒ 00:48:55.130 Gabriel Lam: Right, yeah.

499 00:48:55.130 ⇒ 00:48:59.030 Uttam Kumaran: Or, if your tool unlocks a couple other people to operate at a higher level.

500 00:48:59.160 ⇒ 00:49:01.049 Uttam Kumaran: That’s the math that I’m sort of doing.

501 00:49:01.050 ⇒ 00:49:01.660 Gabriel Lam: Right.

502 00:49:01.660 ⇒ 00:49:02.229 Uttam Kumaran: You know.

503 00:49:02.550 ⇒ 00:49:03.320 Gabriel Lam: Yeah.

504 00:49:05.570 ⇒ 00:49:08.980 Gabriel Lam: Okay, I see where this is going. I think there’s, like, a…

505 00:49:11.250 ⇒ 00:49:16.919 Gabriel Lam: like, this sort of building block thing that I’m thinking through of, like, okay, what are the lowest hanging things that we can get out?

506 00:49:17.130 ⇒ 00:49:21.550 Gabriel Lam: And how… like, what are… how do we build the building blocks so we can then build the rest of it out?

507 00:49:22.030 ⇒ 00:49:28.979 Gabriel Lam: It’s tough, because you could toil in building the platform and never get adoption, or you can ship a bunch of stuff fast.

508 00:49:28.980 ⇒ 00:49:32.409 Uttam Kumaran: But then it, like, doesn’t work well for the future.

509 00:49:32.510 ⇒ 00:49:40.690 Uttam Kumaran: That is your… this is your problem to build the… to think about the optimization. Yeah, yeah. Right? But I’m aware of the problem, so I know that, but…

510 00:49:40.880 ⇒ 00:49:45.530 Uttam Kumaran: I think also, I don’t mind moving slow.

511 00:49:46.030 ⇒ 00:49:50.030 Uttam Kumaran: But what I don’t mind… I don’t… I don’t mind moving slow if it’s intentional.

512 00:49:50.210 ⇒ 00:49:50.810 Gabriel Lam: Yeah.

513 00:49:50.810 ⇒ 00:49:51.400 Uttam Kumaran: Right?

514 00:49:51.750 ⇒ 00:50:01.450 Uttam Kumaran: we had no principled way of speed, so I naturally was, like, one week. And so if I set the deadline, I know that things will occur within the bounds of the deadline.

515 00:50:01.840 ⇒ 00:50:06.530 Uttam Kumaran: But that is just me taking, like, a machete approach to, like.

516 00:50:06.820 ⇒ 00:50:12.889 Uttam Kumaran: velocity, right? Yeah. Like, it’s not very, like, sophisticated. It’s just, like, you have a week, ship the shit in a week.

517 00:50:13.030 ⇒ 00:50:16.769 Uttam Kumaran: So, we can do something more sophisticated, like.

518 00:50:17.310 ⇒ 00:50:22.009 Uttam Kumaran: 3 weeks of shipping, 1 week on platform, 2 and 2, like, we can do things that way.

519 00:50:22.140 ⇒ 00:50:28.779 Uttam Kumaran: Additionally, I think one part that should be underarching of this whole thing is security.

520 00:50:29.380 ⇒ 00:50:30.779 Uttam Kumaran: and adoption.

521 00:50:30.950 ⇒ 00:50:35.499 Uttam Kumaran: Security is, like, at every step of the way, how are we developing these features? This will get…

522 00:50:35.680 ⇒ 00:50:41.970 Uttam Kumaran: more and more important as we grow, so it’s something that, like, I… I want to make sure is on your mind that, like.

523 00:50:42.070 ⇒ 00:50:53.320 Uttam Kumaran: people should have access to the things they only need access to. Some of this stuff is very sensitive. We have obviously not done a good job at, like, restricting things to date, so there is some stuff to do there. Second.

524 00:50:53.590 ⇒ 00:50:55.110 Uttam Kumaran: is adoption.

525 00:50:55.240 ⇒ 00:51:02.039 Uttam Kumaran: So, who is using it? Who are my power users? Who are people that aren’t using it? Are there things for them to even use?

526 00:51:02.310 ⇒ 00:51:04.689 Uttam Kumaran: And so, they’re not using it, there’s nothing there, like…

527 00:51:04.940 ⇒ 00:51:10.190 Uttam Kumaran: There should be some part of our weekly conversation that is around bike.

528 00:51:10.190 ⇒ 00:51:10.520 Gabriel Lam: Yeah.

529 00:51:10.520 ⇒ 00:51:11.720 Uttam Kumaran: The adoption, you know?

530 00:51:12.300 ⇒ 00:51:17.969 Gabriel Lam: Which I think at this stage, is the biggest thing, right? I am with you in the security thing. I think that.

531 00:51:17.970 ⇒ 00:51:19.609 Uttam Kumaran: No, no, no, I agree, I agree, I agree.

532 00:51:19.610 ⇒ 00:51:24.300 Gabriel Lam: Yeah, like, in a big company, like, that’s all it is, really. But,

533 00:51:24.760 ⇒ 00:51:26.290 Gabriel Lam: I feel like at this point.

534 00:51:27.370 ⇒ 00:51:36.870 Gabriel Lam: getting people adopted into the workflow then leaves them open to being like, okay, I switched from, you know…

535 00:51:37.340 ⇒ 00:51:42.359 Gabriel Lam: doing it all raw by myself to, like, this sort of MVP v0 version.

536 00:51:42.960 ⇒ 00:51:43.500 Uttam Kumaran: You’re gonna build a.

537 00:51:43.500 ⇒ 00:51:43.899 Gabriel Lam: That’s important.

538 00:51:43.900 ⇒ 00:51:53.339 Uttam Kumaran: by the way, they’re gonna be pushing you, and they’re gonna love… and the next… and people are gonna… you’re gonna see, there’s gonna be a lot more people joining the company. The nth person joining

539 00:51:53.560 ⇒ 00:52:07.600 Uttam Kumaran: on day one, they should get told to go check out the platform and to start… and to start using things, right? So, you’re totally right. And so this is where, if you’re like, hey, we need to do, like, an adoption-focused sprint, then we can do that too. Yeah, so… you have all the…

540 00:52:08.140 ⇒ 00:52:11.500 Uttam Kumaran: You have the curse of, like, optionality here.

541 00:52:11.930 ⇒ 00:52:16.190 Uttam Kumaran: But what I won’t succumb to is not being ambitious. Like, not being, like.

542 00:52:16.190 ⇒ 00:52:16.670 Gabriel Lam: Right.

543 00:52:16.670 ⇒ 00:52:19.339 Uttam Kumaran: Holy shit, we’re gonna go after this thing, right?

544 00:52:19.340 ⇒ 00:52:19.950 Gabriel Lam: Yeah.

545 00:52:20.160 ⇒ 00:52:27.829 Uttam Kumaran: So, that’s, like, more of, like, I can… I will actually… this is where, for me, I’m better utilized at the…

546 00:52:28.070 ⇒ 00:52:33.580 Uttam Kumaran: Like, the vision level, But I had to… I previously was just playing both, so…

547 00:52:33.580 ⇒ 00:52:33.920 Gabriel Lam: Hmm.

548 00:52:33.920 ⇒ 00:52:40.129 Uttam Kumaran: it’s almost like too many, I have to make too many trade-offs, or there’s all these, like… you don’t want me thinking about constraints.

549 00:52:40.480 ⇒ 00:52:44.749 Uttam Kumaran: Because then I’m gonna start to, like, limit, like, what I think is possible.

550 00:52:44.750 ⇒ 00:52:45.340 Gabriel Lam: Right.

551 00:52:45.710 ⇒ 00:52:55.879 Uttam Kumaran: But that’s also why I need to leave the constraint world and be like, okay, things are happening. Now, think about where do we want to be from an AI perspective, like.

552 00:52:56.610 ⇒ 00:52:58.390 Uttam Kumaran: End of 2026.

553 00:52:58.700 ⇒ 00:53:07.519 Uttam Kumaran: And then I can go… I can go spend time on that, you know? And I can go use all my brainpower to think about how do we accomplish

554 00:53:08.330 ⇒ 00:53:09.630 Uttam Kumaran: How do we accomplish that?

555 00:53:15.970 ⇒ 00:53:16.540 Gabriel Lam: Alright.

556 00:53:16.540 ⇒ 00:53:19.660 Uttam Kumaran: Okay, this was the conversation I wanted to have, so yeah, that was everything.

557 00:53:19.660 ⇒ 00:53:31.109 Gabriel Lam: That’s perfect. I feel like I have a lot to be like, alright, I think I’m sort of with you. I’m like, okay, I can see this happening, what needs to get there, and I’m sort of excited to see what it can become.

558 00:53:31.520 ⇒ 00:53:38.939 Uttam Kumaran: Cool, and I will continue to show you examples of, like… I mean, none of this… none of the ideas I’ve had

559 00:53:38.990 ⇒ 00:53:56.030 Uttam Kumaran: I would say that some of these are new, but they’re built on me researching other companies, like, doing pieces of it. So I’ll start to share with you some of those. I have, like, a long list of, like, where I first read about the ambient agent thing, and a few other. So you’ll see, like, I’m not, like, just pulling these out of my ass, like, a lot of these…

560 00:53:56.030 ⇒ 00:53:56.640 Gabriel Lam: I feel like…

561 00:53:56.640 ⇒ 00:54:01.860 Uttam Kumaran: are possible, but it will help for you to see, like, some of the inspiration, but, like.

562 00:54:02.170 ⇒ 00:54:08.640 Uttam Kumaran: I’m going and talking to other leaders of consultancies, I’m talking to other business leaders, and…

563 00:54:09.440 ⇒ 00:54:14.709 Uttam Kumaran: One, nobody’s thinking like this. Second, the enterprise value of Brainforge

564 00:54:15.170 ⇒ 00:54:19.150 Uttam Kumaran: true IP development is what will boost our enterprise value.

565 00:54:19.650 ⇒ 00:54:26.129 Uttam Kumaran: Right? We… enterprise value in a company can be dependent on just our revenues and things like that, but if we’re talking about, like.

566 00:54:26.540 ⇒ 00:54:30.179 Uttam Kumaran: true ownership, it’s our playbooks, and it’s our IP.

567 00:54:30.420 ⇒ 00:54:37.750 Uttam Kumaran: Our company won’t get value based on people, and our revenue has a multiple, so we can’t really fight that.

568 00:54:37.930 ⇒ 00:54:54.109 Uttam Kumaran: So this is very, very important. And the second piece is, we are also selling this. So the… what we’re discussing here, our clients are two… probably a year or two from us being able to sell this, because they’re just not even, like, at this point. We’re selling, like, train your staff on ChatGBT.

569 00:54:54.520 ⇒ 00:54:55.250 Gabriel Lam: Yeah.

570 00:54:55.250 ⇒ 00:54:59.100 Uttam Kumaran: You know, but I know that, like, if we figure this out, this is, like…

571 00:54:59.420 ⇒ 00:55:03.759 Uttam Kumaran: million dollar, how do you come in and do complete AI transformation from the.

572 00:55:04.160 ⇒ 00:55:09.919 Uttam Kumaran: Right? So that’s what I… that’s what we want to be able to sell next year, and so we are dogfooding that, you know?

573 00:55:09.920 ⇒ 00:55:10.590 Gabriel Lam: Okay.

574 00:55:11.250 ⇒ 00:55:12.350 Uttam Kumaran: Yeah. Yeah.

575 00:55:13.220 ⇒ 00:55:15.519 Gabriel Lam: Alright, I… yes, I see it.

576 00:55:15.520 ⇒ 00:55:17.050 Uttam Kumaran: Okay, okay.

577 00:55:17.050 ⇒ 00:55:17.970 Gabriel Lam: I’m down.

578 00:55:18.730 ⇒ 00:55:19.420 Gabriel Lam: Okay.

579 00:55:19.420 ⇒ 00:55:19.960 Uttam Kumaran: Cool.

580 00:55:20.320 ⇒ 00:55:20.970 Gabriel Lam: Okay.

581 00:55:21.330 ⇒ 00:55:29.839 Gabriel Lam: Yeah, I think… I think maybe for next… or the rest of the day, I think there’s some stuff for the team to figure out with the linear tickets, but I do think…

582 00:55:31.430 ⇒ 00:55:39.209 Gabriel Lam: like, Mustafa and Sam are able to handle a lot of the updates, I’m probably gonna start thinking about next week and…

583 00:55:39.350 ⇒ 00:55:41.959 Gabriel Lam: There’s a sort of migration issue that we’re working through.

584 00:55:41.960 ⇒ 00:55:42.820 Uttam Kumaran: Yeah, yeah, yeah.

585 00:55:42.820 ⇒ 00:55:45.809 Gabriel Lam: Which might just be a… a temporary…

586 00:55:46.090 ⇒ 00:55:55.799 Gabriel Lam: like, one of those building blocks that we’re building out for the future? You should. I feel like I’m, like, a broken record, but you should use the Gantt chart stuff, it helps for you to visualize things. Yeah, yeah.

587 00:55:55.800 ⇒ 00:55:56.730 Uttam Kumaran: Yeah.

588 00:55:57.000 ⇒ 00:55:59.119 Gabriel Lam: I’m more of a FigJam guy, just…

589 00:55:59.120 ⇒ 00:55:59.619 Uttam Kumaran: Okay, then you…

590 00:55:59.620 ⇒ 00:56:02.680 Gabriel Lam: The gang charts have been a hit or miss for me, because…

591 00:56:02.680 ⇒ 00:56:03.310 Uttam Kumaran: Okay.

592 00:56:03.570 ⇒ 00:56:06.309 Gabriel Lam: I come from a world where Gantt charts are never… no, this is.

593 00:56:06.310 ⇒ 00:56:06.870 Uttam Kumaran: Oh, no.

594 00:56:06.870 ⇒ 00:56:10.370 Gabriel Lam: Gantt charts are, like, always changing, and I’m like, oh, man.

595 00:56:10.640 ⇒ 00:56:12.259 Uttam Kumaran: Oh, okay, I see, I see, I see.

596 00:56:12.260 ⇒ 00:56:19.060 Gabriel Lam: So… But yeah, I think… I think even if next week, if we can get people to start

597 00:56:19.260 ⇒ 00:56:30.740 Gabriel Lam: thinking about, like, oh, what are features… almost like a… a help box be like, these are some things that we’re facing, it also would be good. It’s also a cultural thing to get people to start talking about.

598 00:56:31.510 ⇒ 00:56:34.670 Gabriel Lam: But I’m excited for what’s to come, for sure.

599 00:56:35.200 ⇒ 00:56:35.800 Uttam Kumaran: Okay.

600 00:56:36.780 ⇒ 00:56:41.460 Uttam Kumaran: Okay. Cool. Thanks, this is great. I, like, I, like, love talking about this stuff, like.

601 00:56:41.460 ⇒ 00:56:42.220 Gabriel Lam: I see.

602 00:56:42.220 ⇒ 00:56:43.970 Uttam Kumaran: Like, I don’t get to talk to anybody.

603 00:56:44.340 ⇒ 00:56:46.250 Gabriel Lam: It’s tough to talk about it in our.

604 00:56:46.250 ⇒ 00:56:57.890 Uttam Kumaran: company, and let alone other companies. Like, I’ve talked to some people, they’re like, dude, you’re on, like, a different planet. I’m like, you have no fucking clue what we’re doing, pretty much.

605 00:56:58.740 ⇒ 00:57:01.340 Uttam Kumaran: And it’s funny, because I’ll even talk to Robert, and he’s like.

606 00:57:01.340 ⇒ 00:57:03.099 Gabriel Lam: I know you have a ton of meetings, I think I…

607 00:57:03.310 ⇒ 00:57:06.160 Gabriel Lam: I think we ran over one of your meetings, so I want to make sure.

608 00:57:06.160 ⇒ 00:57:06.599 Uttam Kumaran: Yeah, yeah, yeah.

609 00:57:06.600 ⇒ 00:57:08.249 Gabriel Lam: You’re able to get to what you need to do.

610 00:57:08.250 ⇒ 00:57:10.089 Uttam Kumaran: Okay, all good. Okay, alright.

611 00:57:10.090 ⇒ 00:57:12.640 Gabriel Lam: Oh, thank you. You cut off again, it might be me, sorry.

612 00:57:12.890 ⇒ 00:57:14.169 Uttam Kumaran: Sorry, can you hear me?

613 00:57:17.350 ⇒ 00:57:18.250 Uttam Kumaran: Hello, hello, hello.

614 00:57:18.250 ⇒ 00:57:21.459 Gabriel Lam: Okay, there we… we’re back. I don’t know what to do my Wi-Fi today.

615 00:57:21.460 ⇒ 00:57:24.030 Uttam Kumaran: No, no, I’ll let you go, I’ll let you go.

616 00:57:24.030 ⇒ 00:57:25.319 Gabriel Lam: Alright, I’ll catch you later.

617 00:57:25.470 ⇒ 00:57:26.629 Uttam Kumaran: Alright, thanks, dude. Bye.

618 00:57:26.630 ⇒ 00:57:27.880 Gabriel Lam: Thanks. Have a good one.