Meeting Title: Uttam - Robert AI Enablement Date: 2026-03-16 Meeting participants: Robert Tseng, Uttam Kumaran
WEBVTT
1 00:00:11.710 ⇒ 00:00:12.630 Robert Tseng: Hey, dude.
2 00:00:15.320 ⇒ 00:00:16.600 Uttam Kumaran: Yo.
3 00:00:16.710 ⇒ 00:00:17.670 Robert Tseng: Hey.
4 00:00:18.940 ⇒ 00:00:21.309 Robert Tseng: Sorry, that one ran out a little long, too.
5 00:00:21.750 ⇒ 00:00:22.610 Uttam Kumaran: Can you hear me?
6 00:00:22.970 ⇒ 00:00:25.860 Robert Tseng: Yeah, yeah, I can hear you. Very interesting.
7 00:00:25.860 ⇒ 00:00:27.659 Uttam Kumaran: No, you’re okay, you’re okay.
8 00:00:28.660 ⇒ 00:00:30.760 Robert Tseng: I just talked to a guy.
9 00:00:31.010 ⇒ 00:00:31.640 Uttam Kumaran: Okay.
10 00:00:31.960 ⇒ 00:00:35.610 Uttam Kumaran: My… What was it?
11 00:00:38.130 ⇒ 00:00:43.359 Robert Tseng: I’m not sure if you’re, like, 10 seconds behind what I’m saying, or if you’re hearing what I’m saying now.
12 00:00:49.810 ⇒ 00:00:52.770 Robert Tseng: I feel like you’re, like, a few seconds behind.
13 00:00:55.650 ⇒ 00:00:57.100 Robert Tseng: your audio…
14 00:00:57.100 ⇒ 00:00:57.770 Uttam Kumaran: I’m…
15 00:00:58.340 ⇒ 00:00:59.379 Robert Tseng: Okay, this family.
16 00:01:08.350 ⇒ 00:01:10.940 Uttam Kumaran: Okay, yeah, sorry, this coffee shop.
17 00:01:11.220 ⇒ 00:01:14.430 Uttam Kumaran: Okay, you’re saying you met a guy…
18 00:01:15.970 ⇒ 00:01:18.759 Robert Tseng: Yeah, well, so, because of the mix… well…
19 00:01:22.100 ⇒ 00:01:27.799 Robert Tseng: this guy, Eden brought… we were competing with him and Eden for the Mixpanel live work.
20 00:01:28.230 ⇒ 00:01:37.879 Robert Tseng: you know, and I pretty much had Greg at the start of Q1 try to, like, kind of just get stuff from this audit and just kind of stay on track so we don’t lose that work stream.
21 00:01:40.320 ⇒ 00:01:48.169 Robert Tseng: you know, at some point, I was like, why do we need to compete with him? Maybe I’ll just talk to him, see if he wants to come and, like, partner with us or something. So…
22 00:01:48.330 ⇒ 00:01:56.070 Robert Tseng: It’s very interesting. Apparently, he is the only North American, subcontractor for Mixed Panel Enterprise.
23 00:01:57.360 ⇒ 00:02:03.899 Robert Tseng: So, like, he’s been only doing pure mixed panel consulting since 2022, just been in the game for a while.
24 00:02:04.090 ⇒ 00:02:13.800 Robert Tseng: The deal sizes are nothing to sneeze at. It’s kind of like the same ceiling we ran into, like, maybe, like, 10K, 10K a month deals. And
25 00:02:14.490 ⇒ 00:02:20.190 Robert Tseng: Like, that I ran into before we started working together, so… but, like, I’m like, huh, here…
26 00:02:20.350 ⇒ 00:02:25.319 Robert Tseng: And then he’s like, he’s trying to get out of the game, so,
27 00:02:25.450 ⇒ 00:02:28.140 Robert Tseng: I was like, well, I mean, this guy… Or does that mean why?
28 00:02:29.390 ⇒ 00:02:36.310 Robert Tseng: Well, he doesn’t want to only be doing product analytics. He wants to be doing, like, AI transformation and, like…
29 00:02:36.420 ⇒ 00:02:38.839 Robert Tseng: However, I’m like, oh, okay, interesting, yeah.
30 00:02:39.260 ⇒ 00:02:39.940 Robert Tseng: So…
31 00:02:41.530 ⇒ 00:02:54.850 Robert Tseng: But, you know, if he has a book of Mixed Found Enterprise clients, and we can, like… I mean, we can… you can think about maybe buying them out, or, like, taking… taking him in to just do the service leave of product analytics. That could move Greg out of…
32 00:02:54.960 ⇒ 00:03:12.040 Robert Tseng: And this guy seems pretty proven, like, to have, like, built out that relationship with them, and, you know, he has this expertise, whatever, so… and he’s interested in, like, kind of AI engineering work we do, and, you know, so… I’m like, huh, maybe, maybe worth talking to. So,
33 00:03:12.700 ⇒ 00:03:29.750 Robert Tseng: Yeah, I could just bring in a few extra logos, like, pretty quickly, and then we can do our thing of, like, growing those accounts to go and win other business, because he just, like, he just does purely product analytics. They ask for anything else, he says, no, I can’t support it, and it’s just, like, a one-man… one-man show kind of thing, so…
34 00:03:29.860 ⇒ 00:03:31.640 Robert Tseng: Yeah, I gotta…
35 00:03:31.640 ⇒ 00:03:33.429 Uttam Kumaran: Oh, see, there’s no… he doesn’t have any team.
36 00:03:34.260 ⇒ 00:03:39.670 Robert Tseng: Oh, I mean, he has, like, some subcontractors or whatever, but he never, like, built the organization. He just kind of, like.
37 00:03:40.040 ⇒ 00:03:45.059 Robert Tseng: Figured out how to, like, increase his number of hours to do the same scope.
38 00:03:47.740 ⇒ 00:03:48.430 Uttam Kumaran: I mean…
39 00:03:52.070 ⇒ 00:03:55.510 Uttam Kumaran: Yeah, she do teaching work with us.
40 00:03:56.180 ⇒ 00:03:58.659 Robert Tseng: I was like, dude, he should just come in and do…
41 00:03:58.660 ⇒ 00:04:01.330 Uttam Kumaran: You should just own product analytics, yeah, yeah.
42 00:04:01.330 ⇒ 00:04:10.530 Robert Tseng: Yeah, give us… give us his business, and then we’ll grow his accounts, like, whatever, like, or something like that. Like, I feel like that could be… that could be an interesting way to just, like.
43 00:04:11.590 ⇒ 00:04:15.730 Robert Tseng: Add… add more, add more, and…
44 00:04:16.010 ⇒ 00:04:20.880 Robert Tseng: just… just to get a, like, a book of business. Like, I never really thought about that as, like, a… as a…
45 00:04:21.279 ⇒ 00:04:23.430 Robert Tseng: As a possible thing that we could do right now.
46 00:04:23.430 ⇒ 00:04:27.130 Uttam Kumaran: So he’s gonna wind down, and then what’s… what’s… what is his plan?
47 00:04:28.410 ⇒ 00:04:31.489 Robert Tseng: He was just gonna try to go back into the market, like…
48 00:04:31.810 ⇒ 00:04:33.559 Uttam Kumaran: No way. Oh.
49 00:04:33.560 ⇒ 00:04:49.100 Robert Tseng: Yeah, he tried to sell the business in December, it didn’t work out. Someone tried to, like, basically, yeah, buy his book of business from him, but then they wanted him to stay on as a, like, a minority owner and just continue to subcontract the work. And he was like.
50 00:04:49.430 ⇒ 00:04:55.599 Robert Tseng: why would I do that? So, I just, like, it seems like the deal that he was offered was just not very compelling.
51 00:04:57.290 ⇒ 00:05:03.659 Robert Tseng: So, I don’t necessarily think I want to buy him out, but, like, I’d be like, hey, you should, you know, maybe we can stabilize.
52 00:05:03.660 ⇒ 00:05:10.360 Uttam Kumaran: I mean, yeah, basically we should see, like, yeah, if he has any operational stuff immediately that gets absorbed by us.
53 00:05:10.710 ⇒ 00:05:13.169 Uttam Kumaran: He no longer has to do anything on recruiting.
54 00:05:13.500 ⇒ 00:05:14.540 Uttam Kumaran: And…
55 00:05:16.040 ⇒ 00:05:26.950 Robert Tseng: Sounds like he’s just tired of the grind. Just give him a salary, and then if we can just take his accounts, and then, like, if he… whatever that ends up being, and then we can just try to grow those, like, I think that’s an interesting strategy, yeah.
56 00:05:26.950 ⇒ 00:05:29.019 Uttam Kumaran: Yeah, and he could get a long tail of those.
57 00:05:29.470 ⇒ 00:05:30.170 Robert Tseng: Yeah.
58 00:05:30.170 ⇒ 00:05:36.870 Uttam Kumaran: We’re a bigger piece of those. We take up all the recruiting and, like, all the operations and stuff like that.
59 00:05:37.050 ⇒ 00:05:39.439 Uttam Kumaran: And then he gets the opportunity to come
60 00:05:39.980 ⇒ 00:05:42.640 Uttam Kumaran: do also amplitude stuff, or whatever.
61 00:05:43.220 ⇒ 00:05:50.799 Robert Tseng: Yeah. He says he gets 5… 5 to 10 leads a month from… from Amplitude. I’m sorry, sorry.
62 00:05:50.800 ⇒ 00:05:51.330 Uttam Kumaran: We’re mixed fan.
63 00:05:51.790 ⇒ 00:05:52.540 Robert Tseng: Yeah.
64 00:05:53.060 ⇒ 00:05:53.780 Robert Tseng: So…
65 00:05:54.900 ⇒ 00:05:55.620 Uttam Kumaran: No, no.
66 00:05:58.110 ⇒ 00:06:02.159 Uttam Kumaran: Nice, that’s so funny.
67 00:06:02.920 ⇒ 00:06:11.629 Uttam Kumaran: Yeah, you totally should. I mean, we should… we should… we should do that. Like, that’s a great opportunity. That’s sort of the person we were sort of hoping would consider us, you know?
68 00:06:12.170 ⇒ 00:06:12.930 Robert Tseng: Yeah.
69 00:06:16.120 ⇒ 00:06:19.420 Uttam Kumaran: Yeah, one of my thoughts, is, like…
70 00:06:19.530 ⇒ 00:06:23.410 Uttam Kumaran: I was talking to Pranav today, I was, like, thinking about, like, okay, how do we…
71 00:06:24.130 ⇒ 00:06:26.400 Uttam Kumaran: How do we promote people to…
72 00:06:26.510 ⇒ 00:06:29.170 Uttam Kumaran: Start to think about, like, wanting to own a book.
73 00:06:29.940 ⇒ 00:06:36.920 Uttam Kumaran: And I was talking to them about, like, incentives, and I was like, okay, we’re gonna maybe probably set CSO incentives that are related to, like, margin.
74 00:06:37.800 ⇒ 00:06:45.910 Uttam Kumaran: Which is one thing. And then I said the service leads incentives are gonna be more around, like, incidents, quality, and, like, speed, maybe.
75 00:06:46.520 ⇒ 00:06:51.630 Uttam Kumaran: But then I was like, okay, if… maybe he hits a margin target, but then how do I…
76 00:06:51.780 ⇒ 00:07:00.269 Uttam Kumaran: I was like, I don’t really know about doing, like, one large project and two smalls, like, I was like, did you figure that… you figure that out?
77 00:07:00.540 ⇒ 00:07:06.350 Uttam Kumaran: But maybe we do a… we do an incentive based on the amount of revenue that you’re a CSO on.
78 00:07:07.240 ⇒ 00:07:12.270 Uttam Kumaran: And then, naturally, the only way up is to build a portfolio.
79 00:07:12.660 ⇒ 00:07:13.470 Uttam Kumaran: Right?
80 00:07:13.960 ⇒ 00:07:16.539 Uttam Kumaran: Yeah. So, like, maybe that guy comes in.
81 00:07:16.920 ⇒ 00:07:21.689 Uttam Kumaran: Yeah, we take over a lot of the management. Maybe we give them a sweet deal of, like.
82 00:07:21.930 ⇒ 00:07:23.330 Uttam Kumaran: A portion of that.
83 00:07:23.500 ⇒ 00:07:25.730 Uttam Kumaran: But then, we also give them a path towards
84 00:07:26.000 ⇒ 00:07:28.069 Uttam Kumaran: Building, like, a bigger book of business.
85 00:07:28.320 ⇒ 00:07:40.549 Uttam Kumaran: Like, for Pranava, I was like, look, if you start to manage, like… he’s like, yeah, I’m just managing this. I’m like, yeah, but you’ll get two or three. What’s gonna happen? He’s like, yeah, I’m just gonna be kind of full. I’m like, yeah, so your job is to now be, like, now CSOs should come work under you.
86 00:07:40.690 ⇒ 00:07:43.220 Uttam Kumaran: And you manage a portfolio of clients.
87 00:07:43.390 ⇒ 00:07:49.950 Uttam Kumaran: And I’m like, okay, maybe the incentive, there should be incentive on, like, The amount of revenue… Managed?
88 00:07:50.060 ⇒ 00:07:51.620 Uttam Kumaran: As well as the margin.
89 00:07:52.900 ⇒ 00:07:53.390 Robert Tseng: Yeah.
90 00:07:53.390 ⇒ 00:08:03.209 Uttam Kumaran: And so I was like, maybe it’s, like, from 0 to 30% margin, there’s no bonus. From 30 to 50, there’s something. And then from 50 onwards, there’s something.
91 00:08:03.690 ⇒ 00:08:04.390 Robert Tseng: Yeah.
92 00:08:04.390 ⇒ 00:08:09.110 Uttam Kumaran: Because I think we… without… I don’t… I don’t think we should incentivize anything less than 30 now.
93 00:08:09.330 ⇒ 00:08:10.390 Uttam Kumaran: So… Yep.
94 00:08:10.670 ⇒ 00:08:16.630 Uttam Kumaran: Then, it’s like the service lead’s job, I’m like, okay, I’m gonna incentivize based on, like, like…
95 00:08:17.200 ⇒ 00:08:19.679 Uttam Kumaran: I was, like, trying to think about, okay, we could do, like.
96 00:08:19.860 ⇒ 00:08:24.620 Uttam Kumaran: We could do, like, percentage… I was thinking of two… couple things. One, I was like.
97 00:08:24.760 ⇒ 00:08:33.919 Uttam Kumaran: I wanted to think about quality, so, like, number of issues. Second piece, I was, like, adherence to forecasts. Meaning, if you think… if you say a product is gonna take X long.
98 00:08:34.610 ⇒ 00:08:37.689 Uttam Kumaran: like… You shouldn’t be consistently off.
99 00:08:38.090 ⇒ 00:08:38.919 Uttam Kumaran: Right.
100 00:08:40.000 ⇒ 00:08:42.229 Uttam Kumaran: Yeah. Like, so there was something around, like.
101 00:08:42.390 ⇒ 00:08:57.770 Uttam Kumaran: timeline adherence, so it pushes them to set realistic timelines, versus the CSO will always try to push faster, faster, cheaper, cheaper, right? So, and then the last piece, I was like, maybe we should do, like, percent of tickets
102 00:08:58.170 ⇒ 00:08:59.619 Uttam Kumaran: hand it to AI.
103 00:08:59.830 ⇒ 00:09:05.790 Uttam Kumaran: And I was like, I don’t often… I was like, should we be really on the nose? But I was with the AI stuff, because I’m like.
104 00:09:05.980 ⇒ 00:09:11.369 Uttam Kumaran: I think just some service leads are gonna, do things the normal way.
105 00:09:11.520 ⇒ 00:09:16.249 Uttam Kumaran: And I’m like, I’m almost like, you should… we should set an AI target that’s, like.
106 00:09:16.630 ⇒ 00:09:20.110 Uttam Kumaran: 30% of tickets are completely handled by… completely handled by,
107 00:09:20.580 ⇒ 00:09:29.259 Uttam Kumaran: But that way, there’s, like, no avoiding it, because otherwise people… you still could just set better forecasts, do well, but then to get beyond 50%, like…
108 00:09:29.710 ⇒ 00:09:36.030 Uttam Kumaran: maybe we just… I want to say, like, yeah, for projects, 30% of your stuff needs to go through AI.
109 00:09:36.990 ⇒ 00:09:43.059 Uttam Kumaran: And that way, it’s like, you can’t game the number of tickets, no matter what, 30% of them have to get handed off the AI.
110 00:09:44.160 ⇒ 00:09:45.819 Uttam Kumaran: Or something like that. And so…
111 00:09:47.730 ⇒ 00:09:54.800 Uttam Kumaran: Yeah, I was talking to Pranav about it, but in this situation, like, yeah, I mean, that guy could come in and be like, he now comes in with, like, a…
112 00:09:55.510 ⇒ 00:10:02.600 Uttam Kumaran: I don’t know, half a million dollar whatever, or 250K, whatever. Maybe because he’s blocking in with it, we can give him…
113 00:10:02.990 ⇒ 00:10:05.849 Uttam Kumaran: there’s, like, a better structure for him. Of course, he has…
114 00:10:05.850 ⇒ 00:10:06.600 Robert Tseng: Yeah.
115 00:10:06.600 ⇒ 00:10:11.669 Uttam Kumaran: If you’re telling me that he was literally gonna quit, then, like, I feel like we have a lot of leverage.
116 00:10:12.330 ⇒ 00:10:12.890 Robert Tseng: Yeah.
117 00:10:12.890 ⇒ 00:10:17.669 Uttam Kumaran: Versus, like, if he was still gonna keep going, then I don’t… so I don’t think we would have to do, like, dollar to dollar.
118 00:10:18.010 ⇒ 00:10:25.560 Uttam Kumaran: But I also want to know what his incentive is, like, is his incentive, like, he’s burning out, and so he wants to be on a team? Is it that, like, is it the AI piece?
119 00:10:25.680 ⇒ 00:10:27.800 Uttam Kumaran: And then we just need to, like, frame it.
120 00:10:28.550 ⇒ 00:10:41.469 Uttam Kumaran: And then also, of course, we can take on all the operational recruiting stuff, he joins a team, and then we still, I think, we try to do right by him and give him a piece of the business that he brings. Yeah.
121 00:10:42.490 ⇒ 00:10:44.530 Uttam Kumaran: Like, yeah, I don’t know.
122 00:10:46.840 ⇒ 00:10:56.350 Robert Tseng: Yeah, like, I think this is interesting… this is an interesting situation. We haven’t had somebody who’s… who has a book of business, and he’s just trying to wind… he’s trying to wind it down, or they hand it off.
123 00:10:56.350 ⇒ 00:11:02.179 Uttam Kumaran: I wonder if you just do the math on it, like, if we were gonna hire someone like that, I wonder what the salary would be.
124 00:11:02.870 ⇒ 00:11:05.779 Uttam Kumaran: And maybe it ends up being pretty even.
125 00:11:06.440 ⇒ 00:11:12.559 Robert Tseng: I think it would just… he was ready to go in-house. It’d be like… he’d be like a Greg Koss person, is kind of what I would get.
126 00:11:13.740 ⇒ 00:11:17.970 Uttam Kumaran: Yeah, and we have work. I mean, we have… we have active stuff for him to do.
127 00:11:18.590 ⇒ 00:11:19.190 Robert Tseng: Yeah.
128 00:11:20.220 ⇒ 00:11:27.089 Robert Tseng: And if he’s interested in, like, expanding, kind of basically, like, doing what you say, like, he wants something.
129 00:11:27.090 ⇒ 00:11:31.689 Uttam Kumaran: Well, it just depends which angle you want to go through. Yeah, yeah, exactly.
130 00:11:32.220 ⇒ 00:11:33.389 Robert Tseng: And he wants to, like, basically.
131 00:11:33.390 ⇒ 00:11:44.679 Uttam Kumaran: Because the thing is, like, again, I can teach… if he can manage projects and do that, I’ll teach all the AI stuff. Like, it sort of depends on if he wants to continue to, like, drive revenue, if he wants to actually, like, go
132 00:11:44.820 ⇒ 00:11:46.160 Uttam Kumaran: more technical.
133 00:11:47.090 ⇒ 00:11:47.530 Robert Tseng: Yeah.
134 00:11:47.530 ⇒ 00:11:48.200 Uttam Kumaran: Yeah.
135 00:11:48.570 ⇒ 00:11:57.989 Robert Tseng: And he… I think it sounds like he wants to apply AI to his… to his work. He wants to learn how to use AI in product analytics, but he just has not had the…
136 00:11:58.220 ⇒ 00:12:06.589 Robert Tseng: ability to build out the infrastructure, so that he can actually do that. He’s just really just trading time for money right now, right? So… Yeah.
137 00:12:07.120 ⇒ 00:12:07.870 Uttam Kumaran: Yeah.
138 00:12:10.740 ⇒ 00:12:13.069 Uttam Kumaran: Yeah, and again, we get subcontractors.
139 00:12:15.080 ⇒ 00:12:15.610 Robert Tseng: Yeah.
140 00:12:18.180 ⇒ 00:12:19.540 Uttam Kumaran: Yeah, that would be sick.
141 00:12:20.400 ⇒ 00:12:23.810 Uttam Kumaran: We already needed to hire more people for product analytics stuff, so…
142 00:12:23.810 ⇒ 00:12:24.570 Robert Tseng: Sign up.
143 00:12:24.570 ⇒ 00:12:28.659 Uttam Kumaran: If you cost it out and it’s similar, even if it’s more, I feel like it’s worth it.
144 00:12:29.720 ⇒ 00:12:36.960 Robert Tseng: I’ll introduce you over email. Yeah, I’m gonna try to get a meeting in person again next week, and, like, I want to keep that conversation going.
145 00:12:36.960 ⇒ 00:12:37.279 Uttam Kumaran: I love you, man.
146 00:12:37.310 ⇒ 00:12:37.790 Robert Tseng: Yeah.
147 00:12:37.790 ⇒ 00:12:38.420 Uttam Kumaran: person.
148 00:12:39.070 ⇒ 00:12:42.050 Robert Tseng: No, I literally met with him, like, right before this call, so…
149 00:12:42.050 ⇒ 00:12:44.680 Uttam Kumaran: Oh, like, somewhere in New York. Oh, on Zoom, on Zoom, okay, thanks.
150 00:12:44.680 ⇒ 00:12:49.250 Robert Tseng: Yeah, I didn’t know he was in New York. I thought he was in LA, but he’s actually in New York, yeah.
151 00:12:49.250 ⇒ 00:12:49.990 Uttam Kumaran: Sick.
152 00:12:50.450 ⇒ 00:12:51.060 Robert Tseng: Yeah.
153 00:12:51.520 ⇒ 00:12:52.330 Uttam Kumaran: Nice.
154 00:12:53.440 ⇒ 00:12:57.390 Uttam Kumaran: Yeah, I mean, dude, she could just go run those relationships, too, like…
155 00:13:02.660 ⇒ 00:13:08.700 Uttam Kumaran: Yeah, dude, I mean, honestly, for you too, dude, think about the cost of, acquiring those customers.
156 00:13:09.450 ⇒ 00:13:13.820 Robert Tseng: Dude, I know, just a walking book of business is so appealing to me right now.
157 00:13:16.390 ⇒ 00:13:17.780 Uttam Kumaran: Yeah, yeah.
158 00:13:18.150 ⇒ 00:13:19.640 Uttam Kumaran: Yeah, you should ask him.
159 00:13:20.400 ⇒ 00:13:27.159 Robert Tseng: That would… that would… that would put… that would push us up, because like… like you said, the ops… the ops costs don’t really scale, we just need more top-line.
160 00:13:27.160 ⇒ 00:13:27.730 Uttam Kumaran: Oh, yeah.
161 00:13:27.730 ⇒ 00:13:35.040 Robert Tseng: operating margins will be fine. So, if we can just get, like, if we could just walk in with 5, like, that would… that would help a lot.
162 00:13:35.600 ⇒ 00:13:41.049 Uttam Kumaran: I mean, was your… if your conversation with him was pretty candid, then I… if I talk to him, I’m probably gonna be like…
163 00:13:41.780 ⇒ 00:13:43.270 Uttam Kumaran: Tell me what you would need.
164 00:13:44.180 ⇒ 00:13:45.870 Uttam Kumaran: And, yeah…
165 00:13:45.870 ⇒ 00:13:47.170 Robert Tseng: I think that’s the angle you should take.
166 00:13:47.170 ⇒ 00:13:47.770 Uttam Kumaran: Thank you.
167 00:13:48.340 ⇒ 00:13:48.980 Robert Tseng: Yeah.
168 00:13:48.980 ⇒ 00:13:57.250 Uttam Kumaran: Yeah, because I want to be, like, I want to give you, like, a landing pad that you feel is, like, really, really clutch. Going back in the industry is not… don’t do that.
169 00:13:57.410 ⇒ 00:14:03.570 Uttam Kumaran: like… But also, I’m… we both know how hard this thing is, so it’s not a surprise.
170 00:14:04.400 ⇒ 00:14:14.279 Uttam Kumaran: I mean, dude, this was originally, like, kind of when I first started talking to Clarence, he was like, you wanna… you want a model like that, so we could think about having a bigger piece of, like, future revenue or something.
171 00:14:15.080 ⇒ 00:14:16.350 Uttam Kumaran: And then…
172 00:14:17.300 ⇒ 00:14:22.310 Uttam Kumaran: Yeah, I mean, he could easily come in and be like, oh shit, like, you guys actually, like, know what you’re doing.
173 00:14:23.050 ⇒ 00:14:23.590 Robert Tseng: Yeah.
174 00:14:23.740 ⇒ 00:14:25.940 Uttam Kumaran: And we would take a… we’d take a big load off.
175 00:14:28.520 ⇒ 00:14:31.290 Robert Tseng: What’s the risk, do you think, of something like this?
176 00:14:32.500 ⇒ 00:14:38.059 Uttam Kumaran: The risk is that we don’t do… we don’t nail the transition, like, those clients are.
177 00:14:38.060 ⇒ 00:14:38.459 Robert Tseng: Married to him.
178 00:14:38.460 ⇒ 00:14:39.020 Uttam Kumaran: M.
179 00:14:39.600 ⇒ 00:14:40.399 Robert Tseng: Are they trying to risk.
180 00:14:40.400 ⇒ 00:14:48.319 Uttam Kumaran: is that, like, I don’t know any of these subcontractors, so maybe they, like, I don’t know, yeah, whatever.
181 00:14:48.770 ⇒ 00:14:51.740 Uttam Kumaran: The risk is also that, like.
182 00:14:52.270 ⇒ 00:14:55.079 Uttam Kumaran: He comes in and wants to run the business.
183 00:14:55.620 ⇒ 00:14:58.170 Uttam Kumaran: Like, yeah, I get… I mean, I hear.
184 00:14:58.170 ⇒ 00:15:00.199 Robert Tseng: I don’t think they will. Yeah, for sure.
185 00:15:03.900 ⇒ 00:15:10.369 Uttam Kumaran: The risk is that, like, the cost, he says, no, I actually want you guys to buy me, like, that’s not gonna be a rev share thing.
186 00:15:10.890 ⇒ 00:15:19.520 Uttam Kumaran: But other than that, like, he’s already gonna die, then… Yeah.
187 00:15:25.430 ⇒ 00:15:30.030 Uttam Kumaran: He’s already gonna quit, and it’s, like, no-brainer for him. We just have to think about what makes it…
188 00:15:30.270 ⇒ 00:15:32.040 Uttam Kumaran: I mean, like, yeah.
189 00:15:32.240 ⇒ 00:15:36.760 Uttam Kumaran: If he already admitted to you that he’s gonna kill it, then I was like, he shouldn’t have done that.
190 00:15:37.450 ⇒ 00:15:39.329 Uttam Kumaran: Yeah. Basically, is what I’m saying.
191 00:15:39.650 ⇒ 00:15:40.470 Robert Tseng: Yeah.
192 00:15:41.700 ⇒ 00:15:48.199 Uttam Kumaran: But that may show that he’s actually, like, serious, and we can just do right by him and make it a solid switch.
193 00:15:50.550 ⇒ 00:15:56.909 Uttam Kumaran: Yeah, I think the risk is really on the, like, house… is that revenue actually, like, married to him?
194 00:15:57.190 ⇒ 00:16:00.370 Uttam Kumaran: Or can we peel… can we peel it off?
195 00:16:00.690 ⇒ 00:16:05.470 Uttam Kumaran: like, is it gonna be mesh culturally? Like, I’ll figure that out, but…
196 00:16:06.070 ⇒ 00:16:06.630 Robert Tseng: Yeah.
197 00:16:07.840 ⇒ 00:16:08.630 Robert Tseng: Okay.
198 00:16:11.150 ⇒ 00:16:13.300 Robert Tseng: Yeah, well, let’s just see where it goes. I’ll make that.
199 00:16:13.300 ⇒ 00:16:14.089 Uttam Kumaran: I know some crap.
200 00:16:14.820 ⇒ 00:16:15.660 Uttam Kumaran: Okay.
201 00:16:17.050 ⇒ 00:16:24.680 Uttam Kumaran: Yeah, what else today? Oh yeah, so, I mean, dude, I really wanted to just try and, like.
202 00:16:24.830 ⇒ 00:16:27.870 Uttam Kumaran: I wanted to see how much you were actually able to do.
203 00:16:27.870 ⇒ 00:16:31.160 Robert Tseng: Yeah, show me your superpowers, yeah, yeah.
204 00:16:31.160 ⇒ 00:16:35.659 Uttam Kumaran: Kinda like, yeah, let me see…
205 00:16:37.280 ⇒ 00:16:43.209 Uttam Kumaran: I’m gonna join in the Zoom, but I’m not… I’m just gonna stay on my phone, let’s just see if it join works.
206 00:17:15.010 ⇒ 00:17:23.939 Uttam Kumaran: Okay… So, I mean, a couple of things that I’m doing that’s, like.
207 00:17:24.619 ⇒ 00:17:27.670 Uttam Kumaran: So, first of all, I think,
208 00:17:28.500 ⇒ 00:17:32.319 Uttam Kumaran: Trying to see where I can start. So, have you been using any of the, like, skills at all?
209 00:17:33.420 ⇒ 00:17:36.380 Robert Tseng: I just used a couple.
210 00:17:36.530 ⇒ 00:17:49.590 Robert Tseng: Today, because, B helped me set up, the MCPs. Like, I didn’t have it. Like, my root directory was not reinforced platform, so I didn’t really actually have access to the skills yet.
211 00:17:50.110 ⇒ 00:17:51.800 Uttam Kumaran: Okay, cool, so… Yeah.
212 00:17:52.000 ⇒ 00:17:53.540 Uttam Kumaran: I think,
213 00:17:54.360 ⇒ 00:18:04.600 Uttam Kumaran: Let’s just go there now. I think the first thing is, like, I want to kind of show you how the skills live, and I think you’re gonna get the biggest bang for your buck by just actually just, like.
214 00:18:04.720 ⇒ 00:18:06.459 Uttam Kumaran: Developing more of them.
215 00:18:06.890 ⇒ 00:18:10.769 Uttam Kumaran: And so, let me open this,
216 00:18:13.040 ⇒ 00:18:22.200 Uttam Kumaran: So, yeah, as you know, like, everything is in this, like, top-level… platform folder,
217 00:18:22.410 ⇒ 00:18:31.009 Uttam Kumaran: We’ve done a great job at developing a lot of them, and so, yeah, you can just go ahead and hit slash and see a lot of them. Common ones that I use…
218 00:18:31.250 ⇒ 00:18:34.680 Uttam Kumaran: I use this, like, granola…
219 00:18:35.490 ⇒ 00:18:43.469 Uttam Kumaran: this, like, synced to Granola, so it’ll pull all your granola meetings that aren’t in the vaults, and then create a PR with them.
220 00:18:43.610 ⇒ 00:19:02.399 Uttam Kumaran: A lot… another one I use is, like, this EP audit, which basically looks at, like, the client that you mentioned and just checks whether all the tickets are there. Another one I use very common is this document counsel. So, for example, I’ll say, like, run document counsel on a plan in…
221 00:19:02.970 ⇒ 00:19:04.010 Uttam Kumaran: Plans.
222 00:19:13.690 ⇒ 00:19:18.740 Uttam Kumaran: And Document Council is nice because, these days I don’t…
223 00:19:18.990 ⇒ 00:19:24.220 Uttam Kumaran: I don’t… it actually just, like, helps you take a document from, like, B to, like, an A.
224 00:19:24.710 ⇒ 00:19:26.669 Uttam Kumaran: And the way it works is it’ll…
225 00:19:26.830 ⇒ 00:19:28.729 Uttam Kumaran: Right now, it’s picking a plan.
226 00:19:29.850 ⇒ 00:19:32.800 Uttam Kumaran: And it’ll go find one.
227 00:19:33.140 ⇒ 00:19:41.730 Uttam Kumaran: Brainforge website to… website to… website migration. It’s gonna pick a bunch of,
228 00:19:41.860 ⇒ 00:19:44.360 Uttam Kumaran: As you can see, it’s gonna pick a bunch of personas.
229 00:19:44.630 ⇒ 00:19:47.100 Uttam Kumaran: So it looked, it picked me.
230 00:19:47.300 ⇒ 00:19:49.099 Uttam Kumaran: I picked a delivery lead.
231 00:19:49.200 ⇒ 00:19:52.729 Uttam Kumaran: I picked ops, I picked implementer, and I picked a stakeholder.
232 00:19:53.490 ⇒ 00:19:59.700 Uttam Kumaran: And it’s gonna run all of those things as separate agents, and actually just, like, give you a take.
233 00:20:00.140 ⇒ 00:20:01.860 Uttam Kumaran: So it’s like, here’s my take.
234 00:20:06.850 ⇒ 00:20:08.659 Uttam Kumaran: How do you know 8 weeks is enough?
235 00:20:09.180 ⇒ 00:20:12.469 Uttam Kumaran: I don’t know if SEO measurable gains is specific enough.
236 00:20:13.590 ⇒ 00:20:28.619 Uttam Kumaran: Delivery lead says revise. So I think it’s just helpful, because I think a lot… oftentimes when I’m writing something, I just need to see, like, a hundred ideas, like, about where the edges are on something, and then I can pick and be like, okay, that’s right, okay, that’s right.
237 00:20:30.630 ⇒ 00:20:41.169 Uttam Kumaran: And so, for example, like, I… so, this is probably… this is an internal plan, but Document Council is really helpful, for that. The other thing is, dude, nowadays, I also, you can…
238 00:20:41.310 ⇒ 00:20:42.509 Uttam Kumaran: So that’s just, like.
239 00:20:43.100 ⇒ 00:20:49.150 Uttam Kumaran: running this on any type of documents. I think what we’re gonna do is I’m actually gonna set up
240 00:20:49.500 ⇒ 00:20:53.589 Uttam Kumaran: For all of our clients, like, the people that are in that
241 00:20:53.950 ⇒ 00:21:06.879 Uttam Kumaran: on the client, as the personas. So, like, it’ll show, like, what is… what is Shivani’s take on something, things like that, because we have enough, I think, material from all of them. So that’s something that’s been helpful for me.
242 00:21:07.030 ⇒ 00:21:13.840 Uttam Kumaran: The other thing I’ve been doing is, like, I’ve… right now, the Google Workspace,
243 00:21:14.120 ⇒ 00:21:21.859 Uttam Kumaran: thing is working really well, so you can… like, I now create calendar invites directly from chat. I can drop.
244 00:21:21.860 ⇒ 00:21:22.280 Robert Tseng: I don’t know.
245 00:21:22.280 ⇒ 00:21:23.700 Uttam Kumaran: directly from chat.
246 00:21:23.980 ⇒ 00:21:29.059 Uttam Kumaran: So, I don’t have to leave, like, at all.
247 00:21:29.210 ⇒ 00:21:37.069 Uttam Kumaran: So, I’ll be constantly trying to push as much through this as possible. There’s still some hiccups, like the Google Calendar doesn’t attach to Zoom.
248 00:21:37.300 ⇒ 00:21:39.380 Uttam Kumaran: But the emails works perfectly.
249 00:21:39.710 ⇒ 00:21:42.940 Uttam Kumaran: So whether it’s, like, look for an email, or draft an email.
250 00:21:43.050 ⇒ 00:21:45.180 Uttam Kumaran: It works really, really well.
251 00:21:45.320 ⇒ 00:21:47.980 Uttam Kumaran: So if that’s something.
252 00:21:47.980 ⇒ 00:21:48.500 Robert Tseng: Oh, yeah.
253 00:21:48.840 ⇒ 00:21:49.550 Robert Tseng: this now?
254 00:21:50.200 ⇒ 00:21:55.169 Uttam Kumaran: No, it actually just goes and calls an API and finds emails, so it’s not, like, in here.
255 00:21:55.170 ⇒ 00:21:56.140 Robert Tseng: So, yeah.
256 00:21:56.140 ⇒ 00:21:56.720 Uttam Kumaran: Yeah.
257 00:21:57.930 ⇒ 00:22:01.030 Uttam Kumaran: You can also completely interact with HubSpot.
258 00:22:01.250 ⇒ 00:22:02.430 Uttam Kumaran: whether it’s.
259 00:22:02.430 ⇒ 00:22:02.870 Robert Tseng: Great.
260 00:22:02.870 ⇒ 00:22:04.150 Uttam Kumaran: Or reading things.
261 00:22:04.150 ⇒ 00:22:06.350 Robert Tseng: I didn’t see that in the MCP, like, the.
262 00:22:06.350 ⇒ 00:22:11.790 Uttam Kumaran: Yeah, there’s a… if you tell it to look and help you set it up, it’ll… it’ll do that.
263 00:22:12.110 ⇒ 00:22:19.029 Uttam Kumaran: Okay. Yeah, so you can do a lot of that. We automated the data platform spreadsheet.
264 00:22:20.920 ⇒ 00:22:21.540 Robert Tseng: Great.
265 00:22:21.890 ⇒ 00:22:26.940 Uttam Kumaran: So, there’s a whole skill here for, like, Backfilling it and updating it.
266 00:22:27.130 ⇒ 00:22:31.510 Uttam Kumaran: Which is really great.
267 00:22:32.550 ⇒ 00:22:37.630 Uttam Kumaran: So, like, I feel like we made a lot of progress on planning and linear and things like that.
268 00:22:39.960 ⇒ 00:22:43.839 Robert Tseng: I want to, like, test, like, migrating my, like, campaign brief.
269 00:22:44.620 ⇒ 00:22:46.510 Robert Tseng: thing to…
270 00:22:46.780 ⇒ 00:22:51.849 Robert Tseng: Like, turning it into a skill. Do you think I should do that, or should I just keep.
271 00:22:51.850 ⇒ 00:23:00.260 Uttam Kumaran: Yeah, yes, yes. And actually, you should just say exactly that. You should say, like, help me turn this into a skill, and there are… there’s a ton of, like.
272 00:23:00.920 ⇒ 00:23:05.050 Uttam Kumaran: I don’t think we have a new skill creator skill. Maybe we do.
273 00:23:05.330 ⇒ 00:23:11.470 Uttam Kumaran: But Cursor is smart enough, it’ll go figure it out, and it’ll help you do that.
274 00:23:12.040 ⇒ 00:23:13.150 Robert Tseng: Okay, great.
275 00:23:13.150 ⇒ 00:23:18.010 Uttam Kumaran: So, we have things around, like, meeting prep, too. So, like, if you want to prep for a meeting.
276 00:23:18.130 ⇒ 00:23:20.120 Uttam Kumaran: There’s a lot of that.
277 00:23:20.230 ⇒ 00:23:26.629 Uttam Kumaran: The other thing that I’ve been doing in Cursor these days is actually don’t even, like… I don’t even have this open anymore.
278 00:23:26.910 ⇒ 00:23:33.190 Uttam Kumaran: So I completely go, and I, and I go to my editor, I go to my appearance, and I turn this off.
279 00:23:33.490 ⇒ 00:23:35.409 Uttam Kumaran: And I basically have been, like.
280 00:23:35.900 ⇒ 00:23:38.430 Uttam Kumaran: You should’ve seen on Saturday, I was like…
281 00:23:38.690 ⇒ 00:23:41.770 Uttam Kumaran: I was basically running, like,
282 00:23:45.570 ⇒ 00:23:47.309 Uttam Kumaran: Let’s see if I can actually…
283 00:23:47.450 ⇒ 00:23:52.630 Uttam Kumaran: Like, I was just running, like, 5 of these… 5 columns in a row,
284 00:23:53.610 ⇒ 00:24:01.280 Uttam Kumaran: And, like, for example, I’ll add another one here, this is…
285 00:24:02.580 ⇒ 00:24:04.670 Uttam Kumaran: These are all, like, cloud agents.
286 00:24:09.430 ⇒ 00:24:13.389 Uttam Kumaran: That way, because I’m oftentimes, like, working on 3 or 4 things.
287 00:24:13.550 ⇒ 00:24:17.690 Uttam Kumaran: While something’s running, like, you want to just trigger another item.
288 00:24:17.860 ⇒ 00:24:22.230 Uttam Kumaran: And so, my… most of my layout has, like, looked like this.
289 00:24:22.380 ⇒ 00:24:30.470 Uttam Kumaran: more than, like… I don’t do a lot in the file system now, because Purser will automatically put it where it needs to go.
290 00:24:30.610 ⇒ 00:24:33.449 Uttam Kumaran: So, a lot of times, I’m just interacting in the session.
291 00:24:33.740 ⇒ 00:24:40.300 Uttam Kumaran: Whether it’s, like, creating an SOW, drafting a strategy thing, working on some code,
292 00:24:40.510 ⇒ 00:24:42.810 Uttam Kumaran: I’m oftentimes just doing it like this.
293 00:24:43.070 ⇒ 00:24:52.789 Uttam Kumaran: Which is really nice. The other piece that I wanted to share is actually, like, how I’ve been interacting a lot with Linear. So, for example, I…
294 00:24:53.250 ⇒ 00:25:01.449 Uttam Kumaran: Now, my workflow is more about creating all the appropriate linear tickets for… and creating the appropriate project.
295 00:25:01.620 ⇒ 00:25:05.369 Uttam Kumaran: And then actually indicating to me what parts I can assign out to AI.
296 00:25:05.490 ⇒ 00:25:08.829 Uttam Kumaran: So, for example, if I show you, like.
297 00:25:09.220 ⇒ 00:25:12.360 Uttam Kumaran: like, one of the projects that I was working on
298 00:25:12.590 ⇒ 00:25:21.200 Uttam Kumaran: was related to standing up, like, our internal rail, and it has, like, had, like, a couple of phases, and so if I say, like.
299 00:25:21.330 ⇒ 00:25:26.769 Uttam Kumaran: Show me the plan for delivery and finance analytics.
300 00:25:27.680 ⇒ 00:25:32.419 Uttam Kumaran: And if I go to the platform team.
301 00:25:32.610 ⇒ 00:25:40.320 Uttam Kumaran: you should see that there are… there’s, like, these delivery and finance analytics. So I had, like, two first phases that…
302 00:25:40.620 ⇒ 00:25:42.080 Uttam Kumaran: I worked on.
303 00:25:42.300 ⇒ 00:25:46.749 Uttam Kumaran: And so… You can see here that it found the plan.
304 00:25:47.290 ⇒ 00:25:53.659 Uttam Kumaran: So, I actually spent probably, like, 30-40 minutes working on this plan that basically outlined, like.
305 00:25:54.580 ⇒ 00:26:11.680 Uttam Kumaran: five phases of setting up our entire internal data platform, which was, like, making sure Snowflake is cleaned up, which I haven’t done in, like, two years. It’s then, like, setting up the connections to RIL. I then also wanted to make sure that AI can, like, I can actually…
306 00:26:11.850 ⇒ 00:26:22.739 Uttam Kumaran: the cursor or codecs environment can actually set up rail, and that was verified. So, part of this was, like, and making sure dbt was set up. So, as you see here, like, I mean, a lot of these are actually finished.
307 00:26:22.990 ⇒ 00:26:26.879 Uttam Kumaran: Like, I had all of these various pieces.
308 00:26:27.120 ⇒ 00:26:30.500 Uttam Kumaran: And let me actually go see if I can find,
309 00:26:42.160 ⇒ 00:26:44.989 Uttam Kumaran: Yeah, so this one’s still here…
310 00:26:51.100 ⇒ 00:26:52.390 Uttam Kumaran: And let’s see…
311 00:26:59.180 ⇒ 00:27:05.780 Uttam Kumaran: Yeah, and so what you… so, like, let’s… let me show you a good example. So, Add Real Explore for Delivery Help Dashboard.
312 00:27:06.310 ⇒ 00:27:10.970 Uttam Kumaran: Linear helped me write this entire ticket.
313 00:27:11.280 ⇒ 00:27:15.509 Uttam Kumaran: And so what I actually did is I said, I worked on this plan first.
314 00:27:16.300 ⇒ 00:27:31.770 Uttam Kumaran: And then I said, great, now that this plan is ready, go ahead and help me phase it out and create tickets. And so, nicely now, there’s a whole section in Kirscher that explains to agents how we do ticketing.
315 00:27:31.820 ⇒ 00:27:41.420 Uttam Kumaran: So, linear will not only… like, Cursor and any agent will not only phase it out, but it actually will create tickets that are super, super rich.
316 00:27:41.470 ⇒ 00:27:57.510 Uttam Kumaran: and also start to tag them, whether AI can work on it, or it’s a human step. For example, like, I’m just gonna kinda bounce around, but, like, just follow with me. Like, there’s a marketing site, like, I worked on this, like, marketing site,
317 00:27:57.930 ⇒ 00:28:04.670 Uttam Kumaran: migration. So consider working with migration…
318 00:28:09.540 ⇒ 00:28:12.139 Uttam Kumaran: Yeah, so Brainforge website migration.
319 00:28:12.580 ⇒ 00:28:20.309 Uttam Kumaran: So… This is… I’ll show you, this is, should be up.
320 00:28:22.470 ⇒ 00:28:28.420 Uttam Kumaran: Design needs some work, but… It’s like… I didn’t do a single…
321 00:28:28.650 ⇒ 00:28:35.559 Uttam Kumaran: I didn’t touch a single fucking thing for this whole thing. But really, like, what I try to do is,
322 00:28:36.290 ⇒ 00:28:51.830 Uttam Kumaran: break it down into bite-sized chunks that I can hand to individual agents to go work on. Like, I didn’t say, okay, I run for 15 hours and, like, do the whole thing, because that’s not how it works. Like, you will get some… you will get, like, some shitty thing. So instead, what.
323 00:28:51.830 ⇒ 00:28:52.380 Robert Tseng: Yeah.
324 00:28:52.660 ⇒ 00:28:58.649 Uttam Kumaran: What I did is I, like, I first worked on, like, a really comprehensive 3- or 4-phase plan.
325 00:28:58.830 ⇒ 00:29:06.039 Uttam Kumaran: Right, so… Like, and then it’s like the foundation, the core page is content, SEO, and a cutover.
326 00:29:06.380 ⇒ 00:29:13.249 Uttam Kumaran: And we’ve been working through it. So there’s… there’s all these, like, different labels. So, for example.
327 00:29:13.390 ⇒ 00:29:16.019 Uttam Kumaran: This one you’re gonna see has, like,
328 00:29:16.560 ⇒ 00:29:21.940 Uttam Kumaran: Phase 2 core pages. There’s gonna be some stuff also on here that’s, like.
329 00:29:21.940 ⇒ 00:29:22.950 Robert Tseng: Phase 3.
330 00:29:22.950 ⇒ 00:29:29.630 Uttam Kumaran: phase… For my, for my AirPods, I’ll swap to my other one. Phase 4, Phase 5…
331 00:29:30.310 ⇒ 00:29:31.809 Robert Tseng: You, things run out.
332 00:29:33.490 ⇒ 00:29:44.709 Uttam Kumaran: And so, another thing that I did is I also said, anything that needs to have a human review, also leave out, because there’s gonna be pieces that, like, I need to look at and be like, okay, we’re set.
333 00:29:45.010 ⇒ 00:29:51.059 Uttam Kumaran: So there’s something a part of Phase 1, where we got to the end, it’s like, this is a human one, so I’m not gonna assign this to AI.
334 00:29:51.060 ⇒ 00:29:51.500 Robert Tseng: I see.
335 00:29:51.500 ⇒ 00:29:58.130 Uttam Kumaran: But, like, for example, there’s some stuff here where I’m like, like…
336 00:29:59.780 ⇒ 00:30:09.769 Uttam Kumaran: 6 critical ARIA and semantic accessibility blockers. Okay, great. So what I’m gonna do now is I’m gonna go to assign, and I can just sort of click on Codex, or Cursor.
337 00:30:10.120 ⇒ 00:30:20.120 Uttam Kumaran: both of these are live and working, and then if I click on Codex, and I’ll just click on, like, in progress, you’ll see that, it’s starting.
338 00:30:20.890 ⇒ 00:30:27.960 Uttam Kumaran: And so, what you can do is you can go to here, Open… X.
339 00:30:28.300 ⇒ 00:30:31.980 Uttam Kumaran: And you’ll see that it started to work on it. And it’s working on it in the cloud.
340 00:30:32.090 ⇒ 00:30:37.970 Uttam Kumaran: And we get, like, a bunch of codecs, usage as part of our ChatGPT plan.
341 00:30:38.060 ⇒ 00:30:47.910 Uttam Kumaran: And I hooked it all up to use our Azure subscription, of which we have, like, a shitload of money still left. So now it’s gonna work on it. But the actual difficulty here is, like.
342 00:30:47.930 ⇒ 00:30:58.320 Uttam Kumaran: It needs to install the right libraries, it needs to be able to have our environment secrets, so that’s what took, like, about, like, 2 or 3 days to, like, make sure that it can actually take
343 00:30:58.430 ⇒ 00:30:59.110 Uttam Kumaran: You have to continue.
344 00:30:59.110 ⇒ 00:31:00.390 Robert Tseng: Gatorize it, yeah.
345 00:31:00.390 ⇒ 00:31:02.820 Uttam Kumaran: Yeah, so it has all that.
346 00:31:03.000 ⇒ 00:31:08.110 Uttam Kumaran: And it has all the wonderful context that is in this ticket.
347 00:31:08.670 ⇒ 00:31:21.359 Uttam Kumaran: Which literally gives it, like, it’s like a perfect ticket, because a broader AI agent worked with me on creating the tickets, giving it really, really good acceptance criteria, so there’s not… there’s not much room for the AI to, like.
348 00:31:21.520 ⇒ 00:31:23.430 Uttam Kumaran: Mess up or get distracted.
349 00:31:23.740 ⇒ 00:31:32.020 Uttam Kumaran: And so it’ll work on it. Maybe take this one is, like, some type of SEO thing, so it’ll probably take, like, 5-10 minutes.
350 00:31:32.140 ⇒ 00:31:35.549 Uttam Kumaran: And then it’ll just basically send a note that it’s done.
351 00:31:36.060 ⇒ 00:31:37.139 Uttam Kumaran: And so…
352 00:31:37.360 ⇒ 00:31:57.020 Uttam Kumaran: what… what… what kind of the reason I want to show you this is you can do this, actually, for a lot of knowledge work, too. Like, if I want to show you an example of something that I was passing off, like, I always sometimes will have ideas of, like, hey, I’ll go research this thing, or, like, help me think through this. I just kick it off, and then I, like, let it sit. So, for example.
353 00:31:57.360 ⇒ 00:32:03.590 Uttam Kumaran: I was like, I had to work on, let’s see…
354 00:32:07.360 ⇒ 00:32:11.170 Uttam Kumaran: Awesome.
355 00:32:11.310 ⇒ 00:32:12.799 Uttam Kumaran: Surrounding, yeah.
356 00:32:13.470 ⇒ 00:32:23.380 Uttam Kumaran: So, I was, like, I was… I was sitting, I was like, oh, yeah, fuck, we have, like, 3… I was like, I was talking to someone, I’m like, yeah, we have 3 years of audio and video data.
357 00:32:24.050 ⇒ 00:32:24.550 Robert Tseng: Yeah.
358 00:32:24.850 ⇒ 00:32:36.100 Uttam Kumaran: just go, like, look at, think about what, how the fuck we can use that for something. And, I don’t know, I just, like, I literally just typed it in, and I just sent it. And so, it then looked through everything.
359 00:32:36.340 ⇒ 00:32:44.309 Uttam Kumaran: And then gave me, like, some interesting ideas about how we could use the audio and video to do coaching, and, like, digital twin, and, like…
360 00:32:44.760 ⇒ 00:32:49.650 Uttam Kumaran: some interesting stuff. I had it, like, look through papers and stuff like that, but the even more
361 00:32:49.980 ⇒ 00:32:56.099 Uttam Kumaran: thing about how I triggered this is I did it all through Slack. So, I had,
362 00:32:56.420 ⇒ 00:32:57.819 Uttam Kumaran: I had a,
363 00:33:02.940 ⇒ 00:33:04.360 Uttam Kumaran: I had…
364 00:33:05.000 ⇒ 00:33:15.759 Uttam Kumaran: So you’ll see in the platform channel, I literally said, we have years of audio and video and transcript data at Brainforge. Want you to brainstorm all the things we can do, think about short meeting long term, moonshot project, think hard and reference papers.
365 00:33:16.610 ⇒ 00:33:19.010 Uttam Kumaran: Cool. So I just triggered it, and then it said, done.
366 00:33:19.440 ⇒ 00:33:34.379 Uttam Kumaran: So for work like this, I’m actually interested in what ways that, like, people can start triggering work from Slack as well. Whether that’s research, whether that’s, like, go research this person, whether that’s, like, consider this thing, find this thing for me.
367 00:33:34.540 ⇒ 00:33:36.830 Uttam Kumaran: So I’m interested in people.
368 00:33:36.830 ⇒ 00:33:38.859 Robert Tseng: Do they have Codex for Can’t do Purser in here?
369 00:33:39.050 ⇒ 00:33:40.759 Uttam Kumaran: No, you can do cursor, too. I just…
370 00:33:40.760 ⇒ 00:33:41.420 Robert Tseng: Okay.
371 00:33:42.130 ⇒ 00:33:43.940 Uttam Kumaran: Both are set up.
372 00:33:43.940 ⇒ 00:33:44.310 Robert Tseng: Okay.
373 00:33:44.900 ⇒ 00:33:49.040 Uttam Kumaran: And then eventually, I think it’ll all collapse into Brainforge Assistant, but yeah.
374 00:33:49.040 ⇒ 00:33:49.700 Robert Tseng: Yeah.
375 00:33:49.840 ⇒ 00:33:56.650 Uttam Kumaran: So, Kersher, the UX is a little bit nicer, meaning, like, for this one, Kersher would have literally sent me the message here.
376 00:33:57.300 ⇒ 00:34:00.860 Robert Tseng: Yeah, I haven’t use codecs, I’m just gonna keep using First Circ. That’s fine.
377 00:34:00.860 ⇒ 00:34:04.229 Uttam Kumaran: Yeah, I just… I’m more interested in people just using it.
378 00:34:04.380 ⇒ 00:34:07.959 Uttam Kumaran: So I can tell, like, what’s R, what’s different, what’s easy.
379 00:34:08.370 ⇒ 00:34:25.569 Uttam Kumaran: But yeah, so I, for example, I said, I want you to investigate leveraging open code as an alternative to Krisha Cloud Agents. I’m triggering 10 to 30 agent sessions per day, it’s getting expensive, go look through it. I had it literally write me this plan about migrating some stuff to OpenCode.
380 00:34:25.760 ⇒ 00:34:30.699 Uttam Kumaran: As, like, an alternative so that I can start passing it easy tickets.
381 00:34:31.110 ⇒ 00:34:39.239 Uttam Kumaran: And we can host everything on our own. So, it went and did the research, and pulled… it pushed the full investigation
382 00:34:39.449 ⇒ 00:34:42.499 Uttam Kumaran: Into the repo for this whole plan.
383 00:34:42.750 ⇒ 00:34:50.690 Uttam Kumaran: So you can envision a world where I not only trigger an agent to create the plan, I review the plan, I’m like, this plan is great.
384 00:34:50.800 ⇒ 00:35:00.489 Uttam Kumaran: Then what I did is I go from this plan, and I literally just… this is exactly what I did. I went from this plan to, in linear, you’ll find…
385 00:35:01.050 ⇒ 00:35:02.150 Uttam Kumaran: Open code.
386 00:35:02.310 ⇒ 00:35:04.499 Uttam Kumaran: the four phases. So you see…
387 00:35:04.500 ⇒ 00:35:04.870 Robert Tseng: Phase 1.
388 00:35:04.870 ⇒ 00:35:07.980 Uttam Kumaran: And we’re currently in the middle of it. Like, this is what I was working on.
389 00:35:09.290 ⇒ 00:35:21.279 Uttam Kumaran: on Saturday, you’ll see that there’s issues, and you can see I passed one of them here, and the rest, like, I still have to kind of work on some of these. I then work… I think I started on the next phase…
390 00:35:21.740 ⇒ 00:35:27.639 Uttam Kumaran: 2, okay, maybe not.
391 00:35:28.220 ⇒ 00:35:32.929 Uttam Kumaran: So, like, that’s how it’s gonna move. So if you think about, like, one of our,
392 00:35:34.110 ⇒ 00:35:39.370 Uttam Kumaran: If you think about our projects, we are actually very close to…
393 00:35:39.610 ⇒ 00:35:46.259 Uttam Kumaran: being able to trigger probably, like, 60 or 70% of project work through cloud agents.
394 00:35:46.410 ⇒ 00:35:48.140 Uttam Kumaran: And have them one-shot it.
395 00:35:48.370 ⇒ 00:35:58.019 Uttam Kumaran: Like, I, I one-shotted most of our, our reel for our company, kind of from scratch.
396 00:35:58.170 ⇒ 00:36:03.369 Uttam Kumaran: But I had to go through every step. I had to go through Snowflake setup.
397 00:36:03.430 ⇒ 00:36:19.420 Uttam Kumaran: the environment key setup. I went through setting up dbt. I then broke it out like I would break out a project plan for us, where it’s like, okay, make these models. First think about what models we need. Okay, I like those models, go make them. Then… but the nice thing about,
398 00:36:19.740 ⇒ 00:36:37.509 Uttam Kumaran: these cloud agencies, they can see. Like, they will… they’ll bring up whatever they’re building, take a screenshot, and use it to, critique themselves. So, these are, like, our… we have a bunch of delivery analytics, like, dashboards now, and all of these, it ain’t worked on these.
399 00:36:38.170 ⇒ 00:36:46.249 Uttam Kumaran: So… Max climb share, top 3 share, clients above 15%, risk score, longest client tenure.
400 00:36:46.620 ⇒ 00:36:51.320 Uttam Kumaran: And I don’t know, like, I haven’t… I have not yet, like, gone super deep on these, but, like.
401 00:36:51.780 ⇒ 00:36:52.890 Uttam Kumaran: I…
402 00:36:53.540 ⇒ 00:37:01.369 Uttam Kumaran: did what would take, like, for our size business, I did the entire data platform without touching a line of code. I didn’t touch a single thing.
403 00:37:02.550 ⇒ 00:37:06.940 Uttam Kumaran: And it probably took me… It probably took me, like.
404 00:37:08.730 ⇒ 00:37:17.190 Uttam Kumaran: 10 hours of, like, just, like, making sure that I’m… like, I’m assign… I have the plan ready, I’m assigning it in the right order.
405 00:37:17.500 ⇒ 00:37:36.349 Uttam Kumaran: And then in case, like, it got jammed up, I, like, would have another agent help fix it. So I would literally talk to Cursor and say, go look at these cloud agents, because I’ve given Cursor the Codex API and the Cursor API. So if a cloud agent gets stuck, I’m like, yo, the cloud agent’s stuck, or, like, it fucked something up.
406 00:37:36.500 ⇒ 00:37:43.980 Uttam Kumaran: go investigate. First, just fix it, and then tell me why our original plan didn’t
407 00:37:44.180 ⇒ 00:37:51.030 Uttam Kumaran: allow for… like, why did our original plan not, like, indicate that it would get stuck? So, we’ve been improving every time.
408 00:37:51.180 ⇒ 00:37:57.419 Uttam Kumaran: So, literally, dude, I worked with Berser on this plan. I said, I want to work
409 00:37:57.580 ⇒ 00:38:10.130 Uttam Kumaran: on delivery dashboards and finance dashboards, I said, you… spend some time, and you know… I said, you have a really good understanding of my business. Here’s the Elantra deck. Here’s that… that deck from that woman from Vixel that they sent me.
410 00:38:10.130 ⇒ 00:38:10.650 Robert Tseng: Yeah.
411 00:38:10.650 ⇒ 00:38:27.609 Uttam Kumaran: And first, like, let’s work on, like, the core dashboards, like, the scopes. I liked it. And then I said, okay, now go ahead and build it. RIL is a completely BI as code platform, dbt is completely code, Snowflake is completely controllable as code.
412 00:38:27.820 ⇒ 00:38:31.320 Uttam Kumaran: I didn’t need to set up polyatomic, but, like, we could have done that as code.
413 00:38:32.020 ⇒ 00:38:36.839 Uttam Kumaran: And yeah, this dashboard’s, like, a little bit ugly, but I actually didn’t, like.
414 00:38:36.980 ⇒ 00:38:47.080 Uttam Kumaran: spend much time thinking about the dashboard UI, but it was pulling this up, making edits, see that there was real error, pulling it back down, changing it, pulling it back up, checking.
415 00:38:47.420 ⇒ 00:38:48.710 Uttam Kumaran: And so…
416 00:38:49.070 ⇒ 00:38:54.970 Uttam Kumaran: I learned that, like, yeah, as long as you get the ordering right and the tickets are really good, you could do it. So I,
417 00:38:55.220 ⇒ 00:39:04.240 Uttam Kumaran: I… like… Yeah, I mean, I think we could… we’re gonna be able to build most of Data Platform
418 00:39:04.470 ⇒ 00:39:05.410 Uttam Kumaran: Like…
419 00:39:05.840 ⇒ 00:39:17.340 Uttam Kumaran: Right, basically, I think most of, like, people’s job at the company is going to get more towards working on the plans and assigning the AI agents, and our team is not using it right now, because I just don’t think they’re, like.
420 00:39:17.730 ⇒ 00:39:24.880 Uttam Kumaran: Seeing this, but… Like, a way… if these guys need to be working on having comprehensive plans.
421 00:39:25.110 ⇒ 00:39:34.380 Uttam Kumaran: And what I’m gonna build for them is, like, really great dbt testing environments, really great UI, like, dashboard environments. And their job is just to be click and assign.
422 00:39:34.500 ⇒ 00:39:36.000 Uttam Kumaran: And verify when done.
423 00:39:36.390 ⇒ 00:39:41.290 Uttam Kumaran: Like, whether this takes, like, an hour for an AI or 3 hours, it doesn’t really matter.
424 00:39:41.590 ⇒ 00:39:42.639 Uttam Kumaran: You know?
425 00:39:42.940 ⇒ 00:39:43.930 Uttam Kumaran: Yup.
426 00:39:44.910 ⇒ 00:39:53.870 Uttam Kumaran: Because now Awash can assign 5 of these agents, because this is what I… dude, I was doing work the whole time that’s getting built. It’s not like I’m, like, sitting on my ass.
427 00:39:53.870 ⇒ 00:39:54.200 Robert Tseng: Yeah.
428 00:39:54.200 ⇒ 00:39:57.109 Uttam Kumaran: This was built… the background.
429 00:39:57.490 ⇒ 00:40:16.629 Uttam Kumaran: that I’m doing other stuff. And so, the net-net here is that, like, when I’m working now, at any moment, I’m using Cursor to work on phases, I’m using Cursor to say, cool, now assign… go through the phases, and start with the ones, and assign those out to Cursor or Codex.
430 00:40:16.790 ⇒ 00:40:18.899 Uttam Kumaran: And then let me know when those are finished.
431 00:40:19.090 ⇒ 00:40:20.799 Uttam Kumaran: And then I work on other stuff.
432 00:40:21.150 ⇒ 00:40:24.659 Uttam Kumaran: So I’ve effectively maximized my…
433 00:40:24.980 ⇒ 00:40:39.409 Uttam Kumaran: like, the amount of threads I could do, right? So, usually I’m doing 3 or 4 things at a time. Now, technically, if things are running in the cloud, I can do 5 or 6, 7, 8 things at a time. You can now trigger some of these mobile.
434 00:40:39.510 ⇒ 00:40:40.810 Uttam Kumaran: Right?
435 00:40:41.010 ⇒ 00:40:45.330 Uttam Kumaran: And we haven’t even created, like, an auto-trigger system, which is, like.
436 00:40:45.800 ⇒ 00:40:48.400 Uttam Kumaran: You go run the entire project.
437 00:40:49.230 ⇒ 00:40:56.749 Uttam Kumaran: like, I still have to manually trigger stuff. Yeah. I just probably need, like, a few more days, and we can have almost, like, a governor that, like.
438 00:40:56.980 ⇒ 00:41:01.070 Uttam Kumaran: Looks at the project, triggers things, waits for it to get done, verifies the work.
439 00:41:01.220 ⇒ 00:41:04.229 Uttam Kumaran: Tells me if there’s some, like, something that’s super stuck.
440 00:41:05.640 ⇒ 00:41:13.230 Uttam Kumaran: And then our job is gonna be to sort of pace it out, and… Collect.
441 00:41:13.790 ⇒ 00:41:14.590 Uttam Kumaran: You know?
442 00:41:15.820 ⇒ 00:41:24.089 Uttam Kumaran: So, that’s why our… the design choices around things that are, like, infrastructure as code is gonna be very, very important here.
443 00:41:26.250 ⇒ 00:41:32.819 Uttam Kumaran: But, like, for the most part, like, a lot… I feel like this was a really good view of… like, I did this solo.
444 00:41:33.670 ⇒ 00:41:40.559 Uttam Kumaran: from just raw data and stuff, like, you know? And then this is, like, one of, like, 4 or 5. So now I’m working on, like, 4 or 5 of these, like.
445 00:41:40.720 ⇒ 00:41:47.899 Uttam Kumaran: grand projects at a time, because I’ve now… I now am confident in our planning that I can actually, like.
446 00:41:48.090 ⇒ 00:41:50.199 Uttam Kumaran: Take a step above, you know?
447 00:41:50.730 ⇒ 00:41:51.440 Robert Tseng: Yes.
448 00:41:57.740 ⇒ 00:41:58.540 Robert Tseng: Huh.