Meeting Title: All Engineering (AI Uses + Workshop) Date: 2025-08-21 Meeting participants: Samuel Roberts, Uttam Kumaran, Annie Yu, Demilade Agboola, Mustafa Raja, Ryan Brosas, Awaish Kumar, Vashdev Heerani
WEBVTT
1 00:01:05.040 ⇒ 00:01:05.930 Uttam Kumaran: Damn.
2 00:01:06.640 ⇒ 00:01:07.670 Samuel Roberts: Hello.
3 00:01:07.670 ⇒ 00:01:10.800 Uttam Kumaran: Hey, sorry, my dog, my dog is just barking all the time.
4 00:01:10.830 ⇒ 00:01:12.579 Samuel Roberts: Oh, you’re good, you’re good.
5 00:01:16.330 ⇒ 00:01:18.030 Samuel Roberts: How’s it going today?
6 00:01:19.210 ⇒ 00:01:23.890 Uttam Kumaran: It is good. … Yeah, it’s good.
7 00:01:24.290 ⇒ 00:01:25.530 Samuel Roberts: Okay.
8 00:01:28.300 ⇒ 00:01:30.060 Uttam Kumaran: Every day is busy, so….
9 00:01:30.060 ⇒ 00:01:31.389 Samuel Roberts: Yeah, yeah.
10 00:01:32.100 ⇒ 00:01:33.800 Samuel Roberts: A different scale, you know?
11 00:01:34.380 ⇒ 00:01:35.250 Uttam Kumaran: Yeah.
12 00:01:38.630 ⇒ 00:01:44.910 Uttam Kumaran: Yeah, I was thinking today… you know, I could… mainly, I just wanna… sort of…
13 00:01:45.030 ⇒ 00:01:48.499 Uttam Kumaran: Give an opportunity for everyone to sort of work on this.
14 00:01:48.770 ⇒ 00:01:52.149 Uttam Kumaran: like, whiteboard I have on different areas of
15 00:01:52.950 ⇒ 00:01:55.530 Uttam Kumaran: Of automation, and maybe talk about
16 00:01:56.900 ⇒ 00:01:59.600 Uttam Kumaran: I mainly want to talk about, like, where…
17 00:01:59.770 ⇒ 00:02:05.250 Uttam Kumaran: time is going now, and then I kind of want to share our plan …
18 00:02:05.620 ⇒ 00:02:14.769 Uttam Kumaran: on the AI team at Howard, but I’m looking at the whole company, and then I just have, like, you know, I use AI in a lot of different ways. I thought it would just be nice to refresh
19 00:02:14.890 ⇒ 00:02:23.119 Uttam Kumaran: folks and share, like, some of the things that I’m doing on everything from communication to strategy to…
20 00:02:23.310 ⇒ 00:02:29.829 Uttam Kumaran: Code review, and… Yeah, just kind of, like, do a little bit of a working session, so….
21 00:02:29.830 ⇒ 00:02:49.059 Samuel Roberts: Yeah, that sounds good. I mean, I think that’ll be really cool, because I love insight into, like, how you use it, because I’m so used to using it for, like, engineering tasks and, you know, mostly cursor and occasionally, like, chat, but, like, I’m very curious, like, other use cases beyond that for understanding, like, more of the broader company, like, uses and stuff.
22 00:02:50.440 ⇒ 00:02:58.070 Samuel Roberts: I feel like I get pigeonholed into this, like, you know, coding stuff so much when I’m using it, like, that’s where I see so much of it, but I know there’s… I know there’s more, obviously, and it would be good.
23 00:02:58.070 ⇒ 00:03:02.460 Uttam Kumaran: Yeah, and I’m a good guinea pig because I’m so low on time.
24 00:03:02.600 ⇒ 00:03:03.160 Samuel Roberts: Yeah.
25 00:03:03.160 ⇒ 00:03:09.369 Uttam Kumaran: some… some… some… like, I can’t do… I can’t do my job right now without AI. Like, it’s… Right.
26 00:03:09.690 ⇒ 00:03:21.629 Uttam Kumaran: I use it… it saves me, in some situations, 50%, in some situations, 90% of the time. Like, I will actually… I’ll share… I’ll share with everybody, like, a
27 00:03:21.920 ⇒ 00:03:28.120 Uttam Kumaran: like, a scope of work we had to produce yesterday, and how I use AI to do it, and it’s not just, like.
28 00:03:28.610 ⇒ 00:03:32.059 Uttam Kumaran: copy-paste and copy whatever it says. You have to work with it, but it….
29 00:03:32.060 ⇒ 00:03:32.710 Samuel Roberts: Yeah.
30 00:03:32.710 ⇒ 00:03:37.819 Uttam Kumaran: I would not have been able to do this in the 30 minutes it took me to do it, because I was late on doing it, so…
31 00:03:39.560 ⇒ 00:03:40.950 Uttam Kumaran: Okay.
32 00:03:41.540 ⇒ 00:03:51.070 Uttam Kumaran: Let me just get this big jam set up, and… … Everybody in here.
33 00:04:52.270 ⇒ 00:04:54.400 Uttam Kumaran: Okay, so…
34 00:04:59.740 ⇒ 00:05:06.350 Uttam Kumaran: Great, I’m gonna send this… big jam in the Zoom chat, so everyone can join.
35 00:05:06.880 ⇒ 00:05:07.940 Uttam Kumaran: this…
36 00:05:23.220 ⇒ 00:05:30.240 Uttam Kumaran: And, sorry, I’m just, like, hustling to use our retro template here, but I will sort of guide us on how to think about this.
37 00:05:43.770 ⇒ 00:05:44.980 Uttam Kumaran: There’s…
38 00:05:50.870 ⇒ 00:05:52.020 Uttam Kumaran: Okay.
39 00:05:53.030 ⇒ 00:05:54.819 Uttam Kumaran: So, …
40 00:05:55.550 ⇒ 00:06:04.450 Uttam Kumaran: One of the things that we’re starting to offer, for clients is what we’re calling, like, an AI
41 00:06:04.570 ⇒ 00:06:10.159 Uttam Kumaran: growth sprint, and I kind of want to do, like, a little bit of a mini version, …
42 00:06:10.460 ⇒ 00:06:15.380 Uttam Kumaran: for us today, and I think it would be…
43 00:06:16.210 ⇒ 00:06:24.400 Uttam Kumaran: Pretty cool to see, like, what results we think about when we think about, our company. So let me, …
44 00:06:25.440 ⇒ 00:06:27.070 Uttam Kumaran: Let me pull this off.
45 00:06:27.360 ⇒ 00:06:28.570 Uttam Kumaran: Dial?
46 00:06:38.180 ⇒ 00:06:40.070 Uttam Kumaran: So, okay.
47 00:06:45.920 ⇒ 00:07:02.230 Uttam Kumaran: Great. So, basically, kind of the structure of what I want to present today is sort of similar to what we are starting to do for clients, and to give you guys a view of that, this is, like, a view of the kind of, like, demo that we
48 00:07:02.230 ⇒ 00:07:08.439 Uttam Kumaran: the workshop that we’re starting to do for certain clients. So, when we come into a client, you know, let’s say, …
49 00:07:09.030 ⇒ 00:07:16.309 Uttam Kumaran: what’s a good example? That’s simple. Let’s say we’re going to a, … A local pet store.
50 00:07:16.650 ⇒ 00:07:26.799 Uttam Kumaran: Right? And let’s say they’re our client. And so, sometimes we get called, let’s say PetStore calls us and says, hey, like, we’re interested in AI automation, we have several
51 00:07:26.950 ⇒ 00:07:36.029 Uttam Kumaran: things that we’ve tried, like, we give everyone chat to BT, but we’re just not seeing the impact. And we have an internal AI council, but, you know, I don’t think we…
52 00:07:36.230 ⇒ 00:07:45.199 Uttam Kumaran: have the muscle internally to sort of get this done. And so, one of the things that I want to talk about today is
53 00:07:45.280 ⇒ 00:07:59.850 Uttam Kumaran: The work that we do for them is typically we get everyone in the room, we have them talk through what are their biggest challenges, and what are the biggest opportunities, and then we work through, okay, like, what are the ways we could pair the challenges, opportunities, and come up with ideas?
54 00:08:00.100 ⇒ 00:08:09.000 Uttam Kumaran: So today I want to kind of do a little bit of a mini version of that. So the goal here is actually not to think about, necessarily.
55 00:08:09.220 ⇒ 00:08:25.070 Uttam Kumaran: like, the AI implementation, it’s actually just to list down the problems and opportunities. There may be things that can be achieved by AI today, there will be certainly a lot of things that can partly be achieved, and there will be some things that we can’t do. And so.
56 00:08:25.150 ⇒ 00:08:32.030 Uttam Kumaran: one thing that I want to avoid is the sort of pressure of everyone having to think through,
57 00:08:32.120 ⇒ 00:08:47.939 Uttam Kumaran: you know, like, what is the… what is the AI thing here? Instead, what I would like is… and the reason why I have all these things listed here is I’m open to everybody giving feedback on anything, but I want to start to see what the opportunities are and what the existing problems are.
58 00:08:48.170 ⇒ 00:08:54.149 Uttam Kumaran: After that, I think we can go through and start to find general themes, and then, as an AI team.
59 00:08:54.360 ⇒ 00:09:06.580 Uttam Kumaran: You know, on the call here, I think you have me, Sam. I’m not sure if Casey’s on, I think it’s just me, Sam, Mustafa. We can talk a little bit about how we see some of our work pairing on some of these problems.
60 00:09:07.030 ⇒ 00:09:11.309 Uttam Kumaran: Does that roughly make sense? So, the instructions…
61 00:09:12.060 ⇒ 00:09:21.210 Uttam Kumaran: Basically are, one, think about… So, write down… Opportunities.
62 00:09:23.210 ⇒ 00:09:29.319 Uttam Kumaran: In each of the areas, and then write down problem areas.
63 00:09:30.160 ⇒ 00:09:37.599 Uttam Kumaran: And it could be anything, so I don’t worry about, like, is it… can it be AI, whatever. I really just want to hear about each of these areas.
64 00:09:37.800 ⇒ 00:09:56.749 Uttam Kumaran: I know a lot of folks on this call are going to be on the engineering side, but everybody in this call was recruited by the company, was onboarded through operations, or… which was just me before, has some interest in the sales side, of course understands client and project management. Of course, I’m going to add project management here.
65 00:09:57.590 ⇒ 00:10:03.059 Uttam Kumaran: And so I want to see some of these kind of, like, problems and opportunities come out. …
66 00:10:03.240 ⇒ 00:10:18.509 Uttam Kumaran: And yeah, I think at that point we can sort of talk through some… some ideas, and this gives us a really good roadmap for our team, and then I will spend the back half of this call sharing some AI stuff that I’ve been working on, and sort of give some people some… some ways of actually implementing this today.
67 00:10:19.710 ⇒ 00:10:22.249 Uttam Kumaran: Does that make sense? Any questions?
68 00:10:26.850 ⇒ 00:10:28.680 Samuel Roberts: Yeah, no, I think it sounds good.
69 00:10:29.340 ⇒ 00:10:35.690 Samuel Roberts: Yeah, I’m trying to think who… are we… should we… are there people here that are not here that we want here for this?
70 00:10:35.690 ⇒ 00:10:41.319 Uttam Kumaran: Yeah, do you wanna… maybe we can ping, Ryan B? Actually.
71 00:10:42.710 ⇒ 00:10:46.589 Samuel Roberts: I can just… we can share the link on that, message you sent, maybe.
72 00:10:46.590 ⇒ 00:10:49.119 Uttam Kumaran: Yeah, … And I think this was….
73 00:10:49.120 ⇒ 00:10:50.280 Samuel Roberts: Had been to stand in anything.
74 00:10:50.280 ⇒ 00:10:53.580 Uttam Kumaran: Ryan and Rico, ….
75 00:10:54.500 ⇒ 00:10:55.180 Samuel Roberts: ….
76 00:10:55.500 ⇒ 00:10:59.719 Uttam Kumaran: I… Robert and Henry are at a client meeting right now.
77 00:10:59.870 ⇒ 00:11:00.750 Samuel Roberts: Okay.
78 00:11:01.360 ⇒ 00:11:02.760 Samuel Roberts: Yeah, is there…
79 00:11:05.770 ⇒ 00:11:09.110 Samuel Roberts: I’m just trying to see where… where do I find the, like, link to this?
80 00:11:11.240 ⇒ 00:11:13.139 Uttam Kumaran: I sent it in the Zoom chat.
81 00:11:13.830 ⇒ 00:11:18.740 Samuel Roberts: Oh, no, link to the fig… I mean, to the video in general, but… Call.
82 00:11:19.140 ⇒ 00:11:21.230 Uttam Kumaran: Oh, you can get it from the calendar invite.
83 00:11:21.740 ⇒ 00:11:26.609 Samuel Roberts: Oh, okay, yeah, I’m looking through, I’m looking through Zoom, and I’m like… yeah, cool, okay, that makes.
84 00:11:26.610 ⇒ 00:11:45.599 Uttam Kumaran: So maybe with this crew, let’s begin, but yeah, I would just add them, and then they’ll probably get the hang of it. So, like, if I could just go one step further, like, let’s… let’s talk about an example. So one thing I could put here is, like, what is an opportunity on operations? So I would say, currently, like, a problem we have is, like, time to onboard
85 00:11:45.650 ⇒ 00:11:47.579 Uttam Kumaran: New clients.
86 00:11:48.240 ⇒ 00:11:50.490 Uttam Kumaran: Right? That’s, like, a problem we have right now.
87 00:11:50.780 ⇒ 00:11:52.990 Uttam Kumaran: It can be as simple as that.
88 00:11:53.240 ⇒ 00:12:02.809 Uttam Kumaran: On the engineering side, for example, one thing I can do is, like, … … Basically, it’s like, how…
89 00:12:04.270 ⇒ 00:12:10.070 Uttam Kumaran: I can use cursor to… you know, create…
90 00:12:10.270 ⇒ 00:12:14.149 Uttam Kumaran: simple PRs for my data tickets.
91 00:12:14.690 ⇒ 00:12:16.540 Uttam Kumaran: That’s a great problem, right?
92 00:12:16.850 ⇒ 00:12:21.629 Uttam Kumaran: … So, cool. Maybe I’m just gonna start the timer for, like, 10 minutes.
93 00:12:22.220 ⇒ 00:12:23.210 Uttam Kumaran: …
94 00:12:23.630 ⇒ 00:12:30.620 Uttam Kumaran: I’m most interested, especially the folks that are on engineering, to give comments, not only on engineering stuff, but also PM stuff.
95 00:12:31.540 ⇒ 00:12:34.899 Uttam Kumaran: And then I’ll sort of start to add, you know.
96 00:12:35.070 ⇒ 00:12:42.969 Uttam Kumaran: things everywhere, but… great. Cool, so let me put a timer on, and then… Yes, we will begin.
97 00:12:44.590 ⇒ 00:12:46.610 Uttam Kumaran: Okay, so 5 minutes.
98 00:15:05.250 ⇒ 00:15:10.289 Annie Yu: I somehow can’t… really typing anything.
99 00:15:10.550 ⇒ 00:15:12.780 Uttam Kumaran: Oh, okay, let me, …
100 00:15:17.710 ⇒ 00:15:19.140 Uttam Kumaran: Okay, one second.
101 00:15:22.790 ⇒ 00:15:24.470 Uttam Kumaran: Can you try now?
102 00:15:25.780 ⇒ 00:15:26.600 Annie Yu: Okay.
103 00:15:28.530 ⇒ 00:15:30.280 Uttam Kumaran: You may have to refresh.
104 00:15:30.280 ⇒ 00:15:30.925 Annie Yu: Mmm…
105 00:15:48.470 ⇒ 00:15:50.290 Annie Yu: Yes, thank you.
106 00:15:50.290 ⇒ 00:15:51.360 Uttam Kumaran: Great.
107 00:17:24.940 ⇒ 00:17:27.559 Uttam Kumaran: Okay, I’m gonna add 2 more minutes.
108 00:19:13.520 ⇒ 00:19:17.439 Uttam Kumaran: And if you have something that just doesn’t fit in the category, just put it anywhere.
109 00:19:17.700 ⇒ 00:19:18.550 Uttam Kumaran: …
110 00:19:52.670 ⇒ 00:19:54.419 Uttam Kumaran: Okay, how do we feel?
111 00:19:54.710 ⇒ 00:19:56.529 Uttam Kumaran: Some more time, or…?
112 00:19:56.980 ⇒ 00:19:58.560 Uttam Kumaran: Should we chat through stuff?
113 00:20:01.570 ⇒ 00:20:07.669 Uttam Kumaran: It’s okay if you didn’t put anything down, or maybe you got one thing down. I think we’ll continue to do this week.
114 00:20:07.800 ⇒ 00:20:18.839 Uttam Kumaran: Week by week, so it’ll get… it’ll get easier, especially as… as everyone here starts to see how these problems go from problems to proof of concept to solved.
115 00:20:18.940 ⇒ 00:20:24.939 Uttam Kumaran: I think it’ll start to open up everybody’s brain about, like, hey, I’m doing this every day, like, is there a way for me to solve this?
116 00:20:25.110 ⇒ 00:20:44.199 Uttam Kumaran: I think my main point for this is we’re not… now that the company’s big, like, we’re not seeing every process, and so part of this is, like, how can we enable everybody in the company to look at the things they’re doing and say, hey, like, I’m spending time doing this every day. I feel like there’s probably something we can automate here so I can move on.
117 00:20:44.650 ⇒ 00:20:46.199 Uttam Kumaran: Would love to discuss it.
118 00:20:49.300 ⇒ 00:20:50.220 Uttam Kumaran: Cool.
119 00:20:50.480 ⇒ 00:20:57.239 Uttam Kumaran: Okay, so I think what we can do, we don’t have a ton of stickies, but maybe what we can do is…
120 00:20:57.510 ⇒ 00:21:15.619 Uttam Kumaran: just start to group things kind of by area, so I’ll… I can start to do that. So I think there’s some stuff about, like, people onboarding and recruiting, so I want to kind of move that here, related to people.
121 00:21:15.840 ⇒ 00:21:21.259 Uttam Kumaran: So these are all… recruiting related.
122 00:21:24.550 ⇒ 00:21:26.350 Uttam Kumaran: on the PM side.
123 00:21:31.220 ⇒ 00:21:35.800 Uttam Kumaran: Yeah, I agree. Like, I think grooming is really, really tough.
124 00:21:36.610 ⇒ 00:21:38.870 Uttam Kumaran: right now. And I think…
125 00:21:39.140 ⇒ 00:21:51.520 Uttam Kumaran: my feedback is that, like, I think ticket quality is directly related to the quality of the output. We all know, like, when you get a ticket that’s, like, doesn’t have anything, or, like, is so generic.
126 00:21:51.980 ⇒ 00:21:57.830 Uttam Kumaran: Like, you have just nowhere… You just, like, just don’t know where to start, and…
127 00:21:58.220 ⇒ 00:22:08.579 Uttam Kumaran: you know, I mean, on one side, I’m happy that we have tickets at all, but on the other side, they can be so much more better, so much better. So, I kind of want to…
128 00:22:09.210 ⇒ 00:22:15.650 Uttam Kumaran: Kind of, like, have something around… … like, grooming?
129 00:22:16.550 ⇒ 00:22:18.900 Uttam Kumaran: So I’m gonna move this here.
130 00:22:19.360 ⇒ 00:22:25.469 Uttam Kumaran: I also sort of, like, I don’t like stand-ups, like, at all.
131 00:22:25.620 ⇒ 00:22:32.539 Uttam Kumaran: And I think that a lot of the times, We’re, like, rehashing things that…
132 00:22:33.180 ⇒ 00:22:38.680 Uttam Kumaran: I feel like are really, really obvious, but maybe it’s not obvious to the whole team.
133 00:22:38.980 ⇒ 00:22:46.169 Uttam Kumaran: And I don’t know, I feel like there’s gotta be a better way. You know, for me, I want to be a company where we have
134 00:22:46.420 ⇒ 00:22:53.770 Uttam Kumaran: As minimal recurring meetings as possible, especially meetings that have to happen every… every single day or every other day.
135 00:22:53.910 ⇒ 00:22:56.770 Uttam Kumaran: Where we’re not directly meeting with a client.
136 00:22:57.150 ⇒ 00:23:05.579 Uttam Kumaran: you know, it’s just such a breakup of energy and time, and then also, like, just sitting and being like, I didn’t get this done, I got this done, like…
137 00:23:05.850 ⇒ 00:23:14.969 Uttam Kumaran: I don’t know, there’s just gotta be a better way for us to manage this. Part of this is on us as engineers. Like, I don’t do the best job of updating my tickets.
138 00:23:15.350 ⇒ 00:23:19.270 Uttam Kumaran: And I think that causes stress for the PM.
139 00:23:19.560 ⇒ 00:23:26.670 Uttam Kumaran: part of this is also, like, I wish I could just put updates somewhere, and then they can get read.
140 00:23:27.070 ⇒ 00:23:29.950 Uttam Kumaran: So, I don’t know, I kind of, like…
141 00:23:30.510 ⇒ 00:23:36.640 Uttam Kumaran: Maybe you want to just group everything related to, like, stand-ups and, like, recurring updates here.
142 00:23:36.640 ⇒ 00:23:37.850 Samuel Roberts: Yeah….
143 00:23:39.300 ⇒ 00:23:44.029 Uttam Kumaran: I don’t know if anyone else feels… I just went on a huge rant, but if anyone else feels similarly?
144 00:23:44.030 ⇒ 00:23:51.480 Samuel Roberts: No, that’s definitely something… that’s why I put the one that I put there, where I was like, I’m not always sure, and maybe it’s just also, like, kind of being new and not sure where things, like.
145 00:23:51.760 ⇒ 00:24:06.029 Samuel Roberts: does this comment go in linear? Does it comment go in Slack? Does this go in Notion Doc? Like, it’s just… there’s information different places, and I’m never sure, like, where the most, valuable update is, and even if it is, like, at some point, some more automation that, like, oh, the linear stuff gets…
146 00:24:06.110 ⇒ 00:24:24.329 Samuel Roberts: condensed, and I don’t know, I think I can, you know, start to imagine that, but I think just the problem is, yeah, like, the linear tickets aren’t as robust as they could be, so then the comments don’t get there, or things live in other places, so then the information is, you know, fragmented, or just not updated, and yeah, there’s definitely…
147 00:24:24.670 ⇒ 00:24:26.299 Samuel Roberts: I think stuff’s to do there.
148 00:24:26.710 ⇒ 00:24:27.390 Uttam Kumaran: Yeah.
149 00:24:27.390 ⇒ 00:24:30.020 Samuel Roberts: Yeah, I don’t… you’re right, a stand-up every day is not necessarily….
150 00:24:30.880 ⇒ 00:24:36.720 Uttam Kumaran: Yeah, it should be, like, optional, or it should be reserved for, like, an issue, right?
151 00:24:36.720 ⇒ 00:24:37.530 Samuel Roberts: Exactly.
152 00:24:37.530 ⇒ 00:24:43.160 Uttam Kumaran: I’m like… and I know, it’s like, some people will say it’s just 30 minutes, but it’s a lot of time, and like…
153 00:24:43.550 ⇒ 00:24:46.190 Uttam Kumaran: I don’t know, I’d rather not be in a meeting, so….
154 00:24:46.330 ⇒ 00:24:48.090 Samuel Roberts: Yeah, breaks up your day.
155 00:24:48.090 ⇒ 00:24:53.259 Uttam Kumaran: Yeah, so… Tickets should actually be done… so this is more of, like.
156 00:24:53.560 ⇒ 00:24:57.509 Uttam Kumaran: Ticket deep… so ticket, like, orchestration….
157 00:24:57.510 ⇒ 00:24:58.130 Samuel Roberts: Yeah.
158 00:24:58.620 ⇒ 00:25:05.180 Uttam Kumaran: And then this one is a lot about, like, I’m just gonna say, like.
159 00:25:05.340 ⇒ 00:25:08.720 Uttam Kumaran: virtual, or like, let’s just say sprint calls.
160 00:25:09.140 ⇒ 00:25:09.740 Samuel Roberts: Yeah.
161 00:25:10.180 ⇒ 00:25:14.820 Uttam Kumaran: Cool. So, on the engineering side, …
162 00:25:15.110 ⇒ 00:25:19.349 Uttam Kumaran: Yeah, I think there’s, like, how do I use cursor….
163 00:25:19.860 ⇒ 00:25:22.809 Uttam Kumaran: Like, how do we use AI to do some basic stuff?
164 00:25:23.810 ⇒ 00:25:27.570 Uttam Kumaran: This is, yeah, like, more cursor setup.
165 00:25:28.000 ⇒ 00:25:28.900 Samuel Roberts: Yeah, definitely.
166 00:25:29.000 ⇒ 00:25:30.790 Uttam Kumaran: So that’s one piece.
167 00:25:31.080 ⇒ 00:25:32.160 Uttam Kumaran: ….
168 00:25:33.860 ⇒ 00:25:42.189 Samuel Roberts: This is… this is from me of being, like, you know, learning about all the tools we’re using, and every once in a while, something else comes up that I’m like, oh, I didn’t even realize we were using that, and, you know…
169 00:25:42.400 ⇒ 00:25:43.160 Samuel Roberts: …
170 00:25:43.560 ⇒ 00:25:47.629 Samuel Roberts: I don’t know if that’s… that’s a combination of things, I could probably break that into a few different things, but….
171 00:25:47.630 ⇒ 00:25:57.699 Uttam Kumaran: No, this is great. I think for this one, I think this is… this is a, like… I guess, Demolati, can you talk about, like, this piece, the understanding business requirements?
172 00:25:59.320 ⇒ 00:26:06.859 Demilade Agboola: So business requirements actually kind of… I guess it falls more under project management, at that point, but basically, sometimes.
173 00:26:07.070 ⇒ 00:26:10.919 Demilade Agboola: The tasks are given, and sometimes it’s not clear.
174 00:26:11.070 ⇒ 00:26:15.300 Demilade Agboola: And it’s on you in engineering to figure that out. …
175 00:26:16.250 ⇒ 00:26:18.790 Demilade Agboola: And then get things out, basically.
176 00:26:18.930 ⇒ 00:26:25.130 Demilade Agboola: For validation of data, I mean, sometimes building code isn’t the hard part, it’s the…
177 00:26:25.300 ⇒ 00:26:34.709 Demilade Agboola: you need that data to be good before it goes out, into the system, and that’s a huge part of it, I guess, so it’s just a…
178 00:26:35.110 ⇒ 00:26:39.309 Demilade Agboola: What happens, like, what happens on a day-to-day from the data team?
179 00:26:39.550 ⇒ 00:26:45.110 Demilade Agboola: … And how do we ensure that those processes are smooth as possible?
180 00:26:47.510 ⇒ 00:26:48.869 Uttam Kumaran: Yeah, so I guess, like.
181 00:26:49.310 ⇒ 00:26:56.920 Uttam Kumaran: I think we’ll come back to that, because I do want to sort of talk about… I’ll show kind of, like, where this is all going, but I do want to start to break down
182 00:26:57.330 ⇒ 00:27:01.479 Uttam Kumaran: Like, the different types of engineering we do, and the common tasks.
183 00:27:01.850 ⇒ 00:27:05.020 Uttam Kumaran: So, I’m gonna put this here as sort of, like.
184 00:27:05.440 ⇒ 00:27:15.060 Uttam Kumaran: I’m just gonna say, like, documentation… finding things… ….
185 00:27:15.410 ⇒ 00:27:17.329 Samuel Roberts: Yeah, yeah, I have a whole….
186 00:27:17.550 ⇒ 00:27:19.700 Uttam Kumaran: Set of ideas there, and then…
187 00:27:19.950 ⇒ 00:27:25.670 Uttam Kumaran: this is gonna… this, I know, you know, only a couple of us put here, but I think this is gonna be just around, like.
188 00:27:25.940 ⇒ 00:27:35.419 Uttam Kumaran: … I’m gonna sort of highlight this as… … like, AI, enabled engineering.
189 00:27:35.680 ⇒ 00:27:36.220 Samuel Roberts: Yeah.
190 00:27:36.670 ⇒ 00:27:38.919 Uttam Kumaran: And so we can talk about that as well.
191 00:27:39.500 ⇒ 00:27:51.699 Uttam Kumaran: On the operations side, yeah, time to onboard new clients. I… I would like… I think this is something worth discussing, but Rico isn’t here, so maybe I’ll kind of leave this for now.
192 00:27:51.780 ⇒ 00:27:54.189 Samuel Roberts: On the sales side, ….
193 00:27:54.530 ⇒ 00:27:55.880 Uttam Kumaran: Yeah, I think….
194 00:27:56.290 ⇒ 00:27:57.880 Demilade Agboola: So, quick question about….
195 00:27:57.880 ⇒ 00:27:58.250 Uttam Kumaran: Yes.
196 00:27:58.250 ⇒ 00:28:00.430 Demilade Agboola: About, time to onboard new clients.
197 00:28:00.560 ⇒ 00:28:07.489 Demilade Agboola: Are you talking about the process in which we get them to sign, or are you talking about the process in which we get to building for them?
198 00:28:08.080 ⇒ 00:28:09.800 Uttam Kumaran: Yeah, it’s like post-signing.
199 00:28:10.710 ⇒ 00:28:14.240 Demilade Agboola: Is that more of an operations thing, or is that a project management thing?
200 00:28:14.860 ⇒ 00:28:22.320 Uttam Kumaran: it’s kind of… there’s, like, a handoff between ops and PM, so there’s kind of this, like, sales to ops.
201 00:28:22.780 ⇒ 00:28:24.729 Uttam Kumaran: the, like, PM handoff.
202 00:28:25.290 ⇒ 00:28:35.399 Uttam Kumaran: that needs to happen, right? And there’s a bunch of things, like Slack channels get created, people get assigned, roadmaps have to get created, meetings have to get booked.
203 00:28:35.730 ⇒ 00:28:41.690 Uttam Kumaran: prep work, like… Agendas, notions, like…
204 00:28:42.050 ⇒ 00:28:44.109 Uttam Kumaran: All this stuff has to happen.
205 00:28:44.290 ⇒ 00:28:48.789 Uttam Kumaran: And right now, I think it’s… Very, very… it’s entirely manual.
206 00:28:49.120 ⇒ 00:28:49.980 Samuel Roberts: Yeah.
207 00:28:50.180 ⇒ 00:29:02.209 Samuel Roberts: what’s interesting about that is, like, some of those things are gonna be exactly the same for certain clients, and some of those things are going to be, like, wildly dependent on the client and project, so… I could see there being, like, a lot of room for, like, certain kinds of automation automatically.
208 00:29:02.760 ⇒ 00:29:10.919 Samuel Roberts: But also, like, you know, maybe different types of, you know, flows for different clients and different types of work. Yeah, I think there’s…
209 00:29:11.560 ⇒ 00:29:13.040 Samuel Roberts: Yeah, interesting.
210 00:29:14.350 ⇒ 00:29:14.910 Uttam Kumaran: Okay.
211 00:29:18.690 ⇒ 00:29:23.090 Uttam Kumaran: Okay, so I’m gonna just… I’m gonna leave it here for now.
212 00:29:23.410 ⇒ 00:29:27.459 Uttam Kumaran: … So, yes, I think there is…
213 00:29:31.100 ⇒ 00:29:36.160 Uttam Kumaran: Yeah, there’s, like… sort of, like, sales follow-ups, and I know we don’t have
214 00:29:36.880 ⇒ 00:29:41.430 Uttam Kumaran: We mainly have Ryan here from the sales side, but yes, these are all…
215 00:29:41.670 ⇒ 00:29:45.369 Uttam Kumaran: Really good problems, so this is, like, sales momentum.
216 00:29:46.850 ⇒ 00:29:50.730 Uttam Kumaran: … There’s also, like…
217 00:29:50.880 ⇒ 00:29:57.550 Uttam Kumaran: Like, yeah, this sort of stuff, which is, like, kind of like sales operations, which is, like, contract development.
218 00:29:58.020 ⇒ 00:29:59.499 Uttam Kumaran: Things like that. Cool.
219 00:30:00.740 ⇒ 00:30:04.950 Uttam Kumaran: … And then there’s also stuff that’s, like, …
220 00:30:07.440 ⇒ 00:30:15.649 Uttam Kumaran: these are, like, kind of like sales rituals, like, asking our existing clients for referrals, getting updates on things in HubSpot.
221 00:30:15.800 ⇒ 00:30:24.729 Uttam Kumaran: These are just things that have to happen every day automatically. And currently, it’s falling under one or few people, and…
222 00:30:24.900 ⇒ 00:30:31.909 Uttam Kumaran: It gets dropped, or other… it’s just not… it’s just not a good use of folks’ time to do this when automation exists.
223 00:30:32.720 ⇒ 00:30:33.620 Uttam Kumaran: …
224 00:30:34.110 ⇒ 00:30:40.549 Uttam Kumaran: And then, yeah, this one is overall just, like, how do we improve? This is, like, kind of like recruiting ops.
225 00:30:41.060 ⇒ 00:30:46.960 Uttam Kumaran: Like, how do we… for me, the biggest thing is, like, how do we give every candidate a great experience?
226 00:30:47.070 ⇒ 00:30:52.269 Uttam Kumaran: How do we improve the speed at which we can come to a decision on a candidate?
227 00:30:52.520 ⇒ 00:30:58.289 Uttam Kumaran: … And… how do we shorten the time between, like.
228 00:30:58.560 ⇒ 00:31:01.660 Uttam Kumaran: Sort of screening and… and a decision, so…
229 00:31:02.110 ⇒ 00:31:10.329 Uttam Kumaran: These are all things that I’m starting to think of. We’re just going through a big wave of hiring for this PM role right now, and so…
230 00:31:10.530 ⇒ 00:31:14.470 Uttam Kumaran: Kind of deciding on… Some of these will come up.
231 00:31:14.760 ⇒ 00:31:20.220 Uttam Kumaran: But maybe today, I think, with this crew, while we have 10 more minutes.
232 00:31:20.440 ⇒ 00:31:31.689 Uttam Kumaran: One thing that I kind of wanted to… to talk a little bit about was just, like, what are the engineering things that we all do on a day-to-day basis?
233 00:31:31.910 ⇒ 00:31:35.919 Uttam Kumaran: And maybe I can, you know, I know we have…
234 00:31:36.230 ⇒ 00:31:42.429 Uttam Kumaran: some representation from… looks like almost every team, like Annie, Demolade, Vashdev.
235 00:31:42.550 ⇒ 00:31:48.439 Uttam Kumaran: maybe some of the things I would like to hear, and I can start to write them down as we go, is just, like.
236 00:31:48.740 ⇒ 00:31:52.729 Uttam Kumaran: Maybe you can just describe, like, a day in the life
237 00:31:53.020 ⇒ 00:31:55.340 Uttam Kumaran: And I can start to know down
238 00:31:56.140 ⇒ 00:32:14.669 Uttam Kumaran: some of the notable, painful things that you have to do as part of your role on one or many clients. Maybe, Annie, I’ll pick on you first. We can sort of talk at the top of the stack on the analyst analysis side, and then kind of move down.
239 00:32:15.020 ⇒ 00:32:26.250 Uttam Kumaran: So maybe if you have a couple of pain points or things that you would say, like, okay, maybe there’s an opportunity to… to improve this or to innovate, or maybe there’s something where I can get some help to do this.
240 00:32:26.610 ⇒ 00:32:28.819 Uttam Kumaran: Would love to, kind of, hear those.
241 00:32:33.540 ⇒ 00:32:34.730 Annie Yu: ….
242 00:32:34.790 ⇒ 00:32:38.210 Uttam Kumaran: I’m gonna take a moment to think through that.
243 00:32:38.210 ⇒ 00:32:39.540 Annie Yu: Okay. Okay.
244 00:32:39.540 ⇒ 00:32:51.089 Uttam Kumaran: Yeah, take a sec. Maybe I’ll move to, Demolade. If there’s anything in, like, AE world, or DE World, or even client world that you think
245 00:32:51.350 ⇒ 00:32:54.940 Uttam Kumaran: There’s gotta be a better way, or this is a huge time suck.
246 00:32:55.350 ⇒ 00:32:58.590 Uttam Kumaran: … Would love.
247 00:32:58.590 ⇒ 00:32:59.700 Demilade Agboola: …
248 00:33:01.090 ⇒ 00:33:12.660 Demilade Agboola: So, it’s just like, okay, I think one of the things that can be a huge time blocker is documentation, to be honest. That’s always… I think that’s very clear.
249 00:33:12.940 ⇒ 00:33:20.190 Demilade Agboola: Another potential thing that can… take time.
250 00:33:20.580 ⇒ 00:33:37.660 Demilade Agboola: is, just basically testing, which kind of, I mentioned the data validation part. Yeah. Because you… a lot of the things you do are… you’re building models, or tests, or, you know, creating new… modifying existing models.
251 00:33:37.910 ⇒ 00:33:43.410 Demilade Agboola: And you have to ensure that the numbers are, like, the numbers make sense, you’ve not exploded the table.
252 00:33:43.710 ⇒ 00:33:49.020 Demilade Agboola: Like, things are as they should be, and you don’t necessarily want that to go out.
253 00:33:49.380 ⇒ 00:33:53.759 Demilade Agboola: … And I’ll tell you just…
254 00:33:54.420 ⇒ 00:33:58.699 Demilade Agboola: This isn’t necessarily, like, a technical thing, but it’s just the ad hoc-ness of certain things.
255 00:33:58.920 ⇒ 00:34:01.930 Demilade Agboola: So, how people, like, reach out to you?
256 00:34:02.080 ⇒ 00:34:05.810 Demilade Agboola: And there’s, like, a lot that tends to happen in that regard.
257 00:34:06.490 ⇒ 00:34:14.989 Demilade Agboola: … On a day-to-day… I think those are the tricky parts.
258 00:34:15.300 ⇒ 00:34:23.989 Demilade Agboola: Like, once you… once you build out the documentation for stuff, the diagrams, the, you know, case notes, the…
259 00:34:24.130 ⇒ 00:34:27.770 Demilade Agboola: You know, pros and cons list of everything you’ve done.
260 00:34:28.050 ⇒ 00:34:33.039 Demilade Agboola: And you handled the testing, most of the actual, like.
261 00:34:33.150 ⇒ 00:34:39.080 Demilade Agboola: writing of the code is fine. I mean, there’s some complex things from time to time.
262 00:34:39.560 ⇒ 00:34:43.010 Demilade Agboola: Where it’s like, oh, how do we handle that?
263 00:34:43.540 ⇒ 00:34:52.729 Demilade Agboola: But I guess for the most part, those things, especially, like, Cursor and, like, ChatGPT and Google, you can kind of get an idea of how to…
264 00:34:52.880 ⇒ 00:34:59.130 Demilade Agboola: To get ahead of those sort of things, and get an idea of how people handle some of those things in certain scenarios.
265 00:35:00.000 ⇒ 00:35:06.619 Demilade Agboola: I think those will be the large blockers for a lot of large things that people run into on a day.
266 00:35:08.700 ⇒ 00:35:09.270 Uttam Kumaran: Okay.
267 00:35:09.760 ⇒ 00:35:14.279 Uttam Kumaran: Yeah, so I think data validation and testing makes a lot of sense. I think, …
268 00:35:15.730 ⇒ 00:35:20.440 Uttam Kumaran: Yes, the documentation piece also is, like, really, really huge.
269 00:35:24.490 ⇒ 00:35:28.410 Uttam Kumaran: … Okay, maybe I can, …
270 00:35:29.410 ⇒ 00:35:33.819 Uttam Kumaran: Maybe I can ask Vashtav, anything on the data engineering side?
271 00:35:34.380 ⇒ 00:35:37.380 Uttam Kumaran: That you would say, like, is pretty painstaking.
272 00:35:37.660 ⇒ 00:35:42.609 Uttam Kumaran: I can think of a couple, but maybe I’d love to hear any thoughts.
273 00:35:44.120 ⇒ 00:35:48.300 Vashdev Heerani: So, for me, hello, Arun. So, for me, …
274 00:35:49.740 ⇒ 00:36:04.300 Vashdev Heerani: Context understanding things, like, there’s a lot of different things coming to me at the same time, so context switching is kind of, kind of very pain point for me.
275 00:36:04.340 ⇒ 00:36:14.269 Vashdev Heerani: Other than that, yes, data validation and data testing is also another thing for me that take a lot of my time.
276 00:36:14.480 ⇒ 00:36:15.230 Vashdev Heerani: …
277 00:36:15.910 ⇒ 00:36:27.099 Vashdev Heerani: Other than that, I got Aviation by side, so he’s very helpful for me to understand, different things on the different platforms.
278 00:36:27.720 ⇒ 00:36:28.390 Uttam Kumaran: Okay.
279 00:36:28.980 ⇒ 00:36:29.640 Vashdev Heerani: No.
280 00:36:35.100 ⇒ 00:36:38.900 Uttam Kumaran: I think one thing for me here is, like.
281 00:36:39.750 ⇒ 00:36:44.020 Uttam Kumaran: time to triage and resolve DE issues.
282 00:36:44.330 ⇒ 00:36:48.889 Uttam Kumaran: I’m thinking really about, like, DACs or things failing?
283 00:36:49.160 ⇒ 00:36:52.750 Uttam Kumaran: You know, data refresh issues.
284 00:36:52.900 ⇒ 00:37:00.209 Uttam Kumaran: So, like, I just want AI to take a first pass, which is, like, take a look at the logs, take a look at the code.
285 00:37:00.530 ⇒ 00:37:03.790 Uttam Kumaran: Tell me what’s… tell me what I should think about.
286 00:37:04.020 ⇒ 00:37:07.640 Uttam Kumaran: And I want that for every type of DE ticket, you know, because
287 00:37:07.860 ⇒ 00:37:16.179 Uttam Kumaran: compared to the analysis and modeling work, like, I think it’s pretty structured in that usually the solution is in the log somewhere.
288 00:37:16.380 ⇒ 00:37:20.329 Uttam Kumaran: So, I want to see how we can start to have AI
289 00:37:20.480 ⇒ 00:37:23.619 Uttam Kumaran: Take a look at those logs and actually give feedback.
290 00:37:24.470 ⇒ 00:37:28.340 Uttam Kumaran: You know, would be my… sort of pitch
291 00:37:31.000 ⇒ 00:37:35.009 Uttam Kumaran: I guess I’ll come back to you, Annie, if you have any….
292 00:37:37.060 ⇒ 00:37:40.939 Annie Yu: On the engineering side, I would say… and this…
293 00:37:42.120 ⇒ 00:37:50.110 Annie Yu: Also, I wouldn’t say it’s a big deal, but sometimes when I get, like, an ad hoc request, I know that I have to write some kind of query.
294 00:37:50.190 ⇒ 00:38:03.039 Annie Yu: But I’m not sure what data field is in, like, which tables I could use, so I eventually still have to come to AE or DE, to clarify some of those. So that, that’s, …
295 00:38:03.380 ⇒ 00:38:11.250 Annie Yu: something that takes time. And then another thing, I probably don’t have, like, a best solution now, but it’s the….
296 00:38:11.250 ⇒ 00:38:13.800 Uttam Kumaran: That’s fine, you don’t need a solution at all, I just want to hear.
297 00:38:13.800 ⇒ 00:38:19.570 Annie Yu: Yeah. Yeah, so, so it’s the, … The thing with Tableau…
298 00:38:19.740 ⇒ 00:38:30.109 Annie Yu: some of the new requests, like, even I don’t know if I can do that in Tableau without modeling changes.
299 00:38:30.180 ⇒ 00:38:45.500 Annie Yu: So I really just have to spend time exploring solutions to see if it’s doable in Tableau, and sometimes I realize, okay, I can do that on my end. But then sometimes I realize, okay, this has to be done in the back end, and then maybe a day has passed.
300 00:38:45.760 ⇒ 00:38:46.900 Annie Yu: Because of that.
301 00:38:48.800 ⇒ 00:38:52.240 Uttam Kumaran: Can you, can you, like… give an example?
302 00:38:52.240 ⇒ 00:39:10.860 Annie Yu: Like, recently, they said… so, like, Egan team looks at their time duration from one point to another, so maybe, like, orders sent to pharmacy to when they were shipped by the pharmacy, and they said, like, okay, we want that to fix on, like.
303 00:39:10.910 ⇒ 00:39:12.389 Annie Yu: Business hours.
304 00:39:12.750 ⇒ 00:39:20.569 Annie Yu: So that’s something that I’ve never done with Tableau, and I don’t know… like, I imagine I could do that, but then I realized
305 00:39:20.710 ⇒ 00:39:23.129 Annie Yu: Okay, I really don’t have a clean way to do that.
306 00:39:23.130 ⇒ 00:39:28.980 Uttam Kumaran: I see, I see. So it’s kind of like… it’s kind of similar to what Voshdev said, which is, like.
307 00:39:29.370 ⇒ 00:39:32.050 Uttam Kumaran: Tableau or Daxter, it’s like…
308 00:39:33.070 ⇒ 00:39:38.470 Uttam Kumaran: you know, how can we just, like, have an AI help us with tool-specific problems, right? Or, like.
309 00:39:38.710 ⇒ 00:39:41.290 Uttam Kumaran: I don’t think that’s the best way of putting it, but like…
310 00:39:41.810 ⇒ 00:39:50.949 Uttam Kumaran: how can we have AI sort of give suggestions, or almost have a co-pilot for every tool? One thing that we had an idea on, and it’s in the backlog, is I…
311 00:39:51.110 ⇒ 00:40:08.549 Uttam Kumaran: I wanted to scrape all of the documentation for, like, Snowflake, GCP, and basically have an agent for every tool, where if you have a question, right now, I Ctrl-F and I Google. Instead, it would be great to just ask our GCP agent, how do I do this?
312 00:40:08.600 ⇒ 00:40:14.290 Uttam Kumaran: Right? Or similarly, we have a Tableau agent that gives suggestions on how to do things.
313 00:40:14.660 ⇒ 00:40:23.640 Uttam Kumaran: And it’s… it could give visual examples, it can refer to docs, it could give examples from the web. So I… I would say this is more, like, tool-specific.
314 00:40:24.010 ⇒ 00:40:25.300 Uttam Kumaran: how-tos?
315 00:40:26.140 ⇒ 00:40:37.209 Uttam Kumaran: That usually have to get solved via trial… Error… Or… Googling? Yeah, go ahead.
316 00:40:37.210 ⇒ 00:40:40.700 Demilade Agboola: question with that. … doesn’t, like…
317 00:40:42.050 ⇒ 00:40:46.610 Demilade Agboola: ChatGPT and just AI in general tend to hallucinate some responses sometimes.
318 00:40:46.930 ⇒ 00:40:54.870 Demilade Agboola: So, like, for instance, if you were like, oh, can I do this in Tableau? It might say you can, but then you try it and it can’t. It doesn’t work.
319 00:40:54.870 ⇒ 00:41:03.259 Uttam Kumaran: Yes. So there’s a couple ways we kind of solve for that. One is, we force the AI to have a ground truth.
320 00:41:03.370 ⇒ 00:41:09.890 Uttam Kumaran: Right, so… the kind of… a common example is, like, when you ask a question, like, hey.
321 00:41:10.120 ⇒ 00:41:21.189 Uttam Kumaran: how do I do this in Redshift? And refer back to the documentation that confirms that this is possible. So one… that’s a good thing, is that, maybe I’m gonna put this…
322 00:41:21.390 ⇒ 00:41:32.569 Uttam Kumaran: Just here, which is, like… and then, like, how can all answers be rooted I’ll be referencing
323 00:41:32.870 ⇒ 00:41:35.399 Uttam Kumaran: A doc or a how-to article.
324 00:41:35.770 ⇒ 00:41:40.079 Uttam Kumaran: You’re exactly right, it’s like, the AI currently, it will hallucinate.
325 00:41:40.420 ⇒ 00:41:47.870 Uttam Kumaran: And so one way of combating against that is, one, reducing the scope of, like, what the AI is tasked to do.
326 00:41:48.300 ⇒ 00:41:52.639 Uttam Kumaran: Second, it’s also making sure that for any answer, there has to be a reference.
327 00:41:52.880 ⇒ 00:41:56.340 Uttam Kumaran: Reference a doc, or to a video, or something, so…
328 00:41:57.830 ⇒ 00:42:07.680 Uttam Kumaran: So that makes sense. I mean, a couple of things on the overall process, too, is, like, I agree that ad hoc… at the top of the funnel here, there’s so much ad hoc that comes in.
329 00:42:08.000 ⇒ 00:42:14.440 Uttam Kumaran: And I really think that there has to be a… and I don’t think the PMs are ever going to be equipped
330 00:42:14.780 ⇒ 00:42:17.710 Uttam Kumaran: to actually… …
331 00:42:18.400 ⇒ 00:42:28.060 Uttam Kumaran: you know, like, do this discovery where the ticket comes to this team really clean. One of the things that I can think about is, like, hey, why don’t we take the request.
332 00:42:28.470 ⇒ 00:42:35.529 Uttam Kumaran: run it through Cursor with access to our data models and have it sort of produce, like, how… how could you actually
333 00:42:36.040 ⇒ 00:42:41.260 Uttam Kumaran: run a discovery for this, right? So at least that saves a little bit of time.
334 00:42:41.450 ⇒ 00:42:51.630 Uttam Kumaran: And it allows, I think, on the analyst team to maybe get a little bit further before handing it to AEs. I don’t know where this happens, but it’s one thing that we can think about.
335 00:42:51.850 ⇒ 00:42:58.939 Uttam Kumaran: I also think that some of these tool-specific sort of co-pilots are things that we can consider as well.
336 00:42:59.160 ⇒ 00:43:05.399 Uttam Kumaran: … I also think, yes, the documentation side is something that we really struggle with.
337 00:43:05.840 ⇒ 00:43:06.760 Uttam Kumaran: …
338 00:43:07.190 ⇒ 00:43:13.399 Uttam Kumaran: I don’t know, I… in traditional engineering teams, you try to bucket some amount of time for documentation, but…
339 00:43:13.600 ⇒ 00:43:28.659 Uttam Kumaran: It’s just something that’s going to be very, very hard for us. And we have AI, right? So my… my sort of thought process and interest is, how can we use AI over our transcripts to start to pull out opportunities for documentation?
340 00:43:28.780 ⇒ 00:43:32.659 Uttam Kumaran: Because there’s a guarantee we’ve talked about this stuff in a meeting or in Slack.
341 00:43:33.100 ⇒ 00:43:51.260 Uttam Kumaran: It’s something that I want, you know, the AI folks on this call to sort of think about, which is, like, how do we start building more SOPs? How do we start using the transcripts of our meetings to help us build our library of documents and the way things are being done more dynamically?
342 00:43:51.380 ⇒ 00:43:56.609 Uttam Kumaran: … you know, I don’t see us solving this documentation problem
343 00:43:56.770 ⇒ 00:44:01.990 Uttam Kumaran: Meaning people have time to do this anytime soon. That’s a great constraint to have here.
344 00:44:03.490 ⇒ 00:44:13.699 Uttam Kumaran: Cool, okay, this is really, really helpful. I… kind of the reason for the color coding, we’re going to start to look at things, sort of, process by process, and so that kind of goes into my next…
345 00:44:13.950 ⇒ 00:44:20.409 Uttam Kumaran: … My next section, briefly, I just wanted to share
346 00:44:20.560 ⇒ 00:44:27.519 Uttam Kumaran: some of the work that Sam and the team has been doing on our AI,
347 00:44:27.750 ⇒ 00:44:47.289 Uttam Kumaran: sort of platform. So, one of the things that we’re doing is working on this AI platform architecture. And really, this is, like, every place in the company we’re using AI, how… what are our goals, and, you know, how we can start to actually measure and achieve some of these. So, I fill out this top section. These are kind of, like, what I immediately think about, which is, like.
348 00:44:47.390 ⇒ 00:44:54.590 Uttam Kumaran: I want to look at, are people spending less time in meetings? Are generally our clients and team members happy?
349 00:44:54.710 ⇒ 00:45:06.559 Uttam Kumaran: Are people using our AI workflows, indicating that maybe they are helpful? We also have, like, core stakeholders, so we have, like, sales, finance, operations, marketing, engineering, PM,
350 00:45:06.720 ⇒ 00:45:24.330 Uttam Kumaran: And we’re starting to go through process. So, really, what we’re gonna start to do is break down every core process, you know, in the company for these areas. And another thing that we’re gonna add here is both, like, what KPI it ties to, and …
351 00:45:24.390 ⇒ 00:45:43.470 Uttam Kumaran: how we can start to look at, generally, scores per department on how augmented they are with AI. So, an example is, like, project management. A couple of things listed here is, like, can we use AI to see all the meetings associated with a client? Creating project charters, creating project plan, maybe it’s to help grooming.
352 00:45:43.640 ⇒ 00:45:53.629 Uttam Kumaran: For operations, one of the things I’m thinking about, is, like, how can we, use our calendar data to actually help update Clockify?
353 00:45:53.790 ⇒ 00:46:03.899 Uttam Kumaran: Or how can we build some type of more smart way of identifying when people are working on clients, so it saves some time in noting down hours?
354 00:46:04.090 ⇒ 00:46:22.269 Uttam Kumaran: Similarly on the engineering side, reviewing PR, creating PR description, executing tickets, chatting over documentation, planning TDDs. So these are all processes that we’re going to start to know down, understand if AI is being used today, and if not, what… how do we start to impact these?
355 00:46:22.380 ⇒ 00:46:23.890 Uttam Kumaran: So this doc is…
356 00:46:24.080 ⇒ 00:46:44.000 Uttam Kumaran: in Notion, feel free to take a look at it. We’re going to be kind of going through and building this out, and then additionally, we have our existing platform. So what are the existing agents, metrics, tools that we’re using? This includes things like Cursor, Granola, other tools, where… and then, of course, like, technically, like, where is all the code, and how are we building this?
357 00:46:44.360 ⇒ 00:46:47.310 Uttam Kumaran: So this is gonna kind of be the Bible for…
358 00:46:47.420 ⇒ 00:47:03.180 Uttam Kumaran: sort of our AI efforts and how we make sure that this solves it. But everybody on this call, you guys are the stakeholders. So, one thing that I want you guys to all put pressure on this team is to actually make sure that the things we develop
359 00:47:03.350 ⇒ 00:47:18.459 Uttam Kumaran: are impacting you. Like, I’m not here to develop stuff that is nice to have or looks flashy. We want to make sure that the things we’re issuing work for you, and so it’s actually really, really important that you indicate, hey, I know
360 00:47:18.820 ⇒ 00:47:33.520 Uttam Kumaran: I know we can use cursor, but no idea how to use it, or it’s not working. Or, hey, I wanted to use one of our agents, and it wasn’t working. Or, like, it’s not able to do this. That’s it, you don’t have to figure it out at that point, we just need to hear the problems fast.
361 00:47:36.400 ⇒ 00:47:38.750 Annie Yu: Well, Tom, I have a note….
362 00:47:38.900 ⇒ 00:47:39.530 Uttam Kumaran: leaf.
363 00:47:39.770 ⇒ 00:47:45.820 Annie Yu: So, I think this is all great, but I think… sometimes… …
364 00:47:46.820 ⇒ 00:48:00.400 Annie Yu: like, my use case would be, like, getting a ticket, and it’s obviously that something’s from AI, but then when I verify against what stakeholder is saying, it’s… it’s… it’s not it. So I think I… my point is.
365 00:48:00.930 ⇒ 00:48:07.710 Annie Yu: I think using AI is for sure helpful, but I think it’s also important to verify what it’s saying.
366 00:48:08.860 ⇒ 00:48:10.480 Uttam Kumaran: Yeah, so I guess, like, let’s…
367 00:48:10.590 ⇒ 00:48:13.439 Uttam Kumaran: If we can go one step further on that.
368 00:48:13.780 ⇒ 00:48:18.040 Uttam Kumaran: there’s, like, two parts to that. One is, like, the ticket quality, right? So…
369 00:48:18.150 ⇒ 00:48:25.780 Uttam Kumaran: Right now, I know a lot of the tickets that are being groomed are being done using AI directly from a transcript, so the quality is not good.
370 00:48:26.330 ⇒ 00:48:34.259 Uttam Kumaran: So that would be, like, an example of, like, hey, the tickets I’m getting, although, like, it’s better than nothing, they don’t have all the detail.
371 00:48:34.370 ⇒ 00:48:39.079 Uttam Kumaran: Is that, like, a good… would that be, like, a good, like, first part of that point?
372 00:48:39.080 ⇒ 00:48:41.519 Annie Yu: Yeah, yeah, I think you put it perfectly.
373 00:48:41.520 ⇒ 00:48:51.029 Uttam Kumaran: Exactly. So, but that’s an example of, like, you said it way too nicely. Where… where, you know, I think you’re right in that
374 00:48:51.090 ⇒ 00:49:04.140 Uttam Kumaran: we should look at anything that comes out of AI, but actually the core problem is that the way we go from meeting transcripts to tickets, the output tickets are not accurate.
375 00:49:04.270 ⇒ 00:49:17.870 Uttam Kumaran: And that is the problem, right? So now, for the AI team, okay, it’s up to us to go dissect. Where do the inaccuracies come from? Do we have not enough… is there not enough info in the transcript? Is the way we’re writing the transcript verbose?
376 00:49:18.120 ⇒ 00:49:19.010 Uttam Kumaran: …
377 00:49:19.160 ⇒ 00:49:35.009 Uttam Kumaran: do we have to require something that says, hey, the AI can push back and say, we don’t have enough information, half of this is hallucination, I need more info before I can create this, right? So those are, like, some solutions, but it’s actually very helpful
378 00:49:35.100 ⇒ 00:49:39.649 Uttam Kumaran: To hear that, like, hey, some of the tickets are okay, but some of them are just…
379 00:49:40.110 ⇒ 00:49:57.439 Uttam Kumaran: trash. Like, they’re just… they just have a lot of jargon and a lot of, like, big words. That’s what’s perfect to hear, and so for us, it’s also, one, is thinking about a feedback loop, right? So how do we… how can you actually look at those tickets, understand that it was AI-written, and then also say, like, hey, this isn’t good enough, and why?
380 00:49:57.520 ⇒ 00:50:02.280 Uttam Kumaran: Then, for the AI team, it’s to take that information back and refine that process.
381 00:50:02.390 ⇒ 00:50:03.400 Uttam Kumaran: …
382 00:50:07.130 ⇒ 00:50:08.820 Uttam Kumaran: But great, that’s great feedback.
383 00:50:10.400 ⇒ 00:50:23.420 Uttam Kumaran: I do think that that, you know, the transcript or Slack to tickets is a huge unlock. Like, I know it saved the team a lot of time, but I also agree the quality is not good enough.
384 00:50:23.680 ⇒ 00:50:30.410 Uttam Kumaran: So, as an AI team, we have to think about how do we improve the quality? Are we lacking on the input side, on the prompt side?
385 00:50:30.570 ⇒ 00:50:33.870 Uttam Kumaran: On the, you know, like, what access to data we have.
386 00:50:34.460 ⇒ 00:50:36.319 Uttam Kumaran: But that’s something that I think
387 00:50:36.510 ⇒ 00:50:38.949 Uttam Kumaran: We need to improve on, for sure.
388 00:50:43.270 ⇒ 00:50:51.599 Uttam Kumaran: Okay, perfect. And so, one thing I just wanted to maybe close out and share is a couple of things that I have been doing, you know, this week.
389 00:50:51.950 ⇒ 00:50:58.100 Uttam Kumaran: with AI, and sort of give people some ideas about how you could use this …
390 00:50:58.280 ⇒ 00:51:01.149 Uttam Kumaran: In your life, too.
391 00:51:01.340 ⇒ 00:51:05.610 Uttam Kumaran: So, one of the things that, I was tasked
392 00:51:06.110 ⇒ 00:51:15.229 Uttam Kumaran: To do yesterday was we had a potential partner ask us to help produce a,
393 00:51:15.450 ⇒ 00:51:21.810 Uttam Kumaran: To help produce a… Proposal document to assist with one of their clients.
394 00:51:22.330 ⇒ 00:51:28.900 Uttam Kumaran: And so… we were actually… I looked through my list of, sort of, prompts.
395 00:51:29.080 ⇒ 00:51:35.949 Uttam Kumaran: All of which is available in the prompt library in Notion, and I realized I didn’t have anything that was sort of like a solutions architect.
396 00:51:36.080 ⇒ 00:51:42.800 Uttam Kumaran: you know, I have a prompt for, like, helping me think through partnerships, or sales, or writing emails, but…
397 00:51:43.080 ⇒ 00:51:59.580 Uttam Kumaran: Thinking through, like, hey, how do we actually put together, like, a technical proposal with something we didn’t have? And so one of the things that I did, every time I need to create a new prompt is I used our prompt writer. And so one of the… so all I did is I sent it, this.
398 00:51:59.720 ⇒ 00:52:13.850 Uttam Kumaran: So, I sent, and to give you a sense of what the prompt writer is, this is the prompt for it. You are an AI system that helps craft high-leverage prompts for GBT-5. The core instructions are understand the request, apply best practices.
399 00:52:14.020 ⇒ 00:52:19.170 Uttam Kumaran: Create the blueprint, Ask follow-up questions, and here’s an example.
400 00:52:19.380 ⇒ 00:52:36.919 Uttam Kumaran: And so, this is actually, like, how I… it’s actually a little bit of, like, how I use it to create prompts that I need. One of the prompts is for the solutions architect, and so what I did is I basically said, hey, here’s… I need to help creating a prompt that acts as a senior solutions architect and sales engineer.
401 00:52:37.010 ⇒ 00:52:47.620 Uttam Kumaran: Where we’re gonna help us scope projects that we’re gonna work on up with a partner on, or take on ourselves. This is actually something that I do on many of our, you know, leads that come in, is I act as a solutions architect, which is, like.
402 00:52:47.810 ⇒ 00:53:03.519 Uttam Kumaran: what are… what is a high-level deliverable here? I said, I’m going to be able to send things like scopes, pricing, transcripts, documents, and I need help putting together scopes, pricing estimates. I said, it’s kind of like sales engineering and CTO work, but kind of like solutions architect.
403 00:53:03.790 ⇒ 00:53:10.049 Uttam Kumaran: but also, like, sales-focused. I was kind of rambling at this point, and I said, if you have follow-up questions, let me know.
404 00:53:10.420 ⇒ 00:53:18.449 Uttam Kumaran: And it’s like, great, you want a drop-in prompt that turns GPT-5 into a cleaner solutions architect to make this effective. I do a few things before I draft.
405 00:53:18.730 ⇒ 00:53:25.029 Uttam Kumaran: The output format, pricing logic, level of polish, tone, and context.
406 00:53:25.320 ⇒ 00:53:35.809 Uttam Kumaran: So then I answered. I said, yes, I do think it should produce a structured proposal, but it should also prompt me for follow-up questions until there’s enough information to produce that.
407 00:53:35.950 ⇒ 00:53:43.649 Uttam Kumaran: And then I basically said, like, here are all our rates, here’s all of our services, like, just copy this straight from Notion.
408 00:53:43.770 ⇒ 00:53:52.790 Uttam Kumaran: I also gave it a list of, like, how we typically deliver, and it said, perfect. Like, that’s enough what I need to build this, and it built this prompt.
409 00:53:53.130 ⇒ 00:54:03.670 Uttam Kumaran: Which is great. So, the thing is, this is something that I… we do on a weekly basis, and so instead of, like, just creating a new chat window and asking it, I now want
410 00:54:03.820 ⇒ 00:54:05.810 Uttam Kumaran: more tailored
411 00:54:05.950 ⇒ 00:54:16.760 Uttam Kumaran: solutions architect prompt that helps me think through this. This is what I would consider level one automation, right? Level 0, I would say, is just asking ChatGPT.
412 00:54:16.910 ⇒ 00:54:29.220 Uttam Kumaran: Like, just clicking new and saying, hi ChatGPT, I need help doing this. Level 2 is, like, okay, we have use case-specific prompts, right? And so for this, what I… what I asked it is, I said.
413 00:54:29.550 ⇒ 00:54:35.579 Uttam Kumaran: at the top, I said, okay, this is an inbound lead from a partner of ours, and actually, as you can see, there’s a lot
414 00:54:35.880 ⇒ 00:54:43.399 Uttam Kumaran: there’s a lot that I’ve written here, but one thing that I also do is I use Whisper, which I think I’ve shared before, which is great text-to-speech.
415 00:54:43.730 ⇒ 00:54:45.810 Uttam Kumaran: You can literally just open this.
416 00:54:46.060 ⇒ 00:55:03.009 Uttam Kumaran: just talk directly into the microphone, and then end it, and it’ll paste it right in. So, it’s so much faster than me typing sometimes, so I use this quite often. If anyone wants this, happy to get it for you. It’s been, like.
417 00:55:03.160 ⇒ 00:55:06.090 Uttam Kumaran: extremely helpful for me. …
418 00:55:06.380 ⇒ 00:55:18.339 Uttam Kumaran: And then I just sent it some of the documents we got, some of the slacks we got from them, and then it said, cool, here are the clarifying questions. As you remember, the prompt said to ask questions. So it asked me, like, a bunch of questions.
419 00:55:18.520 ⇒ 00:55:19.490 Uttam Kumaran: I…
420 00:55:19.910 ⇒ 00:55:28.830 Uttam Kumaran: you know, word vomited, a bunch of stuff about it, perfect. It did the first draft. This was probably, like, 80% there.
421 00:55:29.180 ⇒ 00:55:34.930 Uttam Kumaran: I then edited it, throw it to a Google Doc, … and…
422 00:55:35.160 ⇒ 00:55:38.549 Uttam Kumaran: Shipped it, and this took me 30 minutes.
423 00:55:38.850 ⇒ 00:55:40.869 Uttam Kumaran: I could not have written this
424 00:55:41.150 ⇒ 00:55:47.070 Uttam Kumaran: in this amount, it would have taken me at least a few hours to have done this, of, like, heads-down time to write this.
425 00:55:47.250 ⇒ 00:55:53.830 Uttam Kumaran: But you can see it’s a… I didn’t just take these docs, send it to ChatGPC, say, write me a solution.
426 00:55:54.060 ⇒ 00:56:08.200 Uttam Kumaran: I, like, methodically understood that, hey, we need a solution for a partner. The partner is selling to another client. It needs to be something that they can then take and adapt, and ask me any other questions you have.
427 00:56:08.350 ⇒ 00:56:12.169 Uttam Kumaran: And so you can see, like, it went from probably the answer would have been, like.
428 00:56:12.320 ⇒ 00:56:17.160 Uttam Kumaran: 20 or 30% accurate, to now, it basically was 80% accurate.
429 00:56:17.350 ⇒ 00:56:25.470 Uttam Kumaran: So this is when folks say, like, hey, the AI, like, didn’t get it right, my challenge back is that you have to spend a little bit of time.
430 00:56:25.560 ⇒ 00:56:38.520 Uttam Kumaran: Like, we have to actually spend time creating these prompts and give it a lot of information, but the outputs can be really good. Using things like speech-to-text, copying and pasting our docs is a great way of achieving that.
431 00:56:39.050 ⇒ 00:56:39.810 Uttam Kumaran: …
432 00:56:40.350 ⇒ 00:56:46.960 Uttam Kumaran: You know, so I use this all the time, like, I use this to create tickets, to edit our content.
433 00:56:47.290 ⇒ 00:56:54.130 Uttam Kumaran: To help me write sales, like, for example, … This was, like.
434 00:56:54.590 ⇒ 00:56:59.409 Uttam Kumaran: I was like, here’s a whole thing I wanted to send to a client, can you just, like.
435 00:56:59.600 ⇒ 00:57:05.520 Uttam Kumaran: check if there’s any grammar issues, or if there’s anything, like, I should edit. Perfect. And it, like, it made some great edits.
436 00:57:05.970 ⇒ 00:57:11.569 Uttam Kumaran: … And then I have things that help me with strategy, with writing project documents.
437 00:57:11.740 ⇒ 00:57:26.060 Uttam Kumaran: So all of these, additionally, we’re going to start to make available in the platform, but I highly encourage you to start thinking about, sort of, some of these common things you ask AI, and try your best to create use case-specific prompts
438 00:57:26.220 ⇒ 00:57:28.070 Uttam Kumaran: For them.
439 00:57:28.770 ⇒ 00:57:35.099 Uttam Kumaran: Just wanted to pause there. Is this, like, going over everybody’s heads, or is everyone, like, with me on, like.
440 00:57:36.450 ⇒ 00:57:43.860 Uttam Kumaran: using the prompt writer to create a prompt, and then use it to get… achieve your objective? Can I, like, answer any questions?
441 00:57:52.740 ⇒ 00:57:58.540 Demilade Agboola: I think my only question is, can you do this within the web app, or is this largely a thing for…
442 00:57:59.780 ⇒ 00:58:03.689 Demilade Agboola: For the, like, The actual desktop app.
443 00:58:03.980 ⇒ 00:58:07.760 Uttam Kumaran: You can do this in the web app, so let me, …
444 00:58:11.260 ⇒ 00:58:20.429 Uttam Kumaran: Yeah, so you can create projects right here. All you need to do is put in the project name, and then it’s gonna prompt you. It’s gonna open this window, and you can just put the instructions right here.
445 00:58:25.150 ⇒ 00:58:26.050 Demilade Agboola: Gotcha.
446 00:58:30.000 ⇒ 00:58:32.689 Uttam Kumaran: And the other thing I will share is…
447 00:58:33.600 ⇒ 00:58:36.120 Uttam Kumaran: our prompt library here in Notion.
448 00:58:36.640 ⇒ 00:58:40.029 Uttam Kumaran: There’s a lot of great prompts here about…
449 00:58:40.340 ⇒ 00:58:44.920 Uttam Kumaran: like, kind of everything. Usually, I immediately take my prompt and put it into here, so…
450 00:58:45.440 ⇒ 00:58:52.060 Uttam Kumaran: like, try to use one here, and also, if you’re building some, contribute. To give you a good example of, like, where I….
451 00:58:52.060 ⇒ 00:58:54.930 Demilade Agboola: You’re not actually sharing your browser.
452 00:58:55.140 ⇒ 00:58:58.040 Uttam Kumaran: Oh, sorry. Out of here.
453 00:58:58.200 ⇒ 00:59:00.920 Uttam Kumaran: I was just sharing before this,
454 00:59:03.410 ⇒ 00:59:05.709 Uttam Kumaran: ChatGPT, so you can actually put
455 00:59:06.710 ⇒ 00:59:09.340 Uttam Kumaran: Here on the left side, in projects.
456 00:59:10.920 ⇒ 00:59:12.800 Uttam Kumaran: You can create new projects here.
457 00:59:14.390 ⇒ 00:59:19.379 Uttam Kumaran: And then… in Notion, we have this prompt library here.
458 00:59:21.270 ⇒ 00:59:22.450 Demilade Agboola: Oh, okay, nice.
459 00:59:22.450 ⇒ 00:59:26.939 Uttam Kumaran: So, all the prompts are here, like, for example, if you want, a sale, like.
460 00:59:27.330 ⇒ 00:59:32.109 Uttam Kumaran: a case study architect. We have a prompt here that helps to write case studies.
461 00:59:32.630 ⇒ 00:59:34.789 Uttam Kumaran: So it can take…
462 00:59:35.110 ⇒ 00:59:44.650 Uttam Kumaran: This is what Hannah actually uses to write case studies. So, where she takes an interview with me, or a long transcript of a meeting we’ve had, and we build the case studies from it.
463 00:59:45.050 ⇒ 00:59:50.789 Uttam Kumaran: But you can see, look, this is something we’ve iterated on multiple times, but now it’s… it gets it pretty accurate, you know?
464 00:59:52.210 ⇒ 00:59:58.810 Uttam Kumaran: So, a couple things that I could see us building is, one, like, we will start probably building prompts for each of the core
465 00:59:58.930 ⇒ 00:59:59.980 Uttam Kumaran: …
466 01:00:01.270 ⇒ 01:00:18.499 Uttam Kumaran: each of the core, technologies, so, like, maybe a Redshift agent or a Snowflake agent. Additionally, we can create, sort of, agents prompts for each of our core functions, like, maybe it is a analytics engineer, senior analytics engineer, senior data analyst, so you can use it as a sounding board.
467 01:00:18.520 ⇒ 01:00:23.109 Uttam Kumaran: That’s a great way to start, by the way, is just to use it as a sounding board for questions.
468 01:00:23.140 ⇒ 01:00:24.210 Uttam Kumaran: …
469 01:00:24.800 ⇒ 01:00:39.520 Uttam Kumaran: you know, as you’re walking through work, and then eventually you’ll start to see, like, okay, I’m using AI every day for this thing, maybe I could ask the AI team to take it one step further. Like, one step further being, like, we… in the platform, we created a UI now that you can
470 01:00:39.650 ⇒ 01:00:44.219 Uttam Kumaran: Create tickets, it comes up as squares, and you can quickly create tickets from it.
471 01:00:44.610 ⇒ 01:01:03.070 Uttam Kumaran: doing that in ChatGPT, you have to then copy-paste, go to linear, open a ticket, put it… paste it in, click good, go to the next one, like, it can be a little bit painful, but you… we have to do the legwork beforehand, you know? We have to know, like, okay, generally, I know that I can create tickets from transcript, and here’s how we’re gonna do that, so…
472 01:01:08.400 ⇒ 01:01:12.610 Uttam Kumaran: Any… Final thoughts or questions?
473 01:01:17.100 ⇒ 01:01:29.640 Uttam Kumaran: I would also love to hear any feedback on, like, this session. Like, how can we make this session a bit better? Like, did people enjoy the first part? Was seeing, like, how I use it?
474 01:01:30.130 ⇒ 01:01:33.659 Uttam Kumaran: more impactful? Should we do, like, a group?
475 01:01:34.000 ⇒ 01:01:41.459 Uttam Kumaran: like, workshop of, like, everybody using Cursor to push a PR or something, like, what… what would be helpful for next time?
476 01:01:49.390 ⇒ 01:01:55.569 Demilade Agboola: I’m not sure about what will be helpful for next time, but I know it’s helpful, like, seeing how people use it.
477 01:01:56.040 ⇒ 01:02:02.340 Demilade Agboola: Especially when there are bits and pieces that can be applicable to you, what you’re doing.
478 01:02:02.490 ⇒ 01:02:07.229 Demilade Agboola: For instance, the case study projects could be very useful for me.
479 01:02:07.680 ⇒ 01:02:13.900 Demilade Agboola: … Yeah, I think that’s… it’s a good start. I also…
480 01:02:14.860 ⇒ 01:02:22.190 Demilade Agboola: I think it might be helpful to see how people in the same, like, field use it, or in similar fields.
481 01:02:22.370 ⇒ 01:02:25.620 Demilade Agboola: Like, for instance, how Annie might use it.
482 01:02:25.810 ⇒ 01:02:34.280 Demilade Agboola: to troubleshoot a data issue or data quality issue might be helpful for me troubleshooting certain… like, like, things can… that can…
483 01:02:34.670 ⇒ 01:02:38.670 Demilade Agboola: Translate across sectors and spaces will be very helpful as well.
484 01:02:40.370 ⇒ 01:02:41.010 Uttam Kumaran: Okay.
485 01:02:48.950 ⇒ 01:02:56.959 Uttam Kumaran: Okay, great. Well, for me, I think I like seeing when people actually do stuff with AI, and it’s not just, like, talking about
486 01:02:57.360 ⇒ 01:03:06.250 Uttam Kumaran: sort of high-level stuff, so I want to prioritize for next time, one, continue to share, like, how the AI team is here to support everyone, but also
487 01:03:06.360 ⇒ 01:03:12.649 Uttam Kumaran: I want to also maybe just take examples where we can walk through together, like, how do we go from
488 01:03:12.840 ⇒ 01:03:19.800 Uttam Kumaran: ticket to cursor, to PR to, you know, try to look at different things in the engineering process.
489 01:03:20.170 ⇒ 01:03:32.159 Uttam Kumaran: But yeah, hopefully this is helpful and allows you to kind of think about your workflows and where AI can be helpful. It’s not a catch-all, and I’m not here to say that, like, you have to use it every single time.
490 01:03:32.160 ⇒ 01:03:40.929 Uttam Kumaran: But similar to the example that Andy gave, you know, we did a good job, I think, getting to the point where we can generate tickets with AI, the quality is not there.
491 01:03:40.940 ⇒ 01:03:46.549 Uttam Kumaran: But we really do need the feedback from this team on, like, what parts of the quality were wrong, and…
492 01:03:46.830 ⇒ 01:03:50.999 Uttam Kumaran: how we can make improvements to that. So, that’s really, really helpful to hear that.
493 01:03:55.320 ⇒ 01:03:55.930 Uttam Kumaran: Cool.
494 01:03:55.930 ⇒ 01:04:00.990 Samuel Roberts: Even hearing that is getting me thinking about how we can more easily streamline that feedback process.
495 01:04:01.630 ⇒ 01:04:04.429 Samuel Roberts: You know, so that it’s not a…
496 01:04:04.560 ⇒ 01:04:10.090 Samuel Roberts: We have to get on this, and someone has to give the feedback that it’s that, or like, you know, you have to go somewhere, but if there’s a way to tie that
497 01:04:10.580 ⇒ 01:04:13.300 Samuel Roberts: Choose what was generated. Yeah, exactly.
498 01:04:13.740 ⇒ 01:04:21.420 Uttam Kumaran: Yeah, like, you know, a nice thing could be, like, hey, if the AI doesn’t get it right, maybe you can tag it and say, like, you missed these things, and then it redoes it.
499 01:04:21.420 ⇒ 01:04:22.880 Samuel Roberts: Yeah, there’s something.
500 01:04:22.880 ⇒ 01:04:24.300 Uttam Kumaran: There’s an opportunity, yeah.
501 01:04:27.290 ⇒ 01:04:28.200 Uttam Kumaran: Okay.
502 01:04:28.450 ⇒ 01:04:41.799 Uttam Kumaran: Great. Well, thanks everyone for the time. I know this is… it was a… it’s an hour-long meeting, so I appreciate it. Please message me or the team if you have any questions, or if I can get you access to anything, like, if anyone is interested in… in that.
503 01:04:41.960 ⇒ 01:04:44.340 Uttam Kumaran: Text the… the speech-to-text thing.
504 01:04:44.860 ⇒ 01:04:47.329 Uttam Kumaran: Happy to, you know, get you invited.
505 01:04:49.940 ⇒ 01:04:50.870 Samuel Roberts: Cool.
506 01:04:51.190 ⇒ 01:04:52.670 Uttam Kumaran: Okay, thanks everyone.
507 01:04:52.670 ⇒ 01:04:53.710 Annie Yu: Thank you.
508 01:04:53.710 ⇒ 01:04:54.609 Demilade Agboola: Thank you.
509 01:04:54.610 ⇒ 01:04:55.060 Samuel Roberts: anytime.
510 01:04:55.060 ⇒ 01:04:55.470 Uttam Kumaran: Thank you.
511 01:04:55.470 ⇒ 01:04:56.290 Mustafa Raja: Thank you.