Meeting Title: AI Team Weekly Planning Date: 2025-02-10 Meeting participants: Uttam Kumaran, Miguel De Veyra, Casie Aviles


WEBVTT

1 00:02:19.550 00:02:20.910 Miguel de Veyra: You, you.

2 00:02:23.580 00:02:24.353 Casie Aviles: Cool little.

3 00:02:31.750 00:02:36.080 Miguel de Veyra: On q. 1. But until March 4 long for March 4th week.

4 00:02:39.900 00:02:40.750 Casie Aviles: Yeah.

5 00:02:46.640 00:02:47.929 Casie Aviles: yes, yes, we do.

6 00:02:51.790 00:03:00.019 Casie Aviles: I think Yung Ningying roadmap is good news.

7 00:03:00.840 00:03:02.850 Miguel de Veyra: From Calhatan, including ABC.

8 00:03:04.020 00:03:06.809 Casie Aviles: And no, no, at least for the Ak okrs loan.

9 00:03:07.506 00:03:09.509 Miguel de Veyra: No, I use as a new Pmo.

10 00:03:13.050 00:03:14.419 Casie Aviles: Dibala to like

11 00:03:17.510 00:03:19.959 Casie Aviles: like it on 3 initiative. Spa.

12 00:03:23.500 00:03:25.330 Miguel de Veyra: Our internal tooling agents.

13 00:03:30.910 00:03:32.950 Miguel de Veyra: Yeah, create roadmap, if at all.

14 00:03:33.050 00:03:40.080 Miguel de Veyra: I guess on somebody to us he was basically tracking to.

15 00:03:41.400 00:03:44.559 Miguel de Veyra: It’s also, I think this 2 goes hand in hand, though.

16 00:03:51.190 00:03:53.820 Casie Aviles: Yeah, Meron Perfor, and along top for Junior.

17 00:03:54.000 00:03:55.380 Miguel de Veyra: Offers in your own.

18 00:03:58.280 00:04:02.529 Miguel de Veyra: That’s a basic.

19 00:04:03.320 00:04:06.870 Casie Aviles: Yes, federal, personal. That’s personal snowflakes.

20 00:04:06.990 00:04:13.200 Miguel de Veyra: Hmm track team references cause in the way we track this is basically I should.

21 00:04:13.950 00:04:24.510 Miguel de Veyra: this is basically having, you know, collecting input and output input and AI outputs.

22 00:04:26.840 00:04:28.060 Miguel de Veyra: They will don’t work.

23 00:04:29.960 00:04:31.939 Casie Aviles: Yes, yes. Wait, wait, wait.

24 00:04:32.820 00:04:43.800 Casie Aviles: This is meeting notes. But yeah, meeting notes. So I I

25 00:04:47.420 00:04:48.814 Casie Aviles: yes, so we

26 00:04:50.250 00:04:52.900 Casie Aviles: I mean, it’s always in a manner.

27 00:04:53.420 00:04:55.070 Casie Aviles: Always negro ranion.

28 00:04:56.790 00:05:00.650 Miguel de Veyra: My my own. I’m hoping there was around knowledge base in.

29 00:05:03.327 00:05:08.690 Casie Aviles: Gamit! Ng, but my supervision.

30 00:05:09.560 00:05:09.910 Miguel de Veyra: So much.

31 00:05:09.910 00:05:10.700 Casie Aviles: High school.

32 00:05:10.900 00:05:15.479 Miguel de Veyra: So now collect na nahat.

33 00:05:15.700 00:05:21.600 Miguel de Veyra: Now for like. So bugger for

34 00:05:23.800 00:05:39.380 Miguel de Veyra: rug for meeting meeting notes, basically for meeting right Dubai and then inputs outputs known.

35 00:05:42.030 00:05:49.209 Miguel de Veyra: and then store store to the snowflakes to me. What’s up.

36 00:05:50.340 00:05:58.097 Casie Aviles: The unclear part is for the the it’s like if you if, how do I track? If someone reads the

37 00:05:58.640 00:05:59.630 Casie Aviles: summary.

38 00:06:01.010 00:06:03.089 Miguel de Veyra: What do you mean? Someone beats something

39 00:06:07.180 00:06:08.000 Miguel de Veyra: tall.

40 00:06:11.290 00:06:18.070 Casie Aviles: I mean not depends on the person, because sometimes we I don’t know. Maybe I’m complicating this. But

41 00:06:19.034 00:06:24.700 Casie Aviles: I was thinking, how do we know. Like, if someone actually reads the summaries being sent, or

42 00:06:25.090 00:06:29.260 Casie Aviles: me. Yeah, because we don’t slack. Doesn’t have, like, you know, a tracking.

43 00:06:30.600 00:06:38.549 Casie Aviles: The reports like. I mean reading read receipts right? Like you won’t know if someone read the message or not unlike in messenger.

44 00:06:42.560 00:06:44.769 Casie Aviles: Sorry do do you get? What do you get? What I.

45 00:06:44.770 00:06:49.910 Miguel de Veyra: Okay, so different than understanding understanding

46 00:06:50.160 00:07:05.669 Miguel de Veyra: meetings. Basically in a meeting notes.

47 00:07:06.020 00:07:06.690 Miguel de Veyra: Yes.

48 00:07:07.245 00:07:07.800 Casie Aviles: Hmm.

49 00:07:07.800 00:07:10.070 Miguel de Veyra: So. So that was my thought about it.

50 00:07:11.180 00:07:18.130 Miguel de Veyra: But if we’re here, I guess the the other thing we can do is

51 00:07:18.510 00:07:21.980 Miguel de Veyra: we can wait. Now. It’s just pull up.

52 00:07:23.090 00:07:31.450 Miguel de Veyra: Nothing going. Detail, I would say, is zoom react, click. Okay? Good.

53 00:07:31.930 00:07:34.590 Casie Aviles: Yeah, yeah, that I was thinking about that as well.

54 00:07:34.980 00:07:37.096 Casie Aviles: But no one’s really reacting.

55 00:07:37.520 00:07:41.740 Miguel de Veyra: Hey, you guys need to react to it.

56 00:07:44.940 00:07:46.970 Miguel de Veyra: If you guys actually attend

57 00:07:47.210 00:07:54.520 Miguel de Veyra: a long question, is that what they is that what this is, or do we need to develop a new agent

58 00:07:55.470 00:07:57.419 Miguel de Veyra: for meeting? Run? Do you get it?

59 00:07:57.420 00:07:58.230 Casie Aviles: For.

60 00:07:59.510 00:08:05.970 Miguel de Veyra: Operations team to review and do the work you can get to it.

61 00:08:06.020 00:08:08.329 Casie Aviles: This is for Ops, not ours.

62 00:08:08.850 00:08:18.390 Miguel de Veyra: Oh, like AI team to track. How often members refer to content in summarized notes.

63 00:08:22.330 00:08:23.149 Miguel de Veyra: doing well, please?

64 00:08:23.620 00:08:24.839 Miguel de Veyra: Or orange, too.

65 00:08:26.000 00:08:27.230 Miguel de Veyra: Yeah, you know.

66 00:08:27.230 00:08:28.949 Miguel de Veyra: Include that on Galington at all.

67 00:08:35.230 00:08:37.180 Miguel de Veyra: I do a long react tracking.

68 00:08:38.840 00:08:39.559 Casie Aviles: Yes.

69 00:08:41.070 00:08:45.090 Miguel de Veyra: Yeah. Well, actually, there’s no other way to do it. So it’s along your only way

70 00:08:47.710 00:08:54.399 Miguel de Veyra: attract you into some team references.

71 00:08:54.710 00:08:56.329 Casie Aviles: Yes, yes, it’s a.

72 00:08:56.330 00:08:57.669 Miguel de Veyra: Delete another, one.

73 00:08:57.940 00:09:00.228 Casie Aviles: To prevent micromanagement issues.

74 00:09:01.290 00:09:08.430 Miguel de Veyra: So tracking. Okay, so comment below now.

75 00:09:10.660 00:09:17.710 Miguel de Veyra: Oh, oh, man, so not in other ways. I’m meeting notes. We smoke

76 00:09:19.080 00:09:23.889 Miguel de Veyra: or so, general. So so the box Marion meeting notes for this, and then Saturday.

77 00:09:25.022 00:09:27.679 Casie Aviles: I mean I guess both. I don’t know.

78 00:09:29.090 00:09:37.410 Miguel de Veyra: Inside the tread inside the thread. 6.

79 00:09:42.870 00:09:44.779 Miguel de Veyra: And then, yeah, that should be it.

80 00:09:48.000 00:09:51.279 Miguel de Veyra: Measure the quality it it depends on and out in the room.

81 00:09:53.600 00:10:02.420 Casie Aviles: Measure the quality you want along response long AI pero dinga natin shake nagame.

82 00:10:02.860 00:10:09.530 Miguel de Veyra: AI to measure quality and quantity in the quality and quantity. So we are not on quantity. Right?

83 00:10:10.980 00:10:12.520 Miguel de Veyra: You know, Jerry, check not, then.

84 00:10:12.520 00:10:12.880 Casie Aviles: Right.

85 00:10:12.880 00:10:26.240 Miguel de Veyra: And then I guess the quality is just connecting the click on it, young golden reply, golden reply.

86 00:10:29.220 00:10:33.377 Casie Aviles: Yeah, yeah. But I guess the challenge there is.

87 00:10:35.060 00:10:41.029 Casie Aviles: since you know, everyone is be basically sending random messages there. So

88 00:10:41.480 00:10:46.830 Casie Aviles: in in the channel I mean, like in in our slack channel. If everyone’s sending random messages.

89 00:10:47.410 00:10:56.469 Casie Aviles: how do like does that each message the does? Sorry. Does each message need to have, like, you know, their own golden output. Get golden message.

90 00:10:56.850 00:11:15.360 Miguel de Veyra: So is it like the golden output only for our output, because basically, we’re collecting every message of either clients or us. And then we’re only sending one reply. Right? So we only need one golden reply, analyzer for the 4 Pm. Message. Right?

91 00:11:16.960 00:11:25.479 Casie Aviles: Hmm, yeah, I guess it could work more like, you know, we define like, what qualities do we have for? Do we want to have for each message.

92 00:11:26.490 00:11:33.360 Miguel de Veyra: And measure the partner that we’re sending to clients that we are sending to clients.

93 00:11:33.570 00:11:34.129 Casie Aviles: There you go!

94 00:11:35.460 00:11:37.064 Uttam Kumaran: Hey? Hey? Guys.

95 00:11:37.830 00:11:41.909 Miguel de Veyra: Wait. We’re sending to clients. Hmm!

96 00:11:42.800 00:11:44.130 Miguel de Veyra: Wait, I’m thinking.

97 00:11:47.330 00:11:55.810 Miguel de Veyra: because how do we even know if they’re from our client, from our message, is there, there’s a way, but in notion to do it in any 10

98 00:11:56.200 00:12:00.450 Miguel de Veyra: to check. If you know, this has, this user has a Brainforge email account.

99 00:12:01.400 00:12:07.490 Casie Aviles: Yeah, don’t worry about that part, since everyone there is from Rainforge. I mean the channel.

100 00:12:08.200 00:12:10.029 Miguel de Veyra: But we’re sending the clients right.

101 00:12:18.250 00:12:21.030 Uttam Kumaran: Yeah, that’s it.

102 00:12:21.030 00:12:22.330 Miguel de Veyra: Or I guess, do we here.

103 00:12:23.877 00:12:28.002 Uttam Kumaran: Wait. What? Sorry? Send it again.

104 00:12:28.920 00:12:31.330 Uttam Kumaran: So I’m not following Miguel. Sorry.

105 00:12:31.785 00:12:32.240 Casie Aviles: Awesome.

106 00:12:33.930 00:12:36.509 Casie Aviles: Yeah, we were talking about the

107 00:12:36.710 00:12:40.350 Casie Aviles: like, measuring the quality and quantity for the client.

108 00:12:40.350 00:12:40.760 Miguel de Veyra: Should show.

109 00:12:40.760 00:12:41.760 Casie Aviles: Client project.

110 00:12:41.830 00:12:43.310 Casie Aviles: Yeah, I can share.

111 00:12:43.530 00:12:46.659 Uttam Kumaran: Oh, okay, yeah. I have no idea what you’re talking about.

112 00:12:47.360 00:12:50.530 Miguel de Veyra: Your background noises like there’s so many people.

113 00:12:50.870 00:12:51.990 Uttam Kumaran: Oh, really.

114 00:12:51.990 00:12:52.760 Miguel de Veyra: Yeah.

115 00:12:54.560 00:12:57.549 Casie Aviles: Yeah, I just want to confirm, like everyone here.

116 00:12:57.550 00:13:00.619 Miguel de Veyra: I can see your your slack.

117 00:13:01.260 00:13:03.020 Casie Aviles: Yeah, we’re. I’m I’m sharing this luck.

118 00:13:03.330 00:13:04.120 Miguel de Veyra: Okay.

119 00:13:05.040 00:13:12.030 Casie Aviles: So yeah, like everyone, here’s from Brainforge, right? Like, except for I guess, Lam Lambrini, there’s no client here. Technically.

120 00:13:14.660 00:13:15.920 Uttam Kumaran: Oh, I’m doing value.

121 00:13:16.740 00:13:23.369 Uttam Kumaran: So like, wait like, do you guys, you gotta tell me like what we’re even talking about.

122 00:13:23.370 00:13:24.569 Miguel de Veyra: Show me Ito first.st

123 00:13:25.620 00:13:26.150 Uttam Kumaran: Yeah.

124 00:13:27.860 00:13:28.800 Casie Aviles: Okay. Sure.

125 00:13:30.425 00:13:30.909 Uttam Kumaran: Bye.

126 00:13:31.110 00:13:43.489 Miguel de Veyra: Because we were discussing like, okay, we were discussing this initiative measure, the quality of number and number of messages that we’re sending to clients, basically this 1st one.

127 00:13:44.010 00:13:50.880 Miguel de Veyra: And then basically what we’re, you know, planning around us

128 00:13:52.260 00:14:02.930 Miguel de Veyra: since we’re sending to clients, do we need to know? For we want to know who only has brain forge emails. Right? And then Casey is basically saying that everyone in the client

129 00:14:03.931 00:14:06.820 Miguel de Veyra: chat in that client channel is from brain forge.

130 00:14:07.830 00:14:11.610 Miguel de Veyra: So I’m not sure the extent to which, you know.

131 00:14:14.730 00:14:15.930 Miguel de Veyra: I don’t know which is rich.

132 00:14:16.290 00:14:23.250 Miguel de Veyra: So ideally, if the way I’m thinking about it is if we’re gonna create a slack with the event and from the people from ABC,

133 00:14:23.760 00:14:30.239 Miguel de Veyra: we’re not, you know, we’re we don’t really care what they send since their clients. So we we don’t, wanna, you know.

134 00:14:31.410 00:14:34.969 Miguel de Veyra: check for the quality of their messages. It’s more of ours already.

135 00:14:40.140 00:14:42.220 Uttam Kumaran: Let’s just talk about John.

136 00:14:42.540 00:14:44.790 Uttam Kumaran: Let’s leave ABC to ABC’s email.

137 00:14:46.080 00:14:49.259 Uttam Kumaran: Right? So let’s just talk about Javi.

138 00:14:49.440 00:14:55.460 Uttam Kumaran: can you walk? Can you guys walk me through this. What I’m looking at like sorry I keep.

139 00:14:55.590 00:15:00.180 Uttam Kumaran: I keep like trying to ask, but like I just don’t know what what am I looking at how

140 00:15:00.890 00:15:02.810 Uttam Kumaran: it’s the 1st time I’ve seen it so.

141 00:15:06.980 00:15:11.050 Miguel de Veyra: Yeah. So basically, Casey was the one who, you know.

142 00:15:11.720 00:15:15.750 Miguel de Veyra: separated this once. It’s basically the initiatives. The you know

143 00:15:15.850 00:15:17.700 Miguel de Veyra: what we want to do for q. 1.

144 00:15:18.220 00:15:18.630 Casie Aviles: Yeah.

145 00:15:18.630 00:15:20.450 Miguel de Veyra: And then basically, yeah.

146 00:15:20.890 00:15:22.178 Casie Aviles: Yeah, I was just

147 00:15:23.486 00:15:30.730 Casie Aviles: this to, yeah, to show, like, what are the key deliverables that I need to have? Yeah? And

148 00:15:30.900 00:15:36.060 Casie Aviles: like, what? What are the we need to have? So yeah.

149 00:15:38.240 00:15:40.300 Casie Aviles: I just wanted to organize like.

150 00:15:40.760 00:15:45.710 Casie Aviles: so I can like, come up with the timeline or the roadmap for the okrs.

151 00:15:47.550 00:15:48.200 Uttam Kumaran: Okay.

152 00:15:49.860 00:15:55.730 Uttam Kumaran: So the 1st thing we’re doing is just looking at slack snowflake.

153 00:15:57.137 00:15:58.450 Casie Aviles: I worked on.

154 00:16:01.060 00:16:06.910 Uttam Kumaran: So then, in terms of the Javi we’re just talking about like which channel we want to.

155 00:16:09.990 00:16:13.200 Uttam Kumaran: Like, I guess the question for me is like which channel we want to use.

156 00:16:14.960 00:16:23.479 Casie Aviles: Yeah, but we already clarified that last week. We already talked about it, and that would be this channel right? The client traffic Coffee Channel.

157 00:16:23.480 00:16:29.730 Uttam Kumaran: No, because there isn’t. No, we have. So we have external. We have external channels with the client. I just

158 00:16:29.940 00:16:34.120 Uttam Kumaran: if I I just can’t add you to all of them, because the client may ask like.

159 00:16:35.794 00:16:40.790 Uttam Kumaran: Who is this? Right? So that’s why my suggestion was, if can we

160 00:16:41.390 00:16:46.519 Uttam Kumaran: either you need to use my like credentials, or

161 00:16:46.670 00:16:52.670 Uttam Kumaran: if we create like a brain forge bought, then

162 00:16:53.090 00:16:56.389 Uttam Kumaran: like, I can add that to the channel which will give you access.

163 00:16:56.740 00:17:01.570 Uttam Kumaran: But I don’t wanna have like we can’t have the whole company in every channel. So that’s the thing we’re competing with.

164 00:17:02.260 00:17:02.970 Miguel de Veyra: Yeah.

165 00:17:06.275 00:17:06.849 Casie Aviles: Okay.

166 00:17:08.760 00:17:09.990 Uttam Kumaran: So what should we do?

167 00:17:10.250 00:17:24.169 Miguel de Veyra: So yeah. Kcc, there’s that. That’s where the my point comes in where we we only need to get the ones who has brain for him and brain forge email, basically because it’s with other with external clients. Now.

168 00:17:28.380 00:17:29.040 Casie Aviles: Okay.

169 00:17:30.650 00:17:33.859 Miguel de Veyra: And then I think, utham, what we can do there is

170 00:17:35.129 00:17:42.409 Miguel de Veyra: for us to get. Do we even need the bot for this Casey to build? Or do we just monitor the messages that are being sent.

171 00:17:42.670 00:17:46.260 Uttam Kumaran: But I guess, have we even gone to slack yet? Like do we have that anywhere.

172 00:17:48.860 00:17:49.540 Miguel de Veyra: Or in the world.

173 00:17:49.540 00:17:50.540 Casie Aviles: Update. I.

174 00:17:51.200 00:17:52.230 Uttam Kumaran: Slack data.

175 00:17:53.530 00:17:58.279 Casie Aviles: Yes, but I tracked the incorrect channel.

176 00:17:59.820 00:18:01.360 Uttam Kumaran: Oh, okay, okay, but okay. But.

177 00:18:01.360 00:18:01.880 Casie Aviles: But.

178 00:18:01.880 00:18:08.009 Uttam Kumaran: Also. Here, let me let me. Also. I’m gonna let me invite you right now to our. We have a snowflake so you can.

179 00:18:09.563 00:18:12.309 Uttam Kumaran: I’ll just invite you to the internal template.

180 00:18:23.090 00:18:25.810 Uttam Kumaran: And then how is? But how is this data getting stuff like.

181 00:18:29.590 00:18:30.540 Casie Aviles: Sorry. What’m.

182 00:18:31.260 00:18:33.589 Uttam Kumaran: How is the how is the data getting into Snowflake.

183 00:18:35.125 00:18:37.400 Casie Aviles: I just use a python script. So

184 00:18:38.200 00:18:40.350 Casie Aviles: yeah, so what it looks like is.

185 00:18:44.780 00:18:52.079 Casie Aviles: yeah. But I just use a python script and a snowflake connector, which is just a library that I I imported.

186 00:18:59.010 00:19:02.090 Uttam Kumaran: Is, and is. There? Is there data in those tables.

187 00:19:05.048 00:19:06.880 Casie Aviles: Yeah, there should be.

188 00:19:07.300 00:19:08.060 Casie Aviles: It should.

189 00:19:12.490 00:19:14.714 Casie Aviles: Yeah, it looks like this.

190 00:19:16.250 00:19:21.699 Casie Aviles: there are some that are blank, I think, because of, you know there are screenshots like images, and

191 00:19:21.910 00:19:25.450 Casie Aviles: so it’s not here, but for text messages. It should be here.

192 00:19:26.470 00:19:27.030 Uttam Kumaran: Okay.

193 00:19:31.710 00:19:33.100 Uttam Kumaran: okay, okay, great.

194 00:19:34.490 00:19:39.550 Uttam Kumaran: So let me I’m gonna invite you to our internal snowflake that way.

195 00:19:39.930 00:19:41.960 Uttam Kumaran: You don’t have to use this trial account.

196 00:19:43.118 00:19:46.730 Uttam Kumaran: And if you can point it here, that would be great.

197 00:19:50.090 00:19:53.949 Uttam Kumaran: But then, yeah, so how do we want to handle for other channels like

198 00:19:54.520 00:19:56.130 Uttam Kumaran: you want to use my account?

199 00:19:58.780 00:20:00.240 Uttam Kumaran: Do we want to use an app.

200 00:20:02.030 00:20:09.960 Casie Aviles: Yeah, sure. Because one thing I’m thinking of is through Zapier. But then hmm.

201 00:20:13.800 00:20:16.191 Uttam Kumaran: So yeah, I wanted to try.

202 00:20:18.030 00:20:19.669 Uttam Kumaran: do you want to try using?

203 00:20:20.430 00:20:25.210 Uttam Kumaran: Like, I I’m gonna I may not have time. I was. I was hoping that I could try this. But

204 00:20:25.600 00:20:27.890 Uttam Kumaran: you wanna try this software called Dlp.

205 00:20:28.160 00:20:30.919 Uttam Kumaran: the open source software for data movement.

206 00:20:31.940 00:20:37.089 Uttam Kumaran: And they do have a slack connector, and it’s all in Python.

207 00:20:39.000 00:20:40.489 Casie Aviles: You mean a credential.

208 00:20:42.800 00:20:44.829 Uttam Kumaran: No, no! I just sent it in the zoom chat.

209 00:20:45.630 00:20:48.179 Casie Aviles: Oh, okay, yeah. I see it dlt hub.

210 00:21:10.840 00:21:12.930 Uttam Kumaran: You wanna try that sad.

211 00:21:16.080 00:21:23.590 Casie Aviles: Okay, okay, how? How would how would this be different, like, compared to like

212 00:21:29.210 00:21:33.089 Miguel de Veyra: I think this is just the trigger note. And this is basically the placing Zapier.

213 00:21:34.210 00:21:38.220 Uttam Kumaran: No, no, this is actual. Api. This is replacing. This is replacing everything.

214 00:21:38.750 00:21:43.691 Uttam Kumaran: So you won’t have to. You won’t have to use that here anymore. You can just

215 00:21:44.590 00:21:52.600 Uttam Kumaran: you can just run you should be able to run that to actually grab

216 00:21:52.780 00:21:56.149 Uttam Kumaran: black messages and put them into something.

217 00:22:06.270 00:22:07.330 Casie Aviles: Okay. Okay.

218 00:22:24.440 00:22:28.339 Miguel de Veyra: Load data from slack to know if they can python with the office.

219 00:22:40.730 00:22:42.320 Uttam Kumaran: Yeah. So basically,

220 00:22:43.510 00:22:47.929 Uttam Kumaran: I mean, it should. That should dlt hub should should allow you to move the data from slack

221 00:22:48.180 00:22:49.490 Uttam Kumaran: to snowflake.

222 00:22:49.650 00:22:51.020 Uttam Kumaran: I’m gonna just.

223 00:22:51.850 00:22:54.140 Uttam Kumaran: I’m gonna make sure you’re in the right snowflake right now.

224 00:22:54.777 00:22:57.050 Uttam Kumaran: Let me just let me just go and do that.

225 00:23:04.190 00:23:06.929 Uttam Kumaran: Miguel. I guess I’ll invite you as well to Snowflake.

226 00:23:07.230 00:23:07.863 Miguel de Veyra: Okay. Sure.

227 00:23:10.420 00:23:11.480 Uttam Kumaran: Wow!

228 00:23:19.780 00:23:24.760 Miguel de Veyra: So the flow is. Now gonna be. I’ll just move this a bit.

229 00:23:26.460 00:23:27.210 Casie Aviles: Sure, sure.

230 00:23:28.580 00:23:31.540 Miguel de Veyra: So now the process is phone.

231 00:23:31.840 00:23:33.829 Miguel de Veyra: So it goes to

232 00:23:37.100 00:23:43.469 Miguel de Veyra: deal from slack to snowflake, using the Ot.

233 00:23:49.610 00:23:50.290 Uttam Kumaran: No.

234 00:23:55.790 00:23:58.310 Miguel de Veyra: There you go. So dlt here.

235 00:24:00.010 00:24:01.200 Miguel de Veyra: Okay, there you go.

236 00:24:03.080 00:24:07.179 Uttam Kumaran: Ideally over time. We’re going to use Dlt. Even on the data side for data movement.

237 00:24:07.340 00:24:12.700 Uttam Kumaran: They’re like a new open source. I talked to the co-founder good company.

238 00:24:13.230 00:24:18.869 Uttam Kumaran: So I wanna try to use them because currently they’re they’re they’re basically free if we host it.

239 00:24:20.400 00:24:22.613 Uttam Kumaran: So I want to test it out.

240 00:24:22.930 00:24:23.530 Miguel de Veyra: Okay.

241 00:24:25.370 00:24:26.300 Uttam Kumaran: Oh!

242 00:24:29.220 00:24:31.540 Miguel de Veyra: Casey, do you want me to share screen? I’ll share screen.

243 00:24:32.050 00:24:32.900 Casie Aviles: Yeah, sure.

244 00:24:34.990 00:24:41.419 Miguel de Veyra: So now it it looks like, let me just hide this. Because let me minimize this. Okay.

245 00:24:47.440 00:24:48.710 Miguel de Veyra: okay, there you go.

246 00:24:50.090 00:24:54.649 Miguel de Veyra: And then so slack, using now dlt instead of zapier, it goes to slack.

247 00:24:57.290 00:24:59.402 Uttam Kumaran: Okay. And then I’m gonna send you both

248 00:24:59.930 00:25:01.750 Miguel de Veyra: How do we process the data now?

249 00:25:04.720 00:25:07.619 Miguel de Veyra: Or do we just add the somewhere.

250 00:25:07.620 00:25:08.170 Uttam Kumaran: Hey?

251 00:25:08.530 00:25:10.593 Uttam Kumaran: Wait, wait! Hold on one second. Let me just finish.

252 00:25:11.880 00:25:14.759 Uttam Kumaran: let me just send you guys the details here.

253 00:25:19.900 00:25:26.860 Uttam Kumaran: Okay, okay, so we have slack, we have dlt.

254 00:25:27.280 00:25:36.480 Uttam Kumaran: The AI process is what AI process should come after AI process should come after snowflake.

255 00:25:39.430 00:25:42.090 Miguel de Veyra: Do we not want to? So also store

256 00:25:44.240 00:25:46.339 Miguel de Veyra: the like if it’s good or bad.

257 00:25:47.010 00:25:51.210 Uttam Kumaran: Yeah, but you can. But then it just loops within Snowflake that all has to happen within, like

258 00:25:51.940 00:25:58.370 Uttam Kumaran: you call. So the basically you want to drop the raw data, and then you want to process it and then rewrite it as needed.

259 00:25:58.910 00:26:00.010 Uttam Kumaran: You see what I mean.

260 00:26:01.054 00:26:02.430 Miguel de Veyra: Okay. So it’s after.

261 00:26:04.220 00:26:09.000 Miguel de Veyra: And then, okay, so this one, I guess we’ll just

262 00:26:09.390 00:26:12.039 Miguel de Veyra: use like an Llm prompt or something right?

263 00:26:15.240 00:26:17.930 Uttam Kumaran: Yes, let me let me open the board as well.

264 00:26:23.300 00:26:24.660 Miguel de Veyra: Use the front.

265 00:26:25.790 00:26:26.719 Miguel de Veyra: There we go.

266 00:26:50.400 00:26:52.250 Uttam Kumaran: Wait! What board is this in.

267 00:26:53.040 00:26:54.470 Miguel de Veyra: Junior, Pm.

268 00:26:59.270 00:26:59.950 Uttam Kumaran: Nice.

269 00:27:03.490 00:27:04.220 Uttam Kumaran: Alright.

270 00:27:09.210 00:27:17.040 Uttam Kumaran: yes, so update snowflake and then then, we’re gonna have.

271 00:27:38.770 00:27:43.210 Uttam Kumaran: So so then but I just wanna have it like

272 00:27:43.740 00:27:47.729 Uttam Kumaran: to. I don’t know how you’re using your own. This is insane.

273 00:27:47.730 00:27:51.759 Miguel de Veyra: You can. Just you can just drag it off. There’s an arrow every time.

274 00:27:51.760 00:27:56.909 Uttam Kumaran: Sorry. I’m a hater. I like fig jam. Okay, okay, but update snowflake record. So

275 00:27:57.940 00:28:04.410 Uttam Kumaran: I wanna just note down like what we mean by like, update, right? So we want to look at

276 00:28:06.430 00:28:08.420 Miguel de Veyra: Go compared to the golden reply. Right.

277 00:28:09.070 00:28:12.970 Uttam Kumaran: Well, we almost wanna we almost want like on

278 00:28:13.260 00:28:19.016 Uttam Kumaran: a daily basis. I wanna know one, did we? Message client?

279 00:28:20.350 00:28:29.119 Uttam Kumaran: And 2 were our messages quality overall for the day.

280 00:28:29.350 00:28:31.200 Uttam Kumaran: Right? That’s what I want to know.

281 00:28:31.950 00:28:34.380 Miguel de Veyra: Okay, okay, this is much clearer or not.

282 00:28:34.380 00:28:37.250 Uttam Kumaran: Update snowflake record. You, almost, it’s like

283 00:28:37.760 00:28:46.070 Uttam Kumaran: you, wanna, we we wanna update like the like, it’s like the

284 00:28:48.010 00:28:56.220 Uttam Kumaran: result. It’s like, the yeah. I don’t know. Like client communication, summary table.

285 00:28:56.220 00:28:57.500 Casie Aviles: Have another column.

286 00:28:59.270 00:29:03.639 Uttam Kumaran: Well, yeah, we’ll create a new table. That’s basically like, client like day client.

287 00:29:08.070 00:29:13.159 Uttam Kumaran: Where should I? Where can I put like, where can I put that one of it?

288 00:29:13.480 00:29:22.710 Uttam Kumaran: Day client number of messages from clients.

289 00:29:23.680 00:29:30.240 Uttam Kumaran: Number of messages from? Yes. Okay. Actually, how do I like that

290 00:29:36.210 00:29:39.000 Uttam Kumaran: number of buses from Rainforge.

291 00:29:40.160 00:29:44.920 Uttam Kumaran: Then we’re gonna then basically, we can create a boolean. That’s like, did we send a message right.

292 00:29:45.720 00:29:49.399 Miguel de Veyra: It’s right, and then but the second one is like.

293 00:29:50.020 00:29:51.710 Uttam Kumaran: Quality, score.

294 00:29:53.510 00:29:54.250 Miguel de Veyra: Okay.

295 00:30:06.240 00:30:07.520 Uttam Kumaran: Something like that.

296 00:30:08.040 00:30:09.471 Uttam Kumaran: But ideally, if you

297 00:30:10.470 00:30:16.559 Uttam Kumaran: if you if you can work on case, if you can work on these, then I can help with this part.

298 00:30:16.820 00:30:21.660 Uttam Kumaran: Sorry with, yeah, with this part.

299 00:30:22.830 00:30:24.030 Uttam Kumaran: But data stuff.

300 00:30:24.900 00:30:29.379 Uttam Kumaran: And then the zapier trigger is basically like we should get a message in slack that’s like.

301 00:30:29.670 00:30:32.410 Uttam Kumaran: Didn’t send a message to that, you know.

302 00:30:32.410 00:30:38.909 Miguel de Veyra: Yeah, wait. Sorry, guys, isn’t this what we discussed last week? The 4 Pm. Message that we check.

303 00:30:38.910 00:30:39.530 Uttam Kumaran: Yeah.

304 00:30:42.090 00:30:43.829 Uttam Kumaran: But like right now, we don’t.

305 00:30:48.520 00:30:51.569 Uttam Kumaran: Right now we’re not. We’re not. We’re not checking right? So

306 00:30:52.010 00:30:54.369 Uttam Kumaran: ideally, I want to get an alert. That’s like

307 00:30:59.470 00:31:00.960 Uttam Kumaran: we didn’t send a message.

308 00:31:04.440 00:31:06.950 Miguel de Veyra: Oh, okay, okay. Wait a little bit confused.

309 00:31:07.860 00:31:09.589 Uttam Kumaran: Okay. What do you confuse? Tell me.

310 00:31:09.590 00:31:11.740 Miguel de Veyra: Cause. I think we already have that

311 00:31:13.390 00:31:16.959 Miguel de Veyra: right, Casey. We already have where it checks all the message.

312 00:31:17.320 00:31:21.810 Casie Aviles: Yeah, but it’s on. Yeah, it’s on zappear. But it’s on the wrong channel. And

313 00:31:23.721 00:31:27.590 Casie Aviles: yeah, it just sends the messages on spread on the spreadsheet.

314 00:31:28.020 00:31:28.580 Miguel de Veyra: Yeah, yeah.

315 00:31:28.580 00:31:32.429 Casie Aviles: And also to vellum for quality, for the quality, assessment.

316 00:31:33.300 00:31:34.360 Miguel de Veyra: Okay, yeah.

317 00:31:36.120 00:31:39.390 Miguel de Veyra: So basically, what we’ll add is

318 00:31:39.790 00:31:45.070 Miguel de Veyra: use dlt send it to snowflake instead of sheets, and then

319 00:31:45.830 00:31:54.069 Miguel de Veyra: have basically a daily quality check. We don’t really need each message to check right to them. We need like a. So we need to check the summary of the day.

320 00:31:55.330 00:31:57.019 Uttam Kumaran: Yeah, I would start by the day.

321 00:31:57.190 00:31:58.979 Miguel de Veyra: Okay, yeah, okay, that should.

322 00:31:58.980 00:32:05.509 Uttam Kumaran: So again, that’s that’s like the most crucial part is like, did we send anything at all? The second piece, though I want to look at quality.

323 00:32:07.060 00:32:09.970 Uttam Kumaran: and then, if this is working, then we’ll scale this to all.

324 00:32:10.300 00:32:12.890 Uttam Kumaran: We’ll scale this to all of our client channels.

325 00:32:13.300 00:32:14.120 Miguel de Veyra: Yep, okay.

326 00:32:14.760 00:32:20.839 Miguel de Veyra: yeah. I think that’s much clearer now. So it’s daily wait. Let me just add another note here.

327 00:32:23.390 00:32:26.154 Uttam Kumaran: So where are you? Gonna run dlt Casey.

328 00:32:28.884 00:32:32.359 Casie Aviles: Yeah. Windmill is what I was thinking of.

329 00:32:35.080 00:32:45.149 Casie Aviles: because the python script that I prepared was just local. And yeah, it doesn’t use dlt. It’s using just the connectors, the packages.

330 00:32:46.230 00:32:46.780 Uttam Kumaran: Okay?

331 00:32:51.570 00:32:58.800 Uttam Kumaran: So yeah, I would suggest, yeah, it’s it’s like, we have a, we have a messages table.

332 00:32:58.950 00:33:01.929 Uttam Kumaran: Yeah, ideally, we just have a, we basically just have a messages table.

333 00:33:03.960 00:33:04.750 Uttam Kumaran: Oh.

334 00:33:06.380 00:33:07.240 Miguel de Veyra: There you go.

335 00:33:13.640 00:33:17.370 Miguel de Veyra: Okay, so these are like the 2 main deliverables.

336 00:33:26.140 00:33:27.690 Miguel de Veyra: Is this correct, Lisa?

337 00:33:30.210 00:33:33.039 Miguel de Veyra: No, not to be deliverable? Sorry, but.

338 00:33:33.040 00:33:34.979 Uttam Kumaran: Yeah, yeah, but basically, it’s like.

339 00:33:35.400 00:33:37.220 Miguel de Veyra: That’s basically what we want to do.

340 00:33:46.360 00:33:47.070 Uttam Kumaran: Awesome.

341 00:33:50.380 00:33:54.637 Uttam Kumaran: and then this is like a daily summary.

342 00:33:56.180 00:34:00.300 Uttam Kumaran: Yep, and then this is up here. Yep, yes.

343 00:34:00.300 00:34:07.030 Miguel de Veyra: So we’ll just move this here and then, okay, everything clear.

344 00:34:07.030 00:34:09.589 Uttam Kumaran: Now, now I’m following. Now I’m following.

345 00:34:10.705 00:34:18.009 Uttam Kumaran: Although dude some of this we should do in notion. I’ll let you. I’ll let you. I’ll let you.

346 00:34:19.639 00:34:22.869 Miguel de Veyra: Okay, wait. So the tasks for this one

347 00:34:25.989 00:34:30.699 Miguel de Veyra: post plp on, where are you, gonna host? This Casey.

348 00:34:30.699 00:34:31.659 Uttam Kumaran: On one, now.

349 00:34:31.659 00:34:32.249 Casie Aviles: We don’t know.

350 00:34:32.250 00:34:36.789 Miguel de Veyra: Windmill. Yeah, yeah, is this how you spell windmill?

351 00:34:36.909 00:34:39.260 Miguel de Veyra: Yeah, yeah, okay, okay.

352 00:34:40.540 00:34:48.679 Miguel de Veyra: And then create 2 data, create 2 Dbs create 2 Dbs.

353 00:34:52.760 00:34:54.960 Miguel de Veyra: and then the to do this.

354 00:34:55.368 00:34:59.449 Casie Aviles: Also, Tom. I sent like a screenshot from Zapier on.

355 00:34:59.870 00:35:07.359 Casie Aviles: and the chat I just want to ask which channel again, like, I don’t need to be added there, I just want to know which one. So it.

356 00:35:07.360 00:35:08.717 Casie Aviles: okay, yeah. It’s

357 00:35:09.710 00:35:10.340 Uttam Kumaran: It’s

358 00:35:20.310 00:35:26.480 Miguel de Veyra: And then I think another dependency here is for the quality.

359 00:35:29.680 00:35:35.310 Miguel de Veyra: What’s the measure in which we, you know? Basically, how do we know if it’s quality or not with them.

360 00:35:38.041 00:35:41.579 Uttam Kumaran: Like, Casey. Where do you want me to put the slack channel in this? Is it.

361 00:35:41.970 00:35:46.910 Casie Aviles: Oh, you could just send it on chat, or wherever.

362 00:35:47.220 00:35:48.499 Miguel de Veyra: It’s on, it’s on mute.

363 00:35:48.500 00:35:49.350 Casie Aviles: Miro, yeah.

364 00:35:49.350 00:35:50.550 Miguel de Veyra: Yeah, yeah, but yeah.

365 00:35:50.550 00:35:51.469 Casie Aviles: Okay. I see.

366 00:35:51.812 00:35:54.207 Uttam Kumaran: Quality. I don’t know. I would just

367 00:35:55.050 00:35:56.240 Miguel de Veyra: General, for now.

368 00:35:56.600 00:35:59.740 Uttam Kumaran: I, just yeah, this doesn’t.

369 00:35:59.840 00:36:07.463 Uttam Kumaran: Yeah, this is all new for me. So yeah, something basic. And then.

370 00:36:07.900 00:36:08.315 Miguel de Veyra: Yeah.

371 00:36:14.620 00:36:15.620 Uttam Kumaran: Oh, wow!

372 00:36:16.350 00:36:21.760 Miguel de Veyra: So the 1st is you need to host this on. Dlt

373 00:36:22.170 00:36:27.130 Miguel de Veyra: on. When mail you need to create 2 databases, and then

374 00:36:28.260 00:36:30.879 Miguel de Veyra: all messages to play with clients to do.

375 00:36:32.970 00:36:37.359 Miguel de Veyra: And then another task here is

376 00:36:43.980 00:36:48.640 Miguel de Veyra: analyze. There is somebody skipping a step.

377 00:37:02.440 00:37:04.749 Miguel de Veyra: the stable to somebody.

378 00:37:08.970 00:37:10.809 Miguel de Veyra: daily summary table.

379 00:37:18.100 00:37:19.950 Miguel de Veyra: And then here.

380 00:37:20.990 00:37:25.200 Miguel de Veyra: It’s basically analyze

381 00:37:38.190 00:37:38.990 Miguel de Veyra: so many

382 00:37:45.450 00:37:46.260 Miguel de Veyra: name is

383 00:37:52.540 00:37:59.139 Miguel de Veyra: okay, is he just sorry? Is this pretty clear.

384 00:37:59.550 00:38:02.210 Casie Aviles: Yeah, yeah, it’s it’s much clearer now.

385 00:38:02.912 00:38:04.089 Miguel de Veyra: Okay, nice. Thanks.

386 00:38:09.810 00:38:12.989 Miguel de Veyra: Wait a call recorded. So Dvs.

387 00:38:13.660 00:38:16.389 Miguel de Veyra: let’s color this blue because snowflake is blue.

388 00:38:19.000 00:38:22.390 Miguel de Veyra: What’s the color of windmill purple, right?

389 00:38:23.450 00:38:24.550 Casie Aviles: Also blue.

390 00:38:24.720 00:38:25.830 Uttam Kumaran: They’re blue.

391 00:38:26.210 00:38:31.879 Miguel de Veyra: Okay, lighter blue, then create summary. AI is always purple.

392 00:38:35.060 00:38:39.010 Miguel de Veyra: No, there you go. Purple 4, 2, 4, 2.

393 00:38:44.800 00:38:47.960 Miguel de Veyra: Okay, cool.

394 00:38:48.680 00:38:51.510 Miguel de Veyra: I’ll create the tasks after this call, guys.

395 00:38:58.360 00:38:59.310 Uttam Kumaran: Oh!

396 00:39:01.820 00:39:03.000 Uttam Kumaran: Oh!

397 00:39:06.770 00:39:09.879 Miguel de Veyra: Is anyone else still doing something? Or can we move on.

398 00:39:10.160 00:39:11.330 Uttam Kumaran: Now we’re good. Yeah.

399 00:39:11.330 00:39:11.890 Miguel de Veyra: Okay?

400 00:39:12.050 00:39:20.059 Miguel de Veyra: And then, okay, moving on to this initiative, the measure usage, for this.

401 00:39:20.060 00:39:21.070 Uttam Kumaran: Yes.

402 00:39:21.070 00:39:24.410 Miguel de Veyra: Utam. I think the only way we can do this is through react.

403 00:39:24.850 00:39:29.129 Miguel de Veyra: So if they react something into the summary, then, you know, we assume they read it.

404 00:39:30.630 00:39:38.540 Uttam Kumaran: But also like, oh, okay, let’s let’s list out all the interaction opportunities.

405 00:39:38.890 00:39:40.380 Uttam Kumaran: Right?

406 00:39:41.870 00:39:42.560 Casie Aviles: Red reply.

407 00:39:42.560 00:39:46.270 Uttam Kumaran: The act tracking direct apps.

408 00:39:46.450 00:39:47.230 Uttam Kumaran: Right?

409 00:39:51.948 00:39:56.840 Uttam Kumaran: Replies and thread with agents.

410 00:40:04.130 00:40:11.600 Uttam Kumaran: amount of feedback. So this is also shit. We should just dump into Snowflake.

411 00:40:11.900 00:40:18.619 Uttam Kumaran: But yeah, like this one is a bit harder, because I don’t know. I think, like

412 00:40:19.230 00:40:20.440 Uttam Kumaran: I don’t know where. Like

413 00:40:20.817 00:40:26.020 Uttam Kumaran: we can get this from slack. We can also get this from bellum, because I also want to look at like.

414 00:40:27.070 00:40:29.879 Uttam Kumaran: yeah, I don’t know whether it’s better to get this from Mellum.

415 00:40:31.860 00:40:35.469 Casie Aviles: What? Exactly do we want to get from Valem like the logs and the Evals.

416 00:40:35.470 00:40:40.846 Uttam Kumaran: Yeah, well, no. Like, just like how how the interactions with the agent.

417 00:40:42.410 00:40:54.160 Uttam Kumaran: Okay, like, when people are when people are hitting the agent for tasks like, how often that’s happening.

418 00:40:54.850 00:40:56.490 Casie Aviles: Oh, okay. Okay.

419 00:41:06.760 00:41:08.400 Uttam Kumaran: You want to get that from Bella.

420 00:41:12.346 00:41:17.369 Casie Aviles: I do know that there’s like this executions table for vellum.

421 00:41:18.830 00:41:24.119 Casie Aviles: So oh, yeah, we could export it. This executions.

422 00:41:25.140 00:41:29.840 Uttam Kumaran: But so vellum has an Api right like, can we? We can use Dlt to, maybe grab that too.

423 00:41:31.800 00:41:32.740 Uttam Kumaran: Okay, yeah.

424 00:41:32.740 00:41:33.410 Casie Aviles: I don’t.

425 00:41:35.250 00:41:43.130 Uttam Kumaran: Like, I want to get all the vellum logs into snowflake too, all right.

426 00:41:47.390 00:41:50.729 Miguel de Veyra: Is the is our zoom summarizer already in value.

427 00:41:51.780 00:41:58.604 Casie Aviles: No, actually, not. There’s none of the agents that we have right now are on vellum except for just this.

428 00:41:59.360 00:42:03.109 Casie Aviles: yeah, for the 1st initiative that we were talking about earlier.

429 00:42:04.121 00:42:08.110 Miguel de Veyra: So yeah. So we have to take a step back, because if it’s not even there yet, right.

430 00:42:09.300 00:42:14.589 Casie Aviles: Yeah, that’s the thing. So I’ll have to recreate like each agent on vellum

431 00:42:16.220 00:42:20.919 Casie Aviles: if we want to like. Yeah, track there from bellum.

432 00:42:26.731 00:42:28.218 Uttam Kumaran: I mean, yeah, I

433 00:42:28.760 00:42:34.997 Uttam Kumaran: But, Miguel, I’m waiting on you when I make a decision on like or not.

434 00:42:35.580 00:42:39.260 Miguel de Veyra: Yeah, cause if we could do that and then get the logs.

435 00:42:39.460 00:42:44.769 Miguel de Veyra: But the thing is the cause. We have to build everything in vellum

436 00:42:45.310 00:42:51.510 Miguel de Veyra: right? And then I, we spoke actually with John Casey and I earlier today.

437 00:42:51.780 00:42:56.560 Miguel de Veyra: basically just a simple rag connection to super base and everything.

438 00:42:58.160 00:43:00.709 Miguel de Veyra: And it’s not even production ready yet, like

439 00:43:00.930 00:43:03.739 Miguel de Veyra: she’s been like going above her time.

440 00:43:04.700 00:43:10.360 Miguel de Veyra: And then it’s like a very simple, you know, agent

441 00:43:11.660 00:43:15.729 Miguel de Veyra: like, it’s basically conversational with a bit of rag.

442 00:43:16.070 00:43:21.280 Miguel de Veyra: But, for example, if our lead agent was just like a couple of steps

443 00:43:23.280 00:43:32.609 Miguel de Veyra: like a lot of steps, I think that would take a lot of coding, and then I’m not sure. Then I’m not sure like, is it really worth it to do all that for logs?

444 00:43:34.210 00:43:39.489 Miguel de Veyra: Cause what we can do is just put like a node in the input and the output, and just

445 00:43:39.790 00:43:45.440 Miguel de Veyra: do the golden reply there, send it to to the to Snowflake, or whatever

446 00:43:45.770 00:43:49.969 Miguel de Veyra: or do we really want to code everything, because that takes a lot more time. Of course.

447 00:43:54.270 00:43:59.870 Uttam Kumaran: I don’t know, dude I I we’re we’re coming back to the same conversation that we had last month, like

448 00:44:00.500 00:44:06.550 Uttam Kumaran: I for our AI clients. I want us to have evals, and I want us to have logs.

449 00:44:08.260 00:44:13.720 Miguel de Veyra: Yeah. But building our own logs is better than depending on vellum. Anyways.

450 00:44:15.470 00:44:18.289 Uttam Kumaran: Why I don’t want to build logs

451 00:44:21.330 00:44:26.250 Uttam Kumaran: like I don’t wanna do agent tracing. That’s why we tried to look at trace, loop and stuff like that right.

452 00:44:57.400 00:45:00.279 Uttam Kumaran: I don’t know. You tell me like I just like we I mean.

453 00:45:00.910 00:45:04.320 Uttam Kumaran: we just paid for it. So I mean, that’s but that’s fine like.

454 00:45:04.980 00:45:08.479 Uttam Kumaran: I just like, I just wanna know how we’re gonna move forward.

455 00:45:09.140 00:45:14.019 Miguel de Veyra: Yeah, cause I think, like my biggest concern, this concern. There is like the dev time

456 00:45:16.437 00:45:20.600 Miguel de Veyra: Casey, can you give us like an estimate how long it would take you to

457 00:45:20.740 00:45:25.960 Miguel de Veyra: basically migrate the entire lead research agent, because I think that would be the standard of.

458 00:45:26.200 00:45:32.180 Uttam Kumaran: But I guess, like this is where I don’t want to talk. I don’t wanna like I’m not concerned with, like, how long it’s gonna take.

459 00:45:32.990 00:45:40.110 Uttam Kumaran: I guess from you I’m looking for like, do we want to use vellum or not? Right? I this is where I’m like, I thought we already.

460 00:45:41.380 00:45:47.000 Uttam Kumaran: I thought we were all on board, and it’s clear that we still have some doubts.

461 00:45:49.630 00:45:53.739 Uttam Kumaran: So like, I just wanna know, like, if we’re not using vellum.

462 00:45:54.030 00:45:57.020 Uttam Kumaran: My requirement is still, we need to do evals.

463 00:45:57.840 00:46:04.010 Uttam Kumaran: So what up like, can we? Can we? So I think you just need to. I need you to make some decisions on that

464 00:46:04.170 00:46:11.190 Uttam Kumaran: like, call the vellum guys, or whatever, or just say, Look.

465 00:46:11.750 00:46:14.179 Uttam Kumaran: we don’t want to use it anymore. I’ll cancel it.

466 00:46:17.040 00:46:18.000 Uttam Kumaran: That’s fine.

467 00:46:24.140 00:46:28.170 Uttam Kumaran: I mean. Then then I’ll just go with the with the enterprise, and again.

468 00:46:28.320 00:46:33.150 Miguel de Veyra: Yeah. Yeah, I think, like the biggest thing here is.

469 00:46:33.720 00:46:38.899 Miguel de Veyra: can we cause? I cause cause we discuss. I discuss this with Casey almost daily.

470 00:46:39.860 00:46:42.610 Miguel de Veyra: What type of evils do we actually need

471 00:46:45.150 00:46:55.830 Miguel de Veyra: like? Is it like, you know? Do we just need to save all the inputs outputs and then grade it the quality, or do we do we? Do we really need to know? Like, hey? This request took this amount of tokens.

472 00:46:56.990 00:46:58.469 Uttam Kumaran: Yeah, I want that.

473 00:47:00.550 00:47:05.609 Miguel de Veyra: I guess, for that cause. There’s we tried doing that in in a 10 right? Casey.

474 00:47:05.630 00:47:10.786 Casie Aviles: Yeah, yeah, I sent, like, just an example. I was just playing around with this

475 00:47:11.320 00:47:12.589 Miguel de Veyra: Sorry. What did you send.

476 00:47:12.990 00:47:14.790 Casie Aviles: Sorry it’s in zoom chat.

477 00:47:20.680 00:47:22.290 Miguel de Veyra: Oh, shit! You got execution time.

478 00:47:23.791 00:47:28.320 Casie Aviles: It’s this is just something I tried with. Any 10

479 00:47:29.530 00:47:31.519 Casie Aviles: if we could, you know, have

480 00:47:31.760 00:47:41.770 Casie Aviles: execution times input and then I I hide. I hit the column, for now the output column, because it takes up a lot of the screen.

481 00:47:41.770 00:47:42.640 Miguel de Veyra: Yeah, yeah.

482 00:47:43.650 00:47:51.579 Casie Aviles: But something like this where we have like this Llm. As a judge score. So this.

483 00:47:51.580 00:47:56.950 Uttam Kumaran: So like. Why, but why isn’t this out of the box like, I don’t want our team to develop

484 00:47:57.110 00:48:01.930 Uttam Kumaran: evaluation frameworks like this is something that already exists.

485 00:48:03.550 00:48:05.670 Miguel de Veyra: Wait. Does vellum have this? Casey?

486 00:48:06.800 00:48:07.810 Miguel de Veyra: Yeah, they do.

487 00:48:08.630 00:48:09.590 Uttam Kumaran: No, they do.

488 00:48:09.590 00:48:13.219 Uttam Kumaran: By the way, Bellum has the ability to do test suites.

489 00:48:14.110 00:48:15.900 Miguel de Veyra: But we still have to code it right.

490 00:48:17.720 00:48:18.520 Casie Aviles: And.

491 00:48:19.320 00:48:24.730 Miguel de Veyra: So I think I I explored that with John, I think we still had to. We had to. Still, you know, basically

492 00:48:24.960 00:48:26.349 Miguel de Veyra: put in the work there.

493 00:48:27.680 00:48:33.359 Uttam Kumaran: I mean, dude. Yeah, of course, we have to write the test. Who does? Who the fuck’s gonna write the test for us? But like

494 00:48:33.780 00:48:35.910 Uttam Kumaran: I don’t know how like

495 00:48:36.170 00:48:41.410 Uttam Kumaran: I don’t want to, and I don’t want us to be like thinking about test types and like

496 00:48:41.590 00:48:45.740 Uttam Kumaran: we can use vellum for the test suites. It’s it’s an inbuilt.

497 00:48:46.390 00:48:49.000 Uttam Kumaran: It’s an inbuilt tool

498 00:48:57.970 00:49:00.610 Uttam Kumaran: like I. This is where I’m just like not following like.

499 00:49:01.030 00:49:03.500 Uttam Kumaran: I don’t want us to develop this product.

500 00:49:04.260 00:49:07.080 Uttam Kumaran: What I’m hearing is you want us to build this.

501 00:49:07.430 00:49:10.460 Uttam Kumaran: I don’t want to build testing frameworks.

502 00:49:10.690 00:49:16.610 Uttam Kumaran: That’s why I want to use trace loop, or I want to use telecom, or I want to use vellum they already built.

503 00:49:16.770 00:49:22.490 Uttam Kumaran: What are the most common types of tests. How do you enter them? How do you then get? How do they run

504 00:49:22.740 00:49:25.220 Uttam Kumaran: right? What is the output like?

505 00:49:25.330 00:49:33.140 Uttam Kumaran: Look at? Look at this. This is basically like what what that spreadsheet is, but I don’t. I don’t want us to build this. They these exist.

506 00:49:44.420 00:49:48.370 Uttam Kumaran: This is where, like, I just think, maybe we had like a Miss Myth like

507 00:49:48.650 00:49:55.139 Uttam Kumaran: some sort of miscommunication, because I don’t want to build testing frameworks. I want us to use them.

508 00:49:57.050 00:50:02.410 Uttam Kumaran: Part of the reason why I wanted to go develop is, and then doesn’t have this right.

509 00:50:10.350 00:50:15.739 Uttam Kumaran: I don’t want to recreate like an individual node that’s like super custom for every single thing.

510 00:50:16.780 00:50:19.430 Uttam Kumaran: So I don’t know like, what do you think like is the best move here.

511 00:50:46.280 00:50:48.229 Miguel de Veyra: Well, if that’s our goal, then

512 00:50:49.670 00:50:51.299 Miguel de Veyra: we kind of have to go with

513 00:50:51.440 00:50:53.570 Miguel de Veyra: something else other than an item. No.

514 00:50:58.140 00:51:01.339 Uttam Kumaran: Yeah, dude. That’s like why we even considered it.

515 00:51:02.120 00:51:04.669 Uttam Kumaran: But like, this is what I’m saying.

516 00:51:05.480 00:51:07.910 Uttam Kumaran: I need you to tell me what you want to do.

517 00:51:20.780 00:51:23.880 Uttam Kumaran: I need to think, yeah.

518 00:51:24.050 00:51:31.329 Uttam Kumaran: I researched a few right? I looked on trace loop. I found Helicone. But that’s what I needed. You guys to test it out and tell me

519 00:51:31.760 00:51:36.300 Uttam Kumaran: this is gonna work for us. Right? That’s why we pulled. That’s why I was like.

520 00:51:36.830 00:51:41.990 Uttam Kumaran: Okay, we’re good on vellum. Because I thought we tested it. And we were like, this is gonna be good for us to use.

521 00:51:48.200 00:51:54.870 Uttam Kumaran: So I wanna so I guess like, I, I need 2 decisions. I need one like, how do we want to build this stuff for ABC, like as we’re.

522 00:51:55.270 00:51:58.799 Uttam Kumaran: we’re sort of out of time to make a decision on infrastructure.

523 00:51:58.800 00:51:59.530 Miguel de Veyra: Yes.

524 00:52:00.800 00:52:04.310 Uttam Kumaran: So. And then I need a decision on like, how do I get?

525 00:52:04.500 00:52:09.169 Uttam Kumaran: How are we going to evaluate the models? And how are we going to get logs on the model usage?

526 00:52:11.590 00:52:13.510 Uttam Kumaran: I need those answered. This week.

527 00:52:20.960 00:52:23.719 Miguel de Veyra: How are we gonna evaluate the models and the website.

528 00:52:24.700 00:52:27.930 Uttam Kumaran: How are we gonna evaluate the agents.

529 00:52:28.510 00:52:28.890 Miguel de Veyra: Okay.

530 00:52:28.890 00:52:32.500 Uttam Kumaran: How are we going to get logs on usage?

531 00:52:43.230 00:52:46.910 Uttam Kumaran: Because otherwise, how do I know that the stuff is running? It’s working like

532 00:52:47.060 00:52:52.830 Uttam Kumaran: we can’t deploy. We can’t deploy stuff to production that that we can’t guarantee that.

533 00:53:03.880 00:53:06.210 Uttam Kumaran: So what do you guys wanna do? Do you wanna like

534 00:53:06.570 00:53:09.390 Uttam Kumaran: you want to take some more time to figure that out, Miguel, or like

535 00:53:09.500 00:53:11.479 Uttam Kumaran: talk to me here like, what do you want to do.

536 00:53:11.700 00:53:16.530 Miguel de Veyra: Yeah, I’ll speak with Casey and Jana.

537 00:53:17.200 00:53:20.750 Miguel de Veyra: And then, yeah, we’ll try to figure something out before tomorrow’s meeting.

538 00:53:23.270 00:53:32.150 Uttam Kumaran: Yeah, I mean the the only other. Th-the these are. Again, I mean, I feel like we’re having like a hundred conversations about the same 5 tools, but like arise is the

539 00:53:32.310 00:53:39.479 Uttam Kumaran: is the only other one that I know. But again, these guys, all the Tracy, they they all interact with code. So we need something that

540 00:53:40.000 00:53:45.749 Uttam Kumaran: we can use with mn, right? Because we’re not running. We’re not. We’re not using. We’re not coding our agents

541 00:53:46.270 00:53:46.980 Uttam Kumaran: cool.

542 00:53:51.020 00:53:55.396 Uttam Kumaran: Oh, yeah.

543 00:54:02.065 00:54:08.810 Uttam Kumaran: So I don’t know whether it’s like we use nam for agent building. We use something else for testing.

544 00:54:08.920 00:54:15.300 Uttam Kumaran: and you could send a request to like right, or maybe we should use vellum just for testing

545 00:54:16.120 00:54:21.923 Uttam Kumaran: and use any of them as a back end like you suggested I just need. I just need the answer.

546 00:54:23.310 00:54:24.770 Miguel de Veyra: Actually, there’s a

547 00:54:25.190 00:54:31.599 Miguel de Veyra: there was a way we could do it, but we kind of left it off with. Let me illustrate it.

548 00:54:31.970 00:54:32.699 Uttam Kumaran: I mean. But this

549 00:54:32.700 00:54:38.723 Uttam Kumaran: that’s where, like the I mean. I remember what you said, which is like we just have vellum called

550 00:54:40.260 00:54:45.230 Uttam Kumaran: yeah. But dude. I laughed it off because I didn’t know that we couldn’t do it, that it’s like

551 00:54:45.470 00:54:49.030 Uttam Kumaran: it’s. It’s like shitty to build envelope. I didn’t know that.

552 00:54:51.630 00:54:56.310 Miguel de Veyra: Oh, yeah, I mean, it’s not delicate. It’s yeah, delicious.

553 00:54:57.580 00:55:01.540 Uttam Kumaran: That’s fine. Yeah, let’s see this, then. Then. But here’s what I’m saying

554 00:55:02.410 00:55:08.240 Uttam Kumaran: is is vellum at $500 worth it, because I think if you can’t find, if like.

555 00:55:08.410 00:55:11.180 Uttam Kumaran: if we don’t have another option for testing.

556 00:55:12.600 00:55:17.100 Uttam Kumaran: then we have to do that right.

557 00:55:17.630 00:55:19.040 Uttam Kumaran: We don’t have another option.

558 00:55:19.570 00:55:21.869 Miguel de Veyra: Wait. Wait. I think I have another

559 00:55:22.170 00:55:27.640 Miguel de Veyra: thing in mind, since every tool we’re gonna be doing is basically gonna be in

560 00:55:28.090 00:55:37.170 Miguel de Veyra: deployed somewhere. Right. So it’s either going to be in the app, most likely, or streamlet. It’s gonna be through Api Casey. You played around with

561 00:55:37.500 00:55:43.009 Miguel de Veyra: trace loop right as long as it’s through Via Api, we can integrate it right.

562 00:55:43.900 00:55:44.900 Uttam Kumaran: Oh!

563 00:55:45.660 00:55:46.819 Casie Aviles: I mean. I tried with.

564 00:55:46.820 00:55:52.670 Uttam Kumaran: Trace loop needs like trace loop needs like your your actual Api call.

565 00:55:58.066 00:55:58.660 Miguel de Veyra: Shit. Okay.

566 00:56:00.630 00:56:03.100 Casie Aviles: The address looks not the best option either.

567 00:56:03.360 00:56:07.239 Miguel de Veyra: And no, I mean, just generally speaking, on how you know how it works.

568 00:56:07.520 00:56:14.539 Miguel de Veyra: It has to be, it cause it has to observe the inputs, the question response, right?

569 00:56:15.000 00:56:25.229 Casie Aviles: Yeah, the tricky part is like it has to be actually in the request for like when when we’re calling the the AI like the Api request to the AI,

570 00:56:26.030 00:56:27.070 Casie Aviles: right? Okay?

571 00:56:29.090 00:56:38.820 Casie Aviles: Which is why we couldn’t get the token usage. I mean, we could get the input from from the AI, and then the output, and then we could run evas on that. But we don’t have like the

572 00:56:39.900 00:56:45.009 Casie Aviles: like the token usage and the pricing for the cost.

573 00:56:46.240 00:56:46.690 Miguel de Veyra: Yeah.

574 00:56:46.690 00:56:49.919 Casie Aviles: Which is why it was tricky to integrate directly with Natan.

575 00:57:04.900 00:57:09.349 Uttam Kumaran: Yeah, I mean for me, I’m like.

576 00:57:11.090 00:57:17.249 Uttam Kumaran: Yeah, I I don’t know. I I need you guys to tell me at least one we need to be able to do which is evals.

577 00:57:17.420 00:57:21.789 Uttam Kumaran: So let’s just skip everything. And just let’s just use.

578 00:57:21.990 00:57:26.100 Uttam Kumaran: Let’s just use Velo for the evals. Build the agent, and then.

579 00:57:27.030 00:57:27.640 Miguel de Veyra: Okay.

580 00:57:28.760 00:57:29.810 Uttam Kumaran: Are we good with that.

581 00:57:30.280 00:57:31.050 Miguel de Veyra: Yep. Yep.

582 00:57:31.730 00:57:34.439 Uttam Kumaran: So can you? Can you at least hook up.

583 00:57:36.070 00:57:39.250 Uttam Kumaran: So we’re gonna call the agent through vellum.

584 00:57:41.770 00:57:42.610 Miguel de Veyra: Yes.

585 00:57:43.110 00:57:43.740 Uttam Kumaran: Okay.

586 00:57:44.230 00:57:45.420 Miguel de Veyra: Api. Request.

587 00:57:45.730 00:57:49.259 Uttam Kumaran: So can you make that a reality for

588 00:57:49.890 00:57:52.539 Uttam Kumaran: at least the ABC. Home? One.

589 00:57:53.180 00:57:55.070 Miguel de Veyra: Yeah, we’ll look into.

590 00:57:58.450 00:58:04.589 Uttam Kumaran: Because I mean Miguel broadly. It seems like all roads are pointing to. We have to. We’re gonna have to move up.

591 00:58:04.720 00:58:07.969 Uttam Kumaran: We’re gonna have to end up being like code. 1st from the agent side.

592 00:58:08.750 00:58:09.460 Miguel de Veyra: Yeah.

593 00:58:10.500 00:58:11.120 Uttam Kumaran: Like.

594 00:58:15.420 00:58:16.110 Miguel de Veyra: There’s something in there.

595 00:58:20.480 00:58:21.080 Uttam Kumaran: Okay.

596 00:58:21.610 00:58:24.069 Uttam Kumaran: So I think you have a decision there

597 00:58:24.680 00:58:27.490 Uttam Kumaran: that we’re going to use vellum for the testing.

598 00:58:27.970 00:58:29.740 Uttam Kumaran: But we still don’t have logs.

599 00:58:32.540 00:58:33.540 Uttam Kumaran: That’s fine.

600 00:58:34.420 00:58:38.970 Uttam Kumaran: Like, I need, I really want to try. I really want to use the evals.

601 00:58:39.420 00:58:40.469 Uttam Kumaran: So let’s

602 00:58:43.800 00:58:45.640 Uttam Kumaran: let’s try to do that.

603 00:58:50.380 00:58:53.570 Uttam Kumaran: Okay, alright, okay.

604 00:58:55.080 00:58:56.030 Uttam Kumaran: What else.

605 00:59:03.490 00:59:09.890 Miguel de Veyra: this is basically they just throw it up 6 feedback loop. This is basically is this collecting from

606 00:59:10.780 00:59:12.610 Miguel de Veyra: reinforged team members.

607 00:59:13.280 00:59:14.050 Uttam Kumaran: Yes.

608 00:59:15.341 00:59:21.300 Miguel de Veyra: So basically just establish some sort of form or something like that, or maybe an agent that, hey?

609 00:59:21.840 00:59:23.870 Uttam Kumaran: Well, we already have. We already have that right?

610 00:59:24.220 00:59:29.310 Casie Aviles: Yeah, the one on the feed AI feedback channel where you tag lead researcher.

611 00:59:29.820 00:59:30.440 Miguel de Veyra: Yeah.

612 00:59:32.170 00:59:35.320 Casie Aviles: I guess the only thing lacking is the success metrics.

613 00:59:59.660 01:00:06.940 Uttam Kumaran: Okay. But I. But also like this is, I guess this is a broader question. I want you guys to be meeting with the people on the team.

614 01:00:07.460 01:00:08.620 Miguel de Veyra: Oh, okay.

615 01:00:08.850 01:00:11.150 Uttam Kumaran: Like they are your customers, not me.

616 01:00:11.490 01:00:15.770 Uttam Kumaran: I’m quickly gonna be the least in touch person.

617 01:00:16.320 01:00:17.390 Uttam Kumaran: So

618 01:00:17.900 01:00:28.879 Uttam Kumaran: like Nico is your customer. Robert is your customer, the engineers, your customer, Hannah, is your customer. So how are you guys gonna make sure that

619 01:00:29.040 01:00:32.600 Uttam Kumaran: least things we’re building for them are working.

620 01:00:33.120 01:00:36.270 Uttam Kumaran: That’s basically what this is about.

621 01:00:36.770 01:00:44.939 Miguel de Veyra: Can we? I think just so. We can move on from this. Can we like schedule Utah? What do you think of scheduling like, you know. Maybe a

622 01:00:45.560 01:00:48.839 Miguel de Veyra: 15 min. Call with each department.

623 01:00:49.660 01:00:50.230 Uttam Kumaran: Yeah.

624 01:00:50.460 01:00:55.729 Miguel de Veyra: Yeah, I think that would be the solution. There, I’ll just talk to Mario to you know what’s the best time possible.

625 01:00:57.750 01:01:05.320 Uttam Kumaran: Start with start with, yeah, start with one person like.

626 01:01:05.880 01:01:11.189 Uttam Kumaran: start with Nico, or start with someone where you could be like, would you mind if we met every week

627 01:01:11.390 01:01:21.259 Uttam Kumaran: and just got feedback from you on how you’re using AI, and I want you to run that like, because I want to get out of the loop, because I’ll see it anyways.

628 01:01:21.260 01:01:22.115 Miguel de Veyra: Yeah,

629 01:01:22.970 01:01:31.770 Uttam Kumaran: Like. Also, I know too much, because, like, I’m gonna use everything I want. I want us to help the people who are like nervous or don’t know how to use it, like those are the people.

630 01:01:32.130 01:01:37.419 Miguel de Veyra: Okay, should we? I’m thinking of Connor, since he’s very actively using it.

631 01:01:39.000 01:01:40.100 Uttam Kumaran: Sales.

632 01:01:40.540 01:01:41.460 Uttam Kumaran: I don’t know.

633 01:01:42.840 01:01:46.573 Miguel de Veyra: Or just Nico, because you know ticket here the junior Pm. Aligns with it.

634 01:01:47.930 01:01:49.789 Uttam Kumaran: Exactly. That’s what I’m saying.

635 01:01:50.050 01:01:50.660 Miguel de Veyra: Okay.

636 01:01:53.150 01:01:55.850 Uttam Kumaran: And then we. And then basically see how we

637 01:01:56.470 01:02:02.050 Uttam Kumaran: in this one, we measured the knowledge base. We’re measuring the usage of the transcript summaries.

638 01:02:02.940 01:02:06.249 Uttam Kumaran: But this one is like.

639 01:02:07.520 01:02:13.081 Uttam Kumaran: I want you guys to think about like, how do we measure the the success of any agent? We deploy

640 01:02:14.040 01:02:15.130 Uttam Kumaran: right.

641 01:02:16.350 01:02:18.230 Miguel de Veyra: Because it’s actually from the people.

642 01:02:20.710 01:02:21.596 Miguel de Veyra: Okay, yeah.

643 01:02:24.120 01:02:26.309 Uttam Kumaran: Okay, then let’s talk about. Let’s talk about.

644 01:02:27.510 01:02:31.749 Miguel de Veyra: Wait. I’ll drag something. I’ll drag something close there.

645 01:02:32.070 01:02:34.050 Miguel de Veyra: Ta-ta-da gear!

646 01:02:52.630 01:02:53.689 Miguel de Veyra: Who don’t mean that.

647 01:02:54.670 01:02:55.396 Uttam Kumaran: I’m here.

648 01:02:55.760 01:02:56.469 Miguel de Veyra: Okay. Okay.

649 01:02:58.398 01:03:00.042 Uttam Kumaran: Okay. Thank you.

650 01:03:01.330 01:03:09.260 Miguel de Veyra: So just a quick overview of the callers and everything. This black ones are okay are basically difficulty.

651 01:03:10.120 01:03:12.430 Miguel de Veyra: Higher, of course, the more difficult it is.

652 01:03:13.630 01:03:16.489 Miguel de Veyra: and then blue ones are done.

653 01:03:17.180 01:03:25.959 Miguel de Veyra: I know blue ones are done. Yeah. And then purple ones are deliverables or sorry deliverables, like requirements or dependencies.

654 01:03:26.210 01:03:27.810 Miguel de Veyra: Orange ones are.

655 01:03:29.810 01:03:31.349 Miguel de Veyra: We’re not sure how to do yet.

656 01:03:36.400 01:03:37.050 Uttam Kumaran: Okay.

657 01:03:37.610 01:03:39.389 Miguel de Veyra: Yeah. So let’s talk about.

658 01:03:39.390 01:03:41.680 Uttam Kumaran: Like creating updating notion tickets. So

659 01:03:43.520 01:03:48.259 Uttam Kumaran: I don’t think creating is done like, I don’t know. Can we update all properties.

660 01:03:50.940 01:03:54.089 Miguel de Veyra: Sorry creating an update we can create. But we can’t update.

661 01:03:54.090 01:03:58.639 Uttam Kumaran: But if I was to create saying like, create a ticket

662 01:03:58.890 01:04:03.245 Uttam Kumaran: for this client assigned to this person, it’ll work.

663 01:04:03.730 01:04:14.229 Miguel de Veyra: I think the assigned to this person as long as because right now it’s only trained on me, you and Casey, because AI team is what we prioritize. But the clients? No, not yet.

664 01:04:15.890 01:04:18.290 Uttam Kumaran: Okay. So then it’s not working.

665 01:04:18.690 01:04:20.890 Miguel de Veyra: Because there’s no like option, and

666 01:04:20.890 01:04:23.521 Miguel de Veyra: but you see what I mean.

667 01:04:24.330 01:04:26.940 Uttam Kumaran: But you see what I mean. It’s like we can’t. It’s not usable.

668 01:04:29.680 01:04:32.161 Miguel de Veyra: I don’t think there’s a i’ll double check it.

669 01:04:35.430 01:04:42.150 Uttam Kumaran: But this is what I’m saying it. You can’t if you can’t market as done, if, like, nobody’s using it right

670 01:04:44.852 01:04:49.410 Uttam Kumaran: like today, I had to create 10 tickets manually.

671 01:05:09.230 01:05:12.670 Casie Aviles: I guess. What’s the like. The main reason why we’re we’re

672 01:05:13.370 01:05:15.559 Casie Aviles: you are creating the tickets manually.

673 01:05:17.400 01:05:19.490 Miguel de Veyra: No, because you can’t assign to clients.

674 01:05:20.650 01:05:21.400 Casie Aviles: Oh, okay.

675 01:05:21.650 01:05:22.350 Miguel de Veyra: And.

676 01:05:24.950 01:05:28.840 Uttam Kumaran: So this is what I’m saying. Also, it’s like, if you guys would have. If you talked to Nico.

677 01:05:28.970 01:05:32.140 Uttam Kumaran: you would have seen like, yeah, there’s no way. It’s not possible right now.

678 01:05:35.580 01:05:36.549 Miguel de Veyra: Let me check.

679 01:05:41.880 01:05:43.310 Miguel de Veyra: It’s not available here.

680 01:05:44.320 01:05:47.260 Miguel de Veyra: Status, project, parent.

681 01:05:49.490 01:05:51.539 Uttam Kumaran: So. But that’s what I’m saying, like.

682 01:05:51.890 01:05:54.079 Uttam Kumaran: I don’t think any of these are, Doc.

683 01:05:59.800 01:06:03.609 Uttam Kumaran: because I’m still doing this manually.

684 01:06:06.010 01:06:13.080 Uttam Kumaran: Cause this is what I’m saying. It’s you I you really need to go meet with the people that are creating the tickets every day and be like are

685 01:06:13.260 01:06:18.710 Uttam Kumaran: walk me through creating a ticket. And then you guys just can automate that like, so

686 01:06:20.650 01:06:24.336 Uttam Kumaran: are you, can you guys go? Yeah, like, that’s where we’re at with these.

687 01:06:24.740 01:06:31.040 Uttam Kumaran: I still don’t think these. I don’t think we’re close. I think we’re we’re close. But like, alright.

688 01:06:32.200 01:06:33.300 Uttam Kumaran: yeah, yeah.

689 01:06:33.300 01:06:34.099 Miguel de Veyra: Yeah, yeah.

690 01:06:34.100 01:06:36.191 Uttam Kumaran: So what do you want to do.

691 01:06:38.820 01:06:41.020 Miguel de Veyra: The client, the client.

692 01:06:41.590 01:06:44.939 Uttam Kumaran: Service, probably 3 or 4.

693 01:06:44.940 01:06:47.900 Miguel de Veyra: We need to find the wait. Let me share my screen.

694 01:06:49.180 01:06:53.799 Uttam Kumaran: No dude. We need to be able to modify every single property on the ticket.

695 01:06:56.580 01:06:58.190 Uttam Kumaran: I don’t think you’re understanding like

696 01:06:58.360 01:07:02.600 Uttam Kumaran: I want us to use the ticket bought to create any ticket.

697 01:07:07.800 01:07:08.490 Uttam Kumaran: Okay?

698 01:07:10.710 01:07:17.230 Uttam Kumaran: And I I, it’s not clear to me what what is remaining and what is impossible.

699 01:07:18.120 01:07:20.020 Uttam Kumaran: And that’s what like. So.

700 01:07:20.250 01:07:21.699 Miguel de Veyra: You know. Oh.

701 01:07:21.810 01:07:33.979 Miguel de Veyra: so the thing is basically what we can’t do here, or at least I don’t know what how to do yet is, we can’t assign it to a to a client

702 01:07:34.280 01:07:37.639 Miguel de Veyra: like there’s just no minute. There’s no way in any. Then to do it.

703 01:07:41.550 01:07:44.189 Miguel de Veyra: You can’t change the property to the client.

704 01:07:44.770 01:07:48.230 Miguel de Veyra: There’s there’s no client option, or

705 01:07:48.800 01:07:53.009 Miguel de Veyra: is it? Maybe it’s wait. Let me check if it’s parent item, I’m not sure.

706 01:07:55.550 01:08:00.620 Miguel de Veyra: Maybe it’s parent item, but I don’t know.

707 01:08:08.990 01:08:13.434 Miguel de Veyra: Point it could be parent. Item, wait!

708 01:08:25.979 01:08:29.940 Miguel de Veyra: Who tell me in the tasks like, Wait, let me share my screen. Actually

709 01:08:33.640 01:08:38.699 Miguel de Veyra: like over here. This is considered like the parent item, right or no.

710 01:08:40.100 01:08:42.350 Miguel de Veyra: No, it’s just a it’s just the relation.

711 01:08:44.260 01:08:45.920 Uttam Kumaran: It’s a relationship.

712 01:08:45.920 01:08:50.129 Miguel de Veyra: Yeah, cause, here’s our like, sub, maybe sub item, no, not really.

713 01:08:50.819 01:08:54.029 Uttam Kumaran: No, no, those are. Those are all the names of the columns.

714 01:08:55.729 01:08:56.839 Miguel de Veyra: Yeah, there’s no.

715 01:08:58.765 01:09:04.630 Uttam Kumaran: I know, dude, but like I don’t wanna code wrong. This I I want you to like.

716 01:09:05.611 01:09:11.570 Uttam Kumaran: My. My thing is like I. It’s not clear to me what else we need to do

717 01:09:11.859 01:09:20.539 Uttam Kumaran: like what else is remaining in order for us to start using this to make tickets. One of this is the client, right? But like, I need to see that really, clearly.

718 01:09:21.260 01:09:22.480 Uttam Kumaran: like a list.

719 01:09:24.260 01:09:33.519 Uttam Kumaran: You see what I mean, like, what are all the things people do when they create tickets? How far are we from checking each of those off? Because

720 01:09:33.770 01:09:37.990 Uttam Kumaran: we could use the notion, Api, for something we could use Zapier for something like

721 01:09:38.250 01:09:40.750 Uttam Kumaran: just telling me that it’s not an N. 8 N.

722 01:09:41.140 01:09:42.929 Uttam Kumaran: I’m not. That’s not enough for me.

723 01:09:46.750 01:09:48.040 Miguel de Veyra: Okay.

724 01:09:48.040 01:09:53.090 Uttam Kumaran: There’s a bunch of different ways. We can do it. So take a step back, and, like you, go back to the mural board

725 01:09:54.020 01:10:00.730 Uttam Kumaran: each of these, it’s not enough to say, Oh, okay, you can create a ticket now, it’s like you would.

726 01:10:01.000 01:10:03.648 Uttam Kumaran: There’s a reason why people aren’t using it

727 01:10:05.410 01:10:10.399 Uttam Kumaran: is the reason because they don’t know how to use it? Or is the reason because it can’t do everything that’s for you to find out.

728 01:10:10.710 01:10:10.995 Miguel de Veyra: You.

729 01:10:14.020 01:10:16.630 Uttam Kumaran: So start with one of these, right.

730 01:10:17.190 01:10:23.619 Uttam Kumaran: start with one, and then go the distance, find out all the different ways we create tickets

731 01:10:23.790 01:10:28.149 Uttam Kumaran: and try to label. Okay, we need to automate this and automate this. We need to automate this.

732 01:10:28.430 01:10:28.890 Uttam Kumaran: Oh.

733 01:10:29.740 01:10:36.930 Uttam Kumaran: and I I don’t know if Miro is the best place or notion. But like, yeah, that’s the core of clarity that I need to see.

734 01:10:37.310 01:10:39.555 Miguel de Veyra: Okay. Well, I’ll draft something on.

735 01:10:44.720 01:10:52.729 Uttam Kumaran: Okay. And then I know we’re we’re sort of running low on time. I wanna talk about like the roadmap for like ABC stuff.

736 01:10:54.055 01:10:54.510 Miguel de Veyra: Yeah.

737 01:11:00.913 01:11:01.979 Miguel de Veyra: Can you see my screen?

738 01:11:04.830 01:11:07.000 Uttam Kumaran: Yeah, okay.

739 01:11:07.000 01:11:10.006 Miguel de Veyra: Yeah. So this was done by me and Casey last Friday.

740 01:11:10.470 01:11:12.849 Uttam Kumaran: So basically there’s 2 main parts.

741 01:11:13.500 01:11:18.280 Miguel de Veyra: The rag agent and this other entire path

742 01:11:21.948 01:11:23.901 Miguel de Veyra: so the way I guess this is.

743 01:11:24.560 01:11:31.660 Miguel de Veyra: I was, gonna ask you about this like, how do we want the updates to work? Will we give them like an entire Ui.

744 01:11:33.180 01:11:35.220 Miguel de Veyra: basically, all the records are there.

745 01:11:35.495 01:11:35.770 Uttam Kumaran: And.

746 01:11:35.770 01:11:38.969 Miguel de Veyra: Stored in some sort of database. And then they can.

747 01:11:39.110 01:11:43.680 Miguel de Veyra: We’ll create basically a documentation type of agent. And then they can, you know, update it.

748 01:11:44.740 01:11:49.810 Uttam Kumaran: There’s no but we, we agreed. We’re not. There’s no Ui. It’s all the Google through Google Chat.

749 01:11:50.540 01:11:53.100 Miguel de Veyra: No, no, for the rug, but for the update.

750 01:11:55.700 01:11:59.330 Uttam Kumaran: I guess I’m not convinced that there needs to be a ui, even for the update

751 01:12:04.830 01:12:05.629 Uttam Kumaran: but it’s.

752 01:12:12.250 01:12:14.150 Miguel de Veyra: So where will we store? There.

753 01:12:15.360 01:12:19.458 Uttam Kumaran: Well walk me through, walk me through what we’re doing as part of the update, and then.

754 01:12:20.230 01:12:21.649 Miguel de Veyra: Yeah, okay, so one question.

755 01:12:21.650 01:12:22.230 Uttam Kumaran: Yeah, well.

756 01:12:22.230 01:12:27.730 Miguel de Veyra: Really, yeah, we’re gonna have 2 Dbs, right? This is the normal. dB, not the vectorized one.

757 01:12:28.790 01:12:36.700 Miguel de Veyra: And then this is basically where all their, you know, raw records will be up so all the numbers.

758 01:12:37.072 01:12:38.560 Uttam Kumaran: I wanna pause there.

759 01:12:39.070 01:12:45.269 Uttam Kumaran: We one of the goals in the project is that we’re gonna rewrite their files.

760 01:12:46.300 01:12:51.169 Uttam Kumaran: So did we come to a conclusion on whether this is like one large file.

761 01:12:51.410 01:12:58.130 Uttam Kumaran: We’re gonna have several files like the goal is not to their documents suck.

762 01:12:58.280 01:13:00.960 Uttam Kumaran: So we have to rewrite it like.

763 01:13:01.330 01:13:04.750 Uttam Kumaran: did we come to an agreement on like, what’s what’s the plan? There.

764 01:13:13.770 01:13:17.510 Miguel de Veyra: You mean, like rewrite older documents in a way that it’s

765 01:13:19.250 01:13:22.179 Miguel de Veyra: readable by the factor dB. Or readable by people.

766 01:13:23.256 01:13:27.126 Uttam Kumaran: No, I mean either, because there’s like, look we have.

767 01:13:28.240 01:13:33.800 Uttam Kumaran: I mean, there’s 2 things right. One. We got all these documents but dude. Their document format is bad now.

768 01:13:34.360 01:13:34.990 Miguel de Veyra: Yeah.

769 01:13:34.990 01:13:35.700 Uttam Kumaran: Right?

770 01:13:36.590 01:13:45.689 Uttam Kumaran: So one of the goals is to rewrite is to is to think about a new architecture for their documents.

771 01:13:46.459 01:13:51.819 Uttam Kumaran: My question for you guys, is that is that like one large like Google, Doc.

772 01:13:52.240 01:13:57.250 Uttam Kumaran: is that several databases is that tables like.

773 01:14:01.950 01:14:02.710 Miguel de Veyra: Yeah.

774 01:14:03.350 01:14:03.770 Casie Aviles: Yeah. Dude.

775 01:14:03.770 01:14:04.539 Miguel de Veyra: So the way.

776 01:14:04.540 01:14:05.900 Casie Aviles: Anything to a dB, right?

777 01:14:05.900 01:14:07.280 Miguel de Veyra: To a TV, yeah.

778 01:14:08.600 01:14:11.940 Casie Aviles: And then the reason why we thought of this is like

779 01:14:13.150 01:14:16.609 Casie Aviles: like, for if it’s like purely AI to update the

780 01:14:16.910 01:14:21.190 Casie Aviles: the documents it might cost for some hallucinations. So

781 01:14:21.870 01:14:25.789 Casie Aviles: the idea behind this we wanted to set up like a ui, and we assumed that

782 01:14:26.434 01:14:28.609 Casie Aviles: they didn’t need to chat

783 01:14:29.511 01:14:33.490 Casie Aviles: chat for it like the chat to update. So I mean.

784 01:14:33.890 01:14:40.639 Casie Aviles: yes, I think we did this to add more determinism to it, like, you know, less reliant on the AI part.

785 01:14:41.320 01:14:42.030 Casie Aviles: Yeah.

786 01:14:43.520 01:14:49.669 Uttam Kumaran: But I guess, like I think, that assumption I want to challenge like I don’t. I don’t think that

787 01:14:50.440 01:14:54.819 Uttam Kumaran: I’m I’m not like. Tell me why we need a ui to do that.

788 01:14:57.210 01:15:00.699 Uttam Kumaran: Because who said we need to compare

789 01:15:02.550 01:15:07.110 Uttam Kumaran: right? Like we jumped. You’re jumping, I think. Start with the core problem

790 01:15:07.260 01:15:12.680 Uttam Kumaran: don’t jump to the product solution yet, right? The core fundamental problem is

791 01:15:13.560 01:15:20.720 Uttam Kumaran: when the customer goes and wants to update a record, they don’t know which document to go make the update

792 01:15:21.010 01:15:28.580 Uttam Kumaran: or several documents to make the update right? Putting another ui is not the problem here.

793 01:15:29.340 01:15:37.689 Uttam Kumaran: The problem is the actual update, right? So this is where you have the ability to think creatively about what is the document structure

794 01:15:38.190 01:15:42.179 Uttam Kumaran: that best allows for this sort of update process?

795 01:15:42.570 01:15:43.780 Uttam Kumaran: You wanna have

796 01:15:44.010 01:15:52.700 Uttam Kumaran: 10 like, is there 10 documents? Is there like one document and you upload update multiple places? But I don’t see why we need a ui for this.

797 01:16:06.580 01:16:12.609 Miguel de Veyra: So if they’re gonna update, for example, they wanna change something in on the all of the existing document.

798 01:16:15.200 01:16:20.470 Miguel de Veyra: Do they just say we do we follow the Google drive thing or create a new one?

799 01:16:21.370 01:16:27.249 Uttam Kumaran: But this is what I’m saying. Like I, this is why you need to make a decision on like, are you even going to?

800 01:16:27.540 01:16:31.860 Uttam Kumaran: If you’re gonna keep Google drive, is there gonna be just one master document

801 01:16:38.480 01:16:40.991 Uttam Kumaran: like, that’s what I need like.

802 01:16:41.410 01:16:41.890 Miguel de Veyra: Because.

803 01:16:41.890 01:16:46.169 Uttam Kumaran: Your assumption is that we’re keeping the same structure. I’m telling you that that’s not true.

804 01:16:46.630 01:16:51.209 Uttam Kumaran: Right? Why would we keep the same 50 documents

805 01:16:57.980 01:17:02.720 Uttam Kumaran: like I, I wanna break it down like, do we need to keep the same 50 documents?

806 01:17:04.840 01:17:09.761 Uttam Kumaran: Because fundamentally, the reason why they’re in this problem is because they have 50 documents.

807 01:17:10.220 01:17:11.020 Uttam Kumaran: Right?

808 01:17:13.900 01:17:15.089 Uttam Kumaran: Okay? I see.

809 01:17:16.490 01:17:19.859 Uttam Kumaran: So we have the opportunity to change how the data is stored.

810 01:17:20.060 01:17:24.620 Uttam Kumaran: Now you’re you need to. You need to figure out is that one document

811 01:17:25.050 01:17:36.279 Uttam Kumaran: is that like several? Right? So you need to do like an exercise. You need to look through all the documents they gave, and sort of figure out how. What’s the best way to store

812 01:17:37.040 01:17:40.500 Uttam Kumaran: that information? I don’t know. That’s like one. Yes.

813 01:17:40.910 01:17:45.430 Uttam Kumaran: that’s the knowledge engineering part like. I don’t know what’s the best format to structure this.

814 01:17:55.880 01:18:01.180 Miguel de Veyra: Well. But will they still look at those docs, or is it just purely the bot? Looking at those docs.

815 01:18:07.210 01:18:10.850 Uttam Kumaran: Let’s see, like who’s looking at the docs if they’re using the chat. Bot!

816 01:18:18.550 01:18:23.120 Miguel de Veyra: Then they bring it up in the meeting that sometimes they might want to double check or.

817 01:18:27.110 01:18:29.510 Uttam Kumaran: Unlike? What? What are they? Double, checking.

818 01:18:30.856 01:18:34.969 Miguel de Veyra: It’s something Yvette brought up on the call, like if it’s actually correct.

819 01:18:35.220 01:18:35.820 Uttam Kumaran: All of them.

820 01:18:36.550 01:18:38.750 Miguel de Veyra: Like the what the bot provided. Basically.

821 01:18:41.280 01:18:52.429 Miguel de Veyra: that’s why they wanted, you know, for it. Like to show like where it came from. It came from basically the excerpts, the excerpts it, and then they could click it. There.

822 01:18:52.430 01:18:55.409 Uttam Kumaran: It’s it’s an excerpt from something right.

823 01:18:55.880 01:18:56.540 Miguel de Veyra: Yeah.

824 01:18:57.340 01:18:58.299 Uttam Kumaran: Come back.

825 01:19:00.200 01:19:02.459 Uttam Kumaran: Do you see what I’m saying? Though, overall like

826 01:19:02.940 01:19:05.970 Uttam Kumaran: I I don’t know. I feel like I’m like we’re just missing each other, like

827 01:19:06.930 01:19:14.129 Uttam Kumaran: maintaining 50 documents is wrong, like the 50 documents is why there’s a problem.

828 01:19:15.960 01:19:18.889 Uttam Kumaran: So we have to reduce the amount of documents.

829 01:19:19.300 01:19:30.009 Uttam Kumaran: But you guys need to tell me, does that one document do you need like multiple documents? This is a hard question, like, you have to spend some time and go look at every single document.

830 01:19:30.190 01:19:35.660 Uttam Kumaran: Call Yvette, talk to her about what are each of these documents? Map them all out like

831 01:19:37.230 01:19:46.879 Uttam Kumaran: you have to go through that process simply like being like, okay, we’re gonna like, yeah, dude. If it. If this was purely throw everything into rag, tell me where the knowledge is.

832 01:19:47.480 01:19:49.480 Uttam Kumaran: I would have been done. We’re already done.

833 01:19:49.650 01:19:51.100 Miguel de Veyra: There’s a reason.

834 01:19:51.310 01:19:57.849 Uttam Kumaran: Why, this is a challenge, right? So you need to think through like, is this a large 50 page, Google, Doc?

835 01:19:58.130 01:20:02.309 Uttam Kumaran: And then when when they ask a question, you can say, this is on page 5.

836 01:20:02.660 01:20:05.170 Uttam Kumaran: Is is this a like a mix?

837 01:20:05.540 01:20:09.599 Uttam Kumaran: Is there no Google, Doc like you have to figure. You have to decide that.

838 01:20:13.940 01:20:14.520 Miguel de Veyra: Thank you.

839 01:20:14.520 01:20:15.080 Miguel de Veyra: Yeah.

840 01:20:15.960 01:20:18.880 Uttam Kumaran: Because ultimately you like.

841 01:20:19.150 01:20:29.379 Uttam Kumaran: if people need to go double check, that means the chat bots aren’t working right because all the Csrs need to be using the Chat Bot to get information. Not this, not the documents.

842 01:20:29.500 01:20:34.600 Uttam Kumaran: Then, ultimately it’s up to you to understand. Do the people who are Updating the documents.

843 01:20:34.980 01:20:41.320 Uttam Kumaran: Can they use the chat bot for all of those updates? And in order to be confident in the updates, what do they need?

844 01:20:41.450 01:20:43.760 Uttam Kumaran: Start with? That’s the fundamental question.

845 01:20:44.020 01:20:49.089 Uttam Kumaran: Not like, throw every us other assumption out the window right?

846 01:20:54.640 01:20:57.499 Miguel de Veyra: Because actually, what what I did for

847 01:20:57.910 01:21:09.360 Miguel de Veyra: the knowledge base is that we could. I turn it all into Pdf first, st and then basically even the Powerpoint ones. There’s Powerpoints in there. There’s Google Excels and stuff like that.

848 01:21:10.570 01:21:23.100 Miguel de Veyra: So there’s a lot of different types of documents. So I just converted them all into Pdfs. And then all Powerpoints, all documents and all Csvs. Excels. I concatenated them into one big document.

849 01:21:23.800 01:21:26.319 Miguel de Veyra: and then I just uploaded them to the TV.

850 01:21:26.460 01:21:27.070 Uttam Kumaran: Cool thanks.

851 01:21:27.070 01:21:29.570 Miguel de Veyra: Technically the one file thing I already did.

852 01:21:30.160 01:21:31.549 Uttam Kumaran: Back to the software.

853 01:21:32.390 01:21:34.179 Miguel de Veyra: Now it’s a matter of you know.

854 01:21:35.270 01:21:40.209 Uttam Kumaran: But, for example, when you, if if you just take a Google Sheet and put into a Pdf.

855 01:21:40.750 01:21:43.490 Uttam Kumaran: how do you know that 2 Google sheets are related.

856 01:21:44.840 01:21:45.959 Miguel de Veyra: Yeah, that’s the thing.

857 01:21:47.420 01:21:48.020 Miguel de Veyra: Yeah.

858 01:21:48.020 01:21:48.420 Uttam Kumaran: So.

859 01:21:48.420 01:21:50.030 Miguel de Veyra: I just so all in.

860 01:21:50.030 01:21:51.950 Uttam Kumaran: But yeah, but

861 01:21:53.860 01:22:04.990 Uttam Kumaran: like, I guess. But dude, you’re the you’re the engineer you have to. You need to know what every single document means, you know, right?

862 01:22:07.270 01:22:12.440 Uttam Kumaran: Because just because the demo worked and we could get it. Yeah, we already knew we could do that. This is more of like.

863 01:22:13.650 01:22:17.750 Uttam Kumaran: there’s knowledge in there that’s not just like pdfable and chunkable.

864 01:22:17.970 01:22:28.679 Uttam Kumaran: So what do we do with that? Right? So really, what I’m hoping is that the team can look at every single document and propose a solution for how to store this knowledge long term.

865 01:22:29.550 01:22:30.150 Miguel de Veyra: Yep.

866 01:22:30.660 01:22:40.060 Uttam Kumaran: Because this is where, if they still need the slides, then maybe they still need to be able to maintain those. But that’s a question that you gotta ask them.

867 01:22:40.440 01:22:44.960 Uttam Kumaran: And you gotta think about like, okay, put yourself in their shoes.

868 01:22:45.110 01:22:50.919 Uttam Kumaran: Think about us like we have document like, think about our company. We have slides. We have notion, like, what will.

869 01:22:51.660 01:22:57.359 Uttam Kumaran: how, how will we manage? Right? Okay. Maybe they do need to have a place to keep slides up to date?

870 01:22:58.170 01:23:06.760 Uttam Kumaran: Great. But then do they still need Google Docs? Or do they want to? Maybe? Or maybe what we do is we consolidate all of the text into Google, Doc.

871 01:23:07.290 01:23:10.320 Uttam Kumaran: And then there, it’s like a mat. It’s like a basically like a book.

872 01:23:10.560 01:23:14.079 Uttam Kumaran: Right? This is what I don’t. I don’t know this is for you guys to figure out.

873 01:23:14.520 01:23:15.610 Uttam Kumaran: I don’t know yet.

874 01:23:16.090 01:23:20.439 Uttam Kumaran: I don’t know what the solution is, because unless it’s like that our rag is gonna suck.

875 01:23:27.900 01:23:28.580 Uttam Kumaran: Okay.

876 01:23:35.500 01:23:41.779 Miguel de Veyra: We’ll make for Bible or something.

877 01:23:44.620 01:23:47.819 Miguel de Veyra: And then there’s like, Yeah, we could do that, I think.

878 01:23:48.420 01:23:51.459 Miguel de Veyra: ABC, Bible where everything else. And then.

879 01:23:56.860 01:24:02.020 Miguel de Veyra: yeah, let’s arrange the docs first, st before we even create or update.

880 01:24:05.860 01:24:15.099 Uttam Kumaran: Yeah, go through every single document and like, check out what every single thing is like. There’s no way to like. I don’t have the answer for you.

881 01:24:15.550 01:24:17.920 Miguel de Veyra: Yeah, it’s I’ll look into it.

882 01:24:18.490 01:24:19.080 Uttam Kumaran: Okay.

883 01:24:20.160 01:24:23.359 Uttam Kumaran: So that’s what I want to see. Like, I don’t want us to jump to the solution

884 01:24:23.970 01:24:29.929 Uttam Kumaran: like. And I don’t. My, my assumption is like, we don’t need a ui to do this like

885 01:24:30.630 01:24:33.070 Uttam Kumaran: I don’t want to say, use another application.

886 01:24:33.370 01:24:38.550 Uttam Kumaran: I want them to try like I would like us to constrain to the Google Chatbot.

887 01:24:38.780 01:24:45.970 Uttam Kumaran: But then your question is okay, what can we can we accomplish all the tasks needed? So break down

888 01:24:46.290 01:25:02.470 Uttam Kumaran: what the event needs to do what Janice needs to do, what the Csrs need to do, and then you can verify all of that possible via just the chat or what changes you need to make to the document structure, what change we need to make to rag the technology to make that happen. So start from there.

889 01:25:02.750 01:25:04.659 Uttam Kumaran: But question the assumptions. Yeah.

890 01:25:07.560 01:25:14.719 Uttam Kumaran: So how do we wanna like proceed on the ABC stuff, should we? I wanna talk again at least once before

891 01:25:15.820 01:25:20.510 Uttam Kumaran: Friday, but I don’t. Wanna I wanna give you guys some time to digest all this.

892 01:25:20.510 01:25:21.109 Miguel de Veyra: Yeah, yeah.

893 01:25:21.110 01:25:26.349 Uttam Kumaran: I guess, Miguel, I’m sort of. I know you’re sort of taking on a lot, but luckily it’s just.

894 01:25:27.010 01:25:28.650 Uttam Kumaran: It’s just a team of 3.

895 01:25:28.780 01:25:37.389 Uttam Kumaran: But I do want to hand this off to you, to sort of own to start to build these roadmaps out.

896 01:25:39.370 01:25:43.050 Miguel de Veyra: Okay, yeah, sure. Sure, I’ll see what I can do.

897 01:25:45.236 01:25:54.659 Miguel de Veyra: Yeah, I have to review every document. Because I asked actually asked Jan earlier if she can transition full time just in case we need.

898 01:25:55.700 01:25:58.560 Miguel de Veyra: I don’t think she can. As of now. So yeah.

899 01:26:01.510 01:26:02.510 Uttam Kumaran: Okay.

900 01:26:03.180 01:26:05.630 Miguel de Veyra: Where the hell did I store everything?

901 01:26:08.240 01:26:09.870 Miguel de Veyra: Brain 1st clients ABC.

902 01:26:09.870 01:26:12.489 Uttam Kumaran: Wait. But, dude, that’s another. That’s a separate issue.

903 01:26:12.710 01:26:13.530 Miguel de Veyra: Yeah. Yeah.

904 01:26:13.710 01:26:16.610 Miguel de Veyra: Oh, here’s the files. Okay, yeah.

905 01:26:16.610 01:26:21.050 Uttam Kumaran: I’m a yeah. Yeah. Just tell me, dude, because we’re

906 01:26:21.710 01:26:23.869 Uttam Kumaran: we’re coming up to the middle of the month.

907 01:26:25.903 01:26:38.529 Uttam Kumaran: We’re already one. We’re we’re a few days into ABC, I just really wanna make sure that this can actually happen. Because based on this call, we’re behind, like, I thought we were pretty aligned on like what

908 01:26:39.060 01:26:47.230 Uttam Kumaran: they needed. Have you? Have you listened to the the the loom yet?

909 01:26:47.470 01:26:51.380 Miguel de Veyra: Ever since Friday. No, no good. I I over the weekend.

910 01:26:51.380 01:26:53.500 Uttam Kumaran: Okay, so take a step back.

911 01:26:53.840 01:26:57.480 Uttam Kumaran: look at every document. Listen to the loom

912 01:26:57.870 01:27:02.870 Uttam Kumaran: like, really try to understand this project, because otherwise it’s it’s not gonna happen

913 01:27:04.890 01:27:18.699 Uttam Kumaran: like, Jana is definitely not gonna understand? So you really need to go really understanding what we’re doing for them. We’re still a little bit far. Yeah, that would be really helpful. Because otherwise.

914 01:27:19.270 01:27:24.600 Uttam Kumaran: yeah, I’m gonna have to break it down. It’s gonna be. It’s gonna take a lot of work, I mean, and fundamentally like

915 01:27:25.380 01:27:27.550 Uttam Kumaran: I’m here to. I’m here to help. But I

916 01:27:27.950 01:27:34.920 Uttam Kumaran: I don’t want it to be like, just go do, Xyz. That’s kind of the whole theme of even the Pm. Stuff like, I want you guys to go

917 01:27:35.260 01:27:41.080 Uttam Kumaran: start managing these projects like for Casey.

918 01:27:41.380 01:27:42.650 Uttam Kumaran: Your direct

919 01:27:42.840 01:27:53.659 Uttam Kumaran: clients are on our team. You don’t. There’s no need to wait for me to find out that they’re not using the ticket creation bot, or that the ticket creation bot can’t do everything we need right?

920 01:27:54.410 01:27:58.600 Uttam Kumaran: It shouldn’t be a surprise. So those are the things I want to push you guys to do.

921 01:27:58.720 01:28:00.559 Uttam Kumaran: It’s gonna be tough. But like.

922 01:28:01.180 01:28:03.730 Uttam Kumaran: run towards you guys all all the

923 01:28:04.930 01:28:07.379 Uttam Kumaran: the time and support I can give you. So.

924 01:28:08.280 01:28:11.549 Uttam Kumaran: But I want this team to be to be more self sufficient. We’re

925 01:28:11.830 01:28:19.740 Uttam Kumaran: we don’t have that many clients. We’re only working on a few things. So really want us to go, because before we bring on more people here, we really need to

926 01:28:20.200 01:28:26.365 Uttam Kumaran: feel like we’re we can take on even these. These sorts of smaller things end to end, you know.

927 01:28:27.720 01:28:28.460 Uttam Kumaran: Okay.

928 01:28:29.290 01:28:29.810 Miguel de Veyra: Okay.

929 01:28:32.570 01:28:33.280 Uttam Kumaran: Cool.

930 01:28:33.710 01:28:36.539 Uttam Kumaran: Alright. Then I’ll I mean we’ll talk at. Stand up, but

931 01:28:36.960 01:28:40.589 Uttam Kumaran: let me know if I can come in and review anything else.

932 01:28:45.690 01:28:48.499 Uttam Kumaran: Thank you. Alright, thanks, guys, thanks, guys.

933 01:28:49.010 01:28:49.700 Uttam Kumaran: Bye.