Meeting Title: Daily AI Team Sync Date: 2025-01-21 Meeting participants: Miguel De Veyra, Casie Aviles, Uttam Kumaran
WEBVTT
1 00:02:04.240 ⇒ 00:02:05.830 Miguel de Veyra: Alili, Tushopin.
2 00:02:08.509 ⇒ 00:02:09.149 Casie Aviles: Okay.
3 00:02:17.469 ⇒ 00:02:22.829 Miguel de Veyra: As soon as the details didn’t complete. We need, because I think I need to be like a lot more specific
4 00:02:29.279 ⇒ 00:02:31.099 Miguel de Veyra: to get your processor.
5 00:02:47.979 ⇒ 00:02:57.189 Miguel de Veyra: Indiana ticket processor. Simple pass on Indian Indian Pino parsume data Sota Manganu.
6 00:03:02.489 ⇒ 00:03:03.649 Miguel de Veyra: Make of bits.
7 00:03:07.300 ⇒ 00:03:12.100 Casie Aviles: Yeah, but then nothing, I guess, but then nothing check. I I’ll take a look then.
8 00:03:12.950 ⇒ 00:03:14.380 Casie Aviles: in case so yeah.
9 00:03:14.380 ⇒ 00:03:15.437 Miguel de Veyra: Near the towers.
10 00:03:15.900 ⇒ 00:03:24.270 Miguel de Veyra: Is it? Alright, yeah, not anymore. Multiple tickets to learn in a go on. Yeah, clean up.
11 00:03:31.200 ⇒ 00:03:36.120 Miguel de Veyra: Oh, notion, sa agent.
12 00:03:40.390 ⇒ 00:03:44.779 Miguel de Veyra: never so. I don’t know my share screen now.
13 00:03:50.160 ⇒ 00:03:54.260 Miguel de Veyra: Pardon, share screen transcripts.
14 00:03:57.650 ⇒ 00:03:59.530 Miguel de Veyra: Miguel Miguel.
15 00:04:05.500 ⇒ 00:04:08.990 Miguel de Veyra: ML. 4 PE. Dot T.
16 00:04:09.920 ⇒ 00:04:10.850 Miguel de Veyra: Pandemic.
17 00:04:11.280 ⇒ 00:04:12.150 Casie Aviles: Yes.
18 00:04:19.260 ⇒ 00:04:31.359 Miguel de Veyra: Imprompt. As basically, I added, I wait actually, just give to give you an overview like, basically you
19 00:04:31.950 ⇒ 00:04:39.280 Miguel de Veyra: UN details. Pog, lagrade, informat, note template.
20 00:04:40.550 ⇒ 00:04:46.400 Miguel de Veyra: So each tool.
21 00:04:47.080 ⇒ 00:04:51.119 Miguel de Veyra: Yeah. So, marketing campaign for product launch, stamina.
22 00:04:52.730 ⇒ 00:04:55.900 Miguel de Veyra: Which is, you know.
23 00:04:55.900 ⇒ 00:04:56.319 Casie Aviles: Hey? It wasn’t.
24 00:04:56.320 ⇒ 00:04:58.150 Miguel de Veyra: Sure I’m not right.
25 00:04:58.450 ⇒ 00:04:59.170 Uttam Kumaran: Hey, guys.
26 00:04:59.650 ⇒ 00:05:00.490 Miguel de Veyra: Hey Witham.
27 00:05:00.790 ⇒ 00:05:01.440 Uttam Kumaran: Okay.
28 00:05:03.020 ⇒ 00:05:12.460 Miguel de Veyra: And then the thing. Now wait. Sorry. Now I’ve switched languages and then, now structural, just correct.
29 00:05:12.780 ⇒ 00:05:16.269 Miguel de Veyra: But when it passed it to notion, I don’t understand.
30 00:05:19.270 ⇒ 00:05:20.539 Casie Aviles: Just by the reports.
31 00:05:20.540 ⇒ 00:05:23.540 Miguel de Veyra: It didn’t. Yeah, it didn’t get the output.
32 00:05:27.430 ⇒ 00:05:31.500 Miguel de Veyra: But this is the last part, though it’s like it’s complete. After this.
33 00:05:34.640 ⇒ 00:05:37.640 Casie Aviles: So it’s supposed to create the ticket right?
34 00:05:37.830 ⇒ 00:05:42.459 Miguel de Veyra: It does create the ticket, but it’s randomizing it right now, because
35 00:05:42.670 ⇒ 00:05:47.359 Miguel de Veyra: it’s supposed to be basing the ticket details the one we mapped out individually.
36 00:05:49.180 ⇒ 00:05:50.240 Miguel de Veyra: Wait. Sorry.
37 00:05:50.410 ⇒ 00:05:53.210 Miguel de Veyra: Yeah, this one’s right dependency.
38 00:05:53.660 ⇒ 00:05:57.020 Miguel de Veyra: It’s supposed to come from here, which is the structured one.
39 00:05:58.380 ⇒ 00:06:02.329 Miguel de Veyra: But then I can’t really map this out anywhere.
40 00:06:08.280 ⇒ 00:06:10.070 Uttam Kumaran: Wait. Is that a question for me?
41 00:06:10.430 ⇒ 00:06:12.760 Miguel de Veyra: I don’t know. We’re just debugging me and Casey sorry.
42 00:06:12.760 ⇒ 00:06:13.890 Uttam Kumaran: Oh! God! Damn!
43 00:06:14.265 ⇒ 00:06:14.640 Miguel de Veyra: Okay.
44 00:06:15.830 ⇒ 00:06:21.380 Miguel de Veyra: well, I guess there’s a way we can do this, Casey, just create another workflow and then pass it.
45 00:06:21.500 ⇒ 00:06:25.490 Miguel de Veyra: Pass the details right? So it’s workflows within workflows, within workflows.
46 00:06:26.800 ⇒ 00:06:27.969 Casie Aviles: Yeah, yeah, okay.
47 00:06:27.970 ⇒ 00:06:30.089 Miguel de Veyra: I think that’s the way to do it. But yeah, I think that.
48 00:06:30.250 ⇒ 00:06:32.520 Miguel de Veyra: Okay, let’s be more general now.
49 00:06:36.400 ⇒ 00:06:39.489 Casie Aviles: Yeah, I can go in a deeper debugging session if you want.
50 00:06:41.640 ⇒ 00:06:43.229 Miguel de Veyra: Yeah, let’s do after this call.
51 00:06:48.950 ⇒ 00:06:53.699 Uttam Kumaran: Okay. Cool? Yeah. I guess. Miguel, you wanna go first.st
52 00:06:57.253 ⇒ 00:06:58.620 Miguel de Veyra: Yeah, sure.
53 00:06:59.310 ⇒ 00:07:01.989 Miguel de Veyra: So yeah, right now, just working on the
54 00:07:02.450 ⇒ 00:07:05.809 Miguel de Veyra: ticket to the agent, as you guys saw just a moments earlier.
55 00:07:06.333 ⇒ 00:07:12.860 Miguel de Veyra: It’s almost done. It’s just, you know, some bugs going on. It was working when it wasn’t
56 00:07:12.990 ⇒ 00:07:16.460 Miguel de Veyra: workflows within workflows. But now we have to adjust some stuff around.
57 00:07:17.580 ⇒ 00:07:23.420 Miguel de Veyra: But yeah, that’s pretty much. It already creates the tasks and everything. But yeah, I believe we still need to
58 00:07:24.960 ⇒ 00:07:27.730 Miguel de Veyra: improve it. So it’s production ready.
59 00:07:29.230 ⇒ 00:07:32.660 Miguel de Veyra: And then I think that’s pretty much it on my end.
60 00:07:38.750 ⇒ 00:07:40.060 Miguel de Veyra: Casey, do you want to go next.
61 00:07:40.510 ⇒ 00:07:46.760 Uttam Kumaran: For the ticket. So how far away from the ticketing thing working, I guess. Tell me about the bug that you guys are looking at.
62 00:07:47.967 ⇒ 00:08:09.040 Miguel de Veyra: Yeah. So basically, cause right now, it’s 2. It’s basically 2 workflows. Right? So one is the conversational agent where you know. Cause eventually we want it to be, hey, you know. Here we give it the meeting, and then we’re gonna create 3 tickets. So we need the conversational agent, but the one that actually processes the
63 00:08:10.220 ⇒ 00:08:11.230 Miguel de Veyra: the
64 00:08:11.430 ⇒ 00:08:18.880 Miguel de Veyra: what do you call this that processes the creation of the ticket is just an Llm. Chain, which is just a tool that we pass details into
65 00:08:19.050 ⇒ 00:08:29.260 Miguel de Veyra: individually. They’re working. But now that we connected them, they’re still working. But it’s not accurate. So what another thing we could do is basically convert the
66 00:08:30.190 ⇒ 00:08:36.450 Miguel de Veyra: the notion tool to another workflow that way. It’s like very specific. What we pass it.
67 00:08:37.409 ⇒ 00:08:41.879 Uttam Kumaran: Oh, so the notion tool and right now isn’t like it’s like making stuff up.
68 00:08:43.440 ⇒ 00:08:48.669 Miguel de Veyra: Oh, not really making stuff up, but it’s not accurate.
69 00:08:50.250 ⇒ 00:08:53.690 Uttam Kumaran: Can you show me what the like, what what the input.
70 00:08:54.070 ⇒ 00:08:54.560 Miguel de Veyra: Okay. Sure.
71 00:08:54.560 ⇒ 00:08:55.420 Uttam Kumaran: Mentors are.
72 00:08:56.560 ⇒ 00:08:58.600 Miguel de Veyra: Okay? Sure. Let me share some.
73 00:09:02.350 ⇒ 00:09:08.350 Miguel de Veyra: Okay. So for example, once the agent, you know, delegates like, Hey, create a task. It’s gonna go in like this.
74 00:09:10.680 ⇒ 00:09:16.780 Miguel de Veyra: I already formatted the response of the agent to be.
75 00:09:17.160 ⇒ 00:09:19.549 Miguel de Veyra: you know, basically match our template
76 00:09:22.460 ⇒ 00:09:24.470 Miguel de Veyra: additional notes. Yeah, basically, this one.
77 00:09:26.382 ⇒ 00:09:29.100 Miguel de Veyra: So once we accept it. Here
78 00:09:30.650 ⇒ 00:09:34.580 Miguel de Veyra: you can see we pass the data. We process it here
79 00:09:34.930 ⇒ 00:09:41.130 Miguel de Veyra: right this one is also trained on the template and then return it as a Json, and then we map it here
80 00:09:42.060 ⇒ 00:09:46.380 Miguel de Veyra: so as you can see, this should have been, you know, simpler now.
81 00:09:49.710 ⇒ 00:09:52.340 Miguel de Veyra: Oh, now this one works so it’s not consistent.
82 00:09:53.220 ⇒ 00:09:56.130 Miguel de Veyra: Yeah, it’s ideally, every run should be like this.
83 00:09:57.600 ⇒ 00:10:03.050 Miguel de Veyra: But for some reason the one earlier wasn’t. So let me see there the error one.
84 00:10:05.340 ⇒ 00:10:11.730 Casie Aviles: And sometimes like the output that the agent like the Llm. Chain, inconsistent.
85 00:10:12.070 ⇒ 00:10:15.269 Miguel de Veyra: Yeah. So, for example, this one, the output is.
86 00:10:16.940 ⇒ 00:10:18.930 Miguel de Veyra: I don’t know what happened there. Yeah.
87 00:10:18.930 ⇒ 00:10:23.020 Uttam Kumaran: How there’s no, there shouldn’t be an Llm. At this step, right? Because this is just like
88 00:10:25.500 ⇒ 00:10:27.230 Uttam Kumaran: filling in the properties.
89 00:10:28.560 ⇒ 00:10:31.389 Miguel de Veyra: No. This one is the basically the formatter.
90 00:10:31.960 ⇒ 00:10:36.389 Uttam Kumaran: Oh, so you’re saying it takes in. Oh, so then this is not working.
91 00:10:37.927 ⇒ 00:10:41.860 Uttam Kumaran: So why is there an error right now like, what’s that error.
92 00:10:42.773 ⇒ 00:10:46.500 Miguel de Veyra: Basically, it needs a title here.
93 00:10:46.500 ⇒ 00:10:49.489 Uttam Kumaran: Okay. So so go back to the.
94 00:10:49.660 ⇒ 00:10:53.509 Uttam Kumaran: So you’re saying the Json output from the processor isn’t always full.
95 00:10:54.320 ⇒ 00:11:00.689 Miguel de Veyra: Doesn’t. All because not always like, I think it’s sometimes doesn’t work, because, as you can see here.
96 00:11:00.940 ⇒ 00:11:04.379 Miguel de Veyra: so this one worked. I believe so if you go here.
97 00:11:04.560 ⇒ 00:11:06.610 Miguel de Veyra: the Json output is actually correct.
98 00:11:07.010 ⇒ 00:11:07.640 Uttam Kumaran: Okay.
99 00:11:08.270 ⇒ 00:11:09.329 Miguel de Veyra: So it’s just not.
100 00:11:09.330 ⇒ 00:11:12.680 Uttam Kumaran: This is what this is. Here’s a way to. Here’s 1 way to avoid this.
101 00:11:13.230 ⇒ 00:11:19.059 Uttam Kumaran: you basically need to loop until everything is filled. So go back to the canvas.
102 00:11:19.230 ⇒ 00:11:26.039 Uttam Kumaran: You need to have a check between processor and notion to look that all items are filled.
103 00:11:26.180 ⇒ 00:11:33.109 Uttam Kumaran: If they’re not filled, then you need to probably go back and ask the user for more information.
104 00:11:33.430 ⇒ 00:11:40.519 Uttam Kumaran: So this is the thing that I wanted to do. Is not every like, not every message in slack has all the context we need.
105 00:11:41.160 ⇒ 00:11:50.029 Uttam Kumaran: So if you’re going ahead, if you’re gonna tell the ticket processor to go make a ticket, it may be missing like it may have no idea, for example, that this isn’t.
106 00:11:50.490 ⇒ 00:11:56.950 Uttam Kumaran: This is an AI team ticket. Right? So one thing we need to do is verify that all fields
107 00:11:57.170 ⇒ 00:12:01.370 Uttam Kumaran: have values, and for the fields that don’t have values.
108 00:12:01.500 ⇒ 00:12:07.890 Uttam Kumaran: I think the simplest way right now is just to ask the user again, what is the value?
109 00:12:09.370 ⇒ 00:12:11.220 Miguel de Veyra: Okay, okay. So if the actual.
110 00:12:11.220 ⇒ 00:12:12.030 Uttam Kumaran: Right title.
111 00:12:12.050 ⇒ 00:12:16.569 Uttam Kumaran: and I and I would say, an even better way of doing this is actually, if you go back to canvas
112 00:12:17.700 ⇒ 00:12:19.949 Uttam Kumaran: is to not do this step here.
113 00:12:20.350 ⇒ 00:12:20.940 Miguel de Veyra: Which one did you.
114 00:12:20.940 ⇒ 00:12:24.179 Uttam Kumaran: Step before you get to the ticket processor.
115 00:12:25.150 ⇒ 00:12:26.689 Miguel de Veyra: Oh, here you mean over here.
116 00:12:26.690 ⇒ 00:12:30.400 Uttam Kumaran: Because cause go to the ticket processor prompt.
117 00:12:34.110 ⇒ 00:12:37.320 Uttam Kumaran: Can you just zoom in a little bit? Okay? No, that’s fine. That’s fine.
118 00:12:42.780 ⇒ 00:12:48.420 Uttam Kumaran: Yeah, like some tasks you’re not going to get so can so go. Can you go to the result?
119 00:12:49.450 ⇒ 00:12:51.180 Uttam Kumaran: Oh, let’s just scroll down.
120 00:12:51.550 ⇒ 00:12:52.630 Uttam Kumaran: I’m sorry. Yeah.
121 00:12:55.260 ⇒ 00:13:00.290 Uttam Kumaran: So then how does this get processed into Json for the for the notion.
122 00:13:00.877 ⇒ 00:13:03.379 Miguel de Veyra: We have like this structured output parser.
123 00:13:05.110 ⇒ 00:13:07.820 Uttam Kumaran: And then this is also another Llm.
124 00:13:10.410 ⇒ 00:13:13.670 Miguel de Veyra: Not specifically no cause. It’s not connected anywhere.
125 00:13:14.010 ⇒ 00:13:17.090 Miguel de Veyra: It’s just the structure of responses. Basically.
126 00:13:17.640 ⇒ 00:13:20.390 Uttam Kumaran: So this takes the response from.
127 00:13:21.160 ⇒ 00:13:22.890 Miguel de Veyra: From ticket processor.
128 00:13:22.890 ⇒ 00:13:24.820 Uttam Kumaran: So click on ticket processor again.
129 00:13:24.820 ⇒ 00:13:25.120 Miguel de Veyra: Yep.
130 00:13:25.810 ⇒ 00:13:29.010 Uttam Kumaran: So this just takes anything that we send in.
131 00:13:29.860 ⇒ 00:13:30.610 Miguel de Veyra: Yes.
132 00:13:30.900 ⇒ 00:13:32.220 Uttam Kumaran: See? That’s the problem.
133 00:13:32.920 ⇒ 00:13:38.059 Uttam Kumaran: The the goal is not to fill in the blanks. The goal is to go back
134 00:13:39.030 ⇒ 00:13:58.960 Uttam Kumaran: to the customer when there’s not enough information cause I don’t want the AI to hallucinate and be like I was never given specs. So I’m gonna make up the specs that’s useless right instead, the the benefit here is that all of this one can happen in slack and 2 instead of
135 00:13:59.160 ⇒ 00:14:05.049 Uttam Kumaran: instead of you writing a ticket and missing something. The AI is automatically checking the quality of the ticket.
136 00:14:06.080 ⇒ 00:14:08.390 Uttam Kumaran: So I think that you should
137 00:14:08.500 ⇒ 00:14:15.829 Uttam Kumaran: probably move. Have some sort of check before it gets to the processor
138 00:14:16.270 ⇒ 00:14:24.499 Uttam Kumaran: that makes sure that it has or it could be at the processor level. But basically it needs to loop until you have all the information you need.
139 00:14:24.650 ⇒ 00:14:28.499 Uttam Kumaran: It’s almost similar to the thing we built for ABC.
140 00:14:29.000 ⇒ 00:14:37.199 Miguel de Veyra: I think, yeah, I think how I approach that is just I’ll go just here, you know. If you decide, I’ll just prompt this a lot more stricter, I would say.
141 00:14:37.400 ⇒ 00:14:39.550 Miguel de Veyra: Like, if you have, if you
142 00:14:39.830 ⇒ 00:14:44.369 Miguel de Veyra: you know, if you decide to create 4 tickets, if you think there’s 4 tickets to create.
143 00:14:44.750 ⇒ 00:14:48.029 Miguel de Veyra: clarify it before creating the tickets
144 00:14:49.330 ⇒ 00:14:54.289 Miguel de Veyra: and then provide the details. And then, if the user says yes and then create the tickets.
145 00:14:56.060 ⇒ 00:14:58.459 Miguel de Veyra: I think that’s the simplest way to do it.
146 00:15:00.020 ⇒ 00:15:00.869 Miguel de Veyra: But yeah.
147 00:15:00.870 ⇒ 00:15:08.649 Uttam Kumaran: That’s yeah. That’s sort of what I want to see. Because I would say, 90% of the tickets request you’re gonna get are not going to have enough information.
148 00:15:08.650 ⇒ 00:15:09.210 Miguel de Veyra: Yep.
149 00:15:10.985 ⇒ 00:15:16.809 Uttam Kumaran: However, like the what we’re augmenting here is the process of going from that
150 00:15:17.090 ⇒ 00:15:24.219 Uttam Kumaran: going to notion clicking new ticket, writing the ticket out right. It’s probably like a 5 to 10 min process
151 00:15:24.510 ⇒ 00:15:30.697 Uttam Kumaran: instead. That can all happen in slack in front of everybody in 90 seconds.
152 00:15:31.790 ⇒ 00:15:40.619 Uttam Kumaran: The second thing is ideally. This ticket processor long term has understanding about the rest of our notion.
153 00:15:40.750 ⇒ 00:15:44.240 Uttam Kumaran: because most of our tickets are gonna land in like 10 categories.
154 00:15:45.040 ⇒ 00:15:45.720 Miguel de Veyra: Yeah, yeah.
155 00:15:45.720 ⇒ 00:15:48.449 Uttam Kumaran: So it’ll get smarter over time. Hopefully.
156 00:15:48.750 ⇒ 00:15:53.730 Miguel de Veyra: Yep, actually, it’s most probably just gonna be the agent
157 00:15:54.150 ⇒ 00:15:58.640 Miguel de Veyra: that we need to fill. Because the processor is basically just processing.
158 00:15:59.210 ⇒ 00:16:00.659 Miguel de Veyra: It’s not really thinking.
159 00:16:00.970 ⇒ 00:16:02.320 Uttam Kumaran: Oh, okay, okay.
160 00:16:02.600 ⇒ 00:16:08.579 Uttam Kumaran: then the A, yeah. Basically, the agent needs to chat in the same thread until it has all the information.
161 00:16:08.730 ⇒ 00:16:09.290 Miguel de Veyra: Yeah.
162 00:16:10.460 ⇒ 00:16:12.459 Miguel de Veyra: But yeah, that one was
163 00:16:12.710 ⇒ 00:16:21.090 Miguel de Veyra: that one was pretty much we were. Gonna work was gonna work on. It’s just this one. This annoying thing is the one that’s you know.
164 00:16:21.460 ⇒ 00:16:24.069 Uttam Kumaran: Yeah. But you know why it’s happening. It’s because, like.
165 00:16:24.680 ⇒ 00:16:26.530 Uttam Kumaran: I don’t think it has everything
166 00:16:26.870 ⇒ 00:16:34.169 Uttam Kumaran: like, I think if you if you pass, if you pass the ticket processor like a more well defined
167 00:16:35.080 ⇒ 00:16:39.039 Uttam Kumaran: like with every field in, I think it’ll this will get more accurate, is my guess.
168 00:16:40.132 ⇒ 00:16:45.949 Miguel de Veyra: Yep, yep. I mean, it’s kinda all like the specifics are already here. And then, you know.
169 00:16:47.220 ⇒ 00:16:50.200 Miguel de Veyra: avoid summarization. So it’s full context.
170 00:16:51.025 ⇒ 00:16:57.410 Miguel de Veyra: Yeah, I’ll I’ll try passing in like an entire transcript, or be a somewhat of a transcript.
171 00:16:58.430 ⇒ 00:17:01.839 Miguel de Veyra: But yeah, I think that’s probably
172 00:17:03.080 ⇒ 00:17:09.369 Miguel de Veyra: what do I work on next. It’s just I need to sort something. Yeah, but you’re correct here. This was probably the way to do it.
173 00:17:09.630 ⇒ 00:17:11.520 Miguel de Veyra: If there’s no title, don’t do it
174 00:17:12.150 ⇒ 00:17:14.619 Miguel de Veyra: because it’s creating so many tickets right now.
175 00:17:15.270 ⇒ 00:17:17.229 Miguel de Veyra: But yeah, I think that’s pretty much it.
176 00:17:18.109 ⇒ 00:17:18.729 Uttam Kumaran: Okay.
177 00:17:25.563 ⇒ 00:17:32.099 Casie Aviles: Yeah. So do you guys want to talk about the automations next.
178 00:17:33.210 ⇒ 00:17:37.031 Uttam Kumaran: Yeah, I guess for me that I wanna know about like
179 00:17:38.130 ⇒ 00:17:42.499 Uttam Kumaran: like when we can get the AI team automations for the meetings done.
180 00:17:44.860 ⇒ 00:17:45.680 Uttam Kumaran: That is.
181 00:17:46.800 ⇒ 00:17:49.660 Casie Aviles: Or was that? Is it for the Zoom Meeting.
182 00:17:50.200 ⇒ 00:17:50.970 Uttam Kumaran: Yes.
183 00:17:53.557 ⇒ 00:18:02.189 Casie Aviles: What what else do we need for that? By the way, because I mean, whenever there’s a meeting, it’s all already being sent to the AI test channel.
184 00:18:02.870 ⇒ 00:18:06.620 Casie Aviles: So yeah, all all the meetings, regardless of the type.
185 00:18:10.820 ⇒ 00:18:12.240 Uttam Kumaran: Wait! Say that one more time.
186 00:18:12.870 ⇒ 00:18:17.719 Casie Aviles: I mean, for for all the meetings that that have, like a transcript
187 00:18:18.616 ⇒ 00:18:25.259 Casie Aviles: it goes through the process and the the summary I mean, the summary is already sent to the AI Test channel.
188 00:18:25.740 ⇒ 00:18:28.050 Uttam Kumaran: But I guess I want it sent to the AI Channel.
189 00:18:28.840 ⇒ 00:18:32.369 Uttam Kumaran: meaning, if there’s an AI meeting, send to the AI Channel.
190 00:18:32.510 ⇒ 00:18:35.460 Uttam Kumaran: If there’s a sales meeting. Send it to the Sales Channel right.
191 00:18:36.490 ⇒ 00:18:37.910 Casie Aviles: Oh, okay. Okay. Yeah.
192 00:18:38.710 ⇒ 00:18:42.399 Uttam Kumaran: Can we look through? Can we look through a couple of the past examples?
193 00:18:45.030 ⇒ 00:18:47.720 Uttam Kumaran: And just see like cause? I was looking yesterday.
194 00:18:47.940 ⇒ 00:18:50.400 Uttam Kumaran: and let me just pull them up.
195 00:19:04.520 ⇒ 00:19:07.759 Uttam Kumaran: for example, this meeting happened today with, and
196 00:19:08.020 ⇒ 00:19:10.989 Uttam Kumaran: this is a this was a marketing meeting, right?
197 00:19:11.330 ⇒ 00:19:15.199 Uttam Kumaran: So one there’s no, right now. We don’t have the marketing category.
198 00:19:18.380 ⇒ 00:19:23.790 Uttam Kumaran: I don’t know it was our. Did our did a meeting? Did our meeting yesterday pop up here, or.
199 00:19:24.788 ⇒ 00:19:27.850 Casie Aviles: See, for example, this isn’t it? This is our meeting right.
200 00:19:38.940 ⇒ 00:19:43.820 Uttam Kumaran: But also, didn’t we talk about putting this all under just a thread like threading everything.
201 00:19:48.461 ⇒ 00:19:55.270 Casie Aviles: I think that’s for for the sales one or good.
202 00:19:55.490 ⇒ 00:20:03.289 Uttam Kumaran: Can we do that? Can we do it for all of them? Because, again, if this starts getting sent everywhere in slack, it’s gonna be very annoying.
203 00:20:04.570 ⇒ 00:20:07.449 Casie Aviles: What do you want to be threaded? Exactly because
204 00:20:08.380 ⇒ 00:20:15.400 Casie Aviles: we have the transcript only and the the categorization the above. Can you see the one the message above
205 00:20:15.640 ⇒ 00:20:17.749 Casie Aviles: the meeting categorization step.
206 00:20:18.410 ⇒ 00:20:18.920 Uttam Kumaran: Yeah.
207 00:20:18.920 ⇒ 00:20:19.810 Casie Aviles: This one.
208 00:20:19.810 ⇒ 00:20:22.860 Uttam Kumaran: This but but the summary. This message.
209 00:20:23.260 ⇒ 00:20:23.600 Casie Aviles: Yes.
210 00:20:23.600 ⇒ 00:20:24.999 Uttam Kumaran: Would be great to thread.
211 00:20:27.750 ⇒ 00:20:32.370 Casie Aviles: So you want it threaded after the meeting categorization step.
212 00:20:32.750 ⇒ 00:20:34.190 Uttam Kumaran: Yeah, like in here.
213 00:20:37.320 ⇒ 00:20:45.140 Casie Aviles: So do you also want the this message sent to the AI team, because this one’s only for test channel.
214 00:20:46.240 ⇒ 00:20:52.740 Uttam Kumaran: Yeah, no. Like, for example, this should be categorized as an AI team meeting
215 00:20:53.340 ⇒ 00:20:55.530 Uttam Kumaran: and then gets sent to the AI team channel.
216 00:20:56.960 ⇒ 00:20:58.400 Uttam Kumaran: Similarly.
217 00:21:00.220 ⇒ 00:21:02.279 Uttam Kumaran: This is a call with the data team.
218 00:21:03.760 ⇒ 00:21:05.510 Uttam Kumaran: This should get sent to the data channel
219 00:21:15.140 ⇒ 00:21:17.629 Uttam Kumaran: like, how far away are we from from that.
220 00:21:19.240 ⇒ 00:21:22.267 Casie Aviles: It’s just going to be a matter of adding more
221 00:21:22.770 ⇒ 00:21:26.600 Casie Aviles: like flows for the work for the workflow like the path.
222 00:21:26.790 ⇒ 00:21:29.599 Casie Aviles: So I just. I just have to change the channel.
223 00:21:32.980 ⇒ 00:21:35.250 Uttam Kumaran: Okay. Great, yeah, it would be great. Cause
224 00:21:37.540 ⇒ 00:21:46.919 Uttam Kumaran: it’d be great to start having the AI team. And just like, I just want to send the meeting, because right now it’s useless in here that they’re not getting any use.
225 00:21:47.080 ⇒ 00:21:50.240 Uttam Kumaran: So I wanna get them sent to the right channel.
226 00:21:50.860 ⇒ 00:21:53.520 Uttam Kumaran: But I need it to be threaded.
227 00:21:57.410 ⇒ 00:21:58.060 Casie Aviles: Okay.
228 00:22:23.580 ⇒ 00:22:26.690 Uttam Kumaran: So is that like, you think that’s a good, we can do that today.
229 00:22:27.680 ⇒ 00:22:28.400 Casie Aviles: Yeah, yeah.
230 00:22:28.970 ⇒ 00:22:29.740 Uttam Kumaran: Okay, cool.
231 00:22:30.000 ⇒ 00:22:36.929 Uttam Kumaran: And then I guess the only thing is, how’s it going with with stuff for Craig? It’s not the highest priority. But
232 00:22:38.769 ⇒ 00:22:43.719 Uttam Kumaran: like I, I would definitely want to see this done. The stuff for Craig is secondary.
233 00:22:45.360 ⇒ 00:22:52.580 Casie Aviles: Okay, then, yeah, for his, for his automation. And we’re already creating the drafts.
234 00:22:52.860 ⇒ 00:22:56.792 Casie Aviles: So my, I guess I’m just testing a lot right now.
235 00:22:57.720 ⇒ 00:23:02.089 Casie Aviles: I I guess the only problem that I typically encounter is with the
236 00:23:02.210 ⇒ 00:23:07.190 Casie Aviles: is with looking for the email addresses of the the contacts.
237 00:23:07.610 ⇒ 00:23:15.239 Casie Aviles: So sometimes I, I get the email addresses from from Apollo. But
238 00:23:15.430 ⇒ 00:23:20.010 Casie Aviles: sometimes we I it’s not there. So and then. And then.
239 00:23:20.360 ⇒ 00:23:25.389 Casie Aviles: But when I try to just Google, the the contact you could get the Linkedin.
240 00:23:25.750 ⇒ 00:23:30.670 Casie Aviles: And then you could just yeah, there’s it’s just I’m not sure if it’s the best solution right now with
241 00:23:30.980 ⇒ 00:23:34.030 Casie Aviles: what I did. But yeah, I might be
242 00:23:34.160 ⇒ 00:23:37.739 Casie Aviles: just overthinking these, the implementation. But.
243 00:23:38.770 ⇒ 00:23:42.489 Uttam Kumaran: So for this right now we’re not, can. We’re not saving the
244 00:23:44.291 ⇒ 00:23:46.160 Uttam Kumaran: We’re not saving the email here.
245 00:23:47.496 ⇒ 00:23:48.149 Casie Aviles: No no.
246 00:23:49.500 ⇒ 00:23:53.939 Uttam Kumaran: Is that like? Can we also do that cause? I feel like that’s what Craig was also asking for.
247 00:23:53.940 ⇒ 00:23:57.259 Casie Aviles: You mean the by the the email itself, like the draft itself.
248 00:23:57.420 ⇒ 00:23:58.310 Uttam Kumaran: Yeah.
249 00:24:00.502 ⇒ 00:24:05.589 Casie Aviles: Yeah, no, it’s not there, but it it is great being created as a draft on his account.
250 00:24:07.780 ⇒ 00:24:08.510 Uttam Kumaran: Okay.
251 00:24:11.850 ⇒ 00:24:15.980 Casie Aviles: What we’re yeah. What we’re saving here is, he added. This email column.
252 00:24:15.980 ⇒ 00:24:16.580 Uttam Kumaran: That’s where.
253 00:24:16.580 ⇒ 00:24:19.360 Casie Aviles: The email address is being saved.
254 00:24:20.280 ⇒ 00:24:26.789 Uttam Kumaran: Oh, so you’re saying, when someone gets added here, it takes all of this and updates the email.
255 00:24:27.530 ⇒ 00:24:29.909 Casie Aviles: Yes, that’s ideally.
256 00:24:30.060 ⇒ 00:24:31.709 Casie Aviles: That’s the ideal output.
257 00:24:37.390 ⇒ 00:24:38.060 Uttam Kumaran: Okay?
258 00:24:40.900 ⇒ 00:24:42.750 Uttam Kumaran: And then, what’s the blocker right now?
259 00:24:44.440 ⇒ 00:24:50.789 Casie Aviles: Yeah for the email, like, sometimes I get the email through Apollo. But sometimes I don’t.
260 00:24:51.580 ⇒ 00:24:53.650 Casie Aviles: But when I just Google, the
261 00:24:53.780 ⇒ 00:24:56.879 Casie Aviles: the contact, for example, I take a name, and then
262 00:24:57.120 ⇒ 00:25:00.469 Casie Aviles: I find their Linkedin URL, and then I could use that.
263 00:25:01.400 ⇒ 00:25:07.960 Casie Aviles: So I guess the solution that I have is a bit rigid.
264 00:25:08.160 ⇒ 00:25:10.829 Casie Aviles: so it only works one way like it takes
265 00:25:11.160 ⇒ 00:25:15.700 Casie Aviles: just the name and the company name, and then uses those 2 enrich.
266 00:25:26.180 ⇒ 00:25:31.140 Uttam Kumaran: I mean. But isn’t this? This is, we’re not using the same thing. We built for the lead research agent.
267 00:25:32.717 ⇒ 00:25:41.540 Casie Aviles: Yeah, it’s the same one. But yeah, like, sometimes, yeah, sometimes there’s no email at all.
268 00:25:41.720 ⇒ 00:25:45.151 Casie Aviles: Like, for example, there’s also sometimes the company.
269 00:25:46.260 ⇒ 00:25:53.890 Casie Aviles: for example, the one below called Viv, I think, yeah, yeah, that one.
270 00:25:54.160 ⇒ 00:25:58.750 Casie Aviles: When I try to input that into Apollo, it’s not gonna show up.
271 00:25:59.030 ⇒ 00:25:59.900 Casie Aviles: And
272 00:26:04.590 ⇒ 00:26:08.940 Casie Aviles: yeah, like when when you just Google it, you can find the the URL.
273 00:26:13.730 ⇒ 00:26:14.750 Uttam Kumaran: But here’s the email.
274 00:26:15.840 ⇒ 00:26:17.769 Casie Aviles: Yeah, yeah, I mean. But on Apollo.
275 00:26:18.740 ⇒ 00:26:20.599 Uttam Kumaran: I’m in Singapore.
276 00:26:21.650 ⇒ 00:26:25.049 Uttam Kumaran: But was our lead research agent only using Apollo.
277 00:26:25.480 ⇒ 00:26:27.940 Casie Aviles: Yeah, yeah, it’s only using my phone.
278 00:26:29.550 ⇒ 00:26:35.499 Casie Aviles: So I guess that’s where the limitation comes from. Like I said, when you Google it, you’ll find the email. But.
279 00:26:36.140 ⇒ 00:26:37.880 Uttam Kumaran: Okay. So so
280 00:26:45.830 ⇒ 00:26:52.520 Uttam Kumaran: one thing we can do is we have, we have a really robust email finding process in clay
281 00:26:55.640 ⇒ 00:26:58.570 Uttam Kumaran: that I think we should just start to leverage
282 00:27:10.070 ⇒ 00:27:11.729 Uttam Kumaran: like, for example.
283 00:27:22.180 ⇒ 00:27:23.420 Uttam Kumaran: let me find a good question.
284 00:27:45.180 ⇒ 00:27:51.539 Uttam Kumaran: See? We in Apollo we use like 10 things to get the email.
285 00:27:54.700 ⇒ 00:28:00.930 Uttam Kumaran: So how about I ask Ericsson to see whether he can build this process for us in another workbook?
286 00:28:09.240 ⇒ 00:28:10.700 Uttam Kumaran: do you see what I mean? Though.
287 00:28:12.510 ⇒ 00:28:13.040 Casie Aviles: Yeah, yeah.
288 00:28:13.040 ⇒ 00:28:18.924 Uttam Kumaran: Like we’re using Apollo. But we’re also using 10 other things, basically
289 00:28:21.460 ⇒ 00:28:25.900 Uttam Kumaran: like, for example, see this one we got from here.
290 00:28:27.640 ⇒ 00:28:30.170 Uttam Kumaran: So I’m gonna ask him how I can
291 00:28:47.980 ⇒ 00:28:50.820 Uttam Kumaran: let me ask him in sale operation.
292 00:29:21.300 ⇒ 00:29:26.690 Uttam Kumaran: So do like, how does it work? Do we just need to hit a web hook, or something, or like. How do you think this would work.
293 00:29:28.900 ⇒ 00:29:34.709 Casie Aviles: Yeah, I’m sending out the Http. Requests to Apollo.
294 00:29:35.750 ⇒ 00:29:36.440 Uttam Kumaran: Okay.
295 00:29:37.030 ⇒ 00:29:39.260 Casie Aviles: That’s how I it’s working right now
296 00:29:40.120 ⇒ 00:29:46.520 Casie Aviles: and then. The output is, you know, the the email, the contact, the enriched contact details. Basically.
297 00:30:14.030 ⇒ 00:30:19.120 Uttam Kumaran: Okay, cool. Can you follow up here? And then Erickson will help. Because we
298 00:30:19.510 ⇒ 00:30:22.189 Uttam Kumaran: basically, we’re we’re doing this already. So
299 00:30:23.880 ⇒ 00:30:29.109 Uttam Kumaran: but okay, but otherwise the emails are working like, how are you? Are you logging into his email? Or how are you verifying.
300 00:30:30.140 ⇒ 00:30:33.300 Casie Aviles: Us, you mean, how do we verify the email addresses.
301 00:30:33.510 ⇒ 00:30:34.909 Uttam Kumaran: Well, no. The email draft.
302 00:30:37.370 ⇒ 00:30:38.390 Casie Aviles: I have
303 00:30:39.230 ⇒ 00:30:46.829 Casie Aviles: sorry how you mean, how we verify. If the email is working like we can send emails to that email address.
304 00:30:46.830 ⇒ 00:30:51.579 Uttam Kumaran: No, no like. How, how can we verify like? How do you know that the draft was created in Craig’s.
305 00:30:52.000 ⇒ 00:30:56.429 Casie Aviles: Oh, okay, so it’s on Zapier, like you could see.
306 00:30:56.430 ⇒ 00:30:57.540 Uttam Kumaran: Says, success.
307 00:30:57.930 ⇒ 00:30:59.470 Casie Aviles: Yes, yes, but.
308 00:30:59.470 ⇒ 00:31:02.939 Uttam Kumaran: Can. We? Is there? Can you add? Can you put it here as well.
309 00:31:03.897 ⇒ 00:31:05.139 Casie Aviles: Yeah, yeah, we can do that.
310 00:31:06.790 ⇒ 00:31:09.769 Uttam Kumaran: Okay. Cause I don’t wanna log into his email and check.
311 00:31:10.440 ⇒ 00:31:12.430 Uttam Kumaran: I want to be able to like, audit it here.
312 00:31:16.050 ⇒ 00:31:20.170 Uttam Kumaran: Okay, okay. So hopefully, this will unblock you.
313 00:31:21.660 ⇒ 00:31:23.649 Uttam Kumaran: But keep working on the
314 00:31:24.100 ⇒ 00:31:33.670 Uttam Kumaran: the meeting agent. I want us like, I want us to go live with that across all meetings like we’ve been sitting on it for like 2 weeks. Now I want us to get it out out of the door.
315 00:31:34.718 ⇒ 00:31:46.420 Uttam Kumaran: The biggest thing. I’ll just add what we talked about for the so meeting agent, one spread summaries.
316 00:31:46.700 ⇒ 00:31:59.510 Uttam Kumaran: 0 initial message, second is, start classify, start mapping meeting.
317 00:32:36.520 ⇒ 00:32:37.270 Uttam Kumaran: Okay?
318 00:32:39.030 ⇒ 00:32:41.200 Uttam Kumaran: Cool. I’m like, Yeah, let’s get this out the door.
319 00:32:44.520 ⇒ 00:32:47.629 Uttam Kumaran: And then, Miguel, tell me how that notion stuff works.
320 00:32:50.320 ⇒ 00:32:57.680 Uttam Kumaran: and there’s several examples for us to try. So you should just maybe you can even try it. I know you’re trying in the AI test channel. But
321 00:32:58.205 ⇒ 00:33:00.750 Uttam Kumaran: if you want to try it in some other channels. Let me know.
322 00:33:02.430 ⇒ 00:33:05.290 Miguel de Veyra: The sorry, the notion. What do you mean? Right.
323 00:33:05.690 ⇒ 00:33:06.340 Uttam Kumaran: Yeah.
324 00:33:08.210 ⇒ 00:33:11.880 Miguel de Veyra: Wait. I don’t think we’ve connected the notion, but in slack. To be honest.
325 00:33:11.880 ⇒ 00:33:14.150 Uttam Kumaran: No, no, not in the notion. Bot! The ticket tickets here.
326 00:33:14.150 ⇒ 00:33:20.050 Miguel de Veyra: Add a ticket here? Yeah, sure. Where do you? Where would you want me to test it?
327 00:33:25.230 ⇒ 00:33:27.170 Uttam Kumaran: Okay, let me give you a good example.
328 00:33:36.610 ⇒ 00:33:38.630 Miguel de Veyra: I might actually just create a new channel.
329 00:33:39.280 ⇒ 00:33:39.830 Miguel de Veyra: So I don’t.
330 00:33:39.830 ⇒ 00:33:46.240 Uttam Kumaran: No, no, no, I want to use real examples. Otherwise it’s gonna continue to be like, it’s gonna continue to be fake.
331 00:33:51.160 ⇒ 00:33:55.120 Uttam Kumaran: I mean one example, or
332 00:33:55.750 ⇒ 00:33:59.495 Uttam Kumaran: I mean, why don’t we just use like one of the ones in the AI team?
333 00:34:02.310 ⇒ 00:34:04.380 Uttam Kumaran: as an example, I don’t know.
334 00:34:05.275 ⇒ 00:34:10.969 Miguel de Veyra: Yeah, right? Now, I created like a basically, me, you and Casey, just
335 00:34:11.550 ⇒ 00:34:16.550 Miguel de Veyra: demo transcript of us discussing creating a Ui dashboard, basically.
336 00:34:17.370 ⇒ 00:34:19.960 Miguel de Veyra: And it’s yeah. So it should be fine.
337 00:34:21.360 ⇒ 00:34:23.250 Uttam Kumaran: Okay. I’ll send you some examples.
338 00:34:23.250 ⇒ 00:34:25.660 Miguel de Veyra: Okay, sure. Sure. I’m testing it now.
339 00:34:26.429 ⇒ 00:34:27.019 Uttam Kumaran: Okay.
340 00:34:27.559 ⇒ 00:34:28.949 Miguel de Veyra: Can you see guys hop on a call after.
341 00:34:30.800 ⇒ 00:34:31.960 Casie Aviles: Yeah, sure. Sure.
342 00:34:31.960 ⇒ 00:34:33.209 Miguel de Veyra: Okay, thanks guys.
343 00:34:33.520 ⇒ 00:34:34.340 Uttam Kumaran: Thank you.
344 00:34:34.870 ⇒ 00:34:35.900 Miguel de Veyra: Thank you, Tanky. Bye-bye.