Meeting Title: Data x AI | Internal AI Discovery Date: 2025-04-01 Meeting participants: Amber Lin, Demilade Agboola, Miguel De Veyra
WEBVTT
1 00:00:21.780 ⇒ 00:00:22.799 Miguel de Veyra: Hello, Amber!
2 00:00:23.790 ⇒ 00:00:25.300 Amber Lin: Hello!
3 00:00:25.690 ⇒ 00:00:28.649 Miguel de Veyra: I think we’re only gonna be meeting them a lot
4 00:00:29.729 ⇒ 00:00:34.789 Amber Lin: Okay, because it’s really late for a week. Start while
5 00:00:43.078 ⇒ 00:00:45.590 Amber Lin: let’s wait for them to come
6 00:01:07.330 ⇒ 00:01:08.280 Demilade Agboola: Hi! Everyone.
7 00:01:10.280 ⇒ 00:01:11.620 Amber Lin: Hello!
8 00:01:21.570 ⇒ 00:01:28.740 Amber Lin: I think it’s only gonna be us. Bhutan probably has other meetings to do. So.
9 00:01:29.276 ⇒ 00:01:36.980 Amber Lin: Donald, I saw that you sent me a Google Doc. Thank you for that. I’m gonna pull that up right now.
10 00:01:51.240 ⇒ 00:01:55.630 Amber Lin: I’m gonna share this doc with.
11 00:01:57.100 ⇒ 00:02:00.249 Amber Lin: I’ll copy this link. I’ll send it to the chat.
12 00:02:17.510 ⇒ 00:02:19.320 Amber Lin: and whether you see my screen
13 00:02:22.970 ⇒ 00:02:24.050 Demilade Agboola: Yes, I do.
14 00:02:24.360 ⇒ 00:02:26.269 Amber Lin: Okay, perfect.
15 00:02:26.560 ⇒ 00:02:27.570 Amber Lin: So
16 00:02:27.690 ⇒ 00:02:50.759 Amber Lin: essentially, what we want to do is that the data team and the sales team is our main focus because the data team is what drives output and where we incur costs. And the sales team is where we get the revenue from. So it’s really important to us as the internal AI team to see
17 00:02:51.130 ⇒ 00:02:53.660 Amber Lin: what frustrates you the most.
18 00:02:53.980 ⇒ 00:03:07.849 Amber Lin: what time, what is the biggest area that you spend a lot of time doing manual repetitive tasks and kind of what are your wishes for a magic wand? So what do you want to get fixed?
19 00:03:08.120 ⇒ 00:03:15.919 Amber Lin: So let’s just go through the different sections, and I bet new ideas will come up along the way
20 00:03:18.000 ⇒ 00:03:21.580 Demilade Agboola: Yeah, definitely. So we could start
21 00:03:21.580 ⇒ 00:03:22.280 Amber Lin: Hmm!
22 00:03:23.470 ⇒ 00:03:27.000 Amber Lin: No, I don’t have any access. I will have to
23 00:03:27.000 ⇒ 00:03:29.479 Demilade Agboola: Just request I’ll get the email and quickly, just to them
24 00:03:49.500 ⇒ 00:03:49.795 Amber Lin: Oh.
25 00:03:53.680 ⇒ 00:03:54.490 Amber Lin: hmm!
26 00:03:57.800 ⇒ 00:03:58.730 Amber Lin: Great.
27 00:03:59.000 ⇒ 00:04:02.560 Amber Lin: So let’s go over this point.
28 00:04:03.990 ⇒ 00:04:08.779 Amber Lin: Just talk me through, because I bet you have other points. You you want to address
29 00:04:10.371 ⇒ 00:04:24.868 Demilade Agboola: I mean, basically for the data team. And I was kind of thinking of this as the Ae team. The huge win is always when we can do high quality data pipeline. So the data that comes out
30 00:04:26.290 ⇒ 00:04:33.679 Demilade Agboola: is high quality. So they’re not. You know, we don’t have any errors. It’s always up to date. It’s fresh, that sort of thing.
31 00:04:34.050 ⇒ 00:04:39.440 Demilade Agboola: and it’s been built quickly. So you know, time
32 00:04:39.800 ⇒ 00:04:47.279 Demilade Agboola: to time, from when we were assigned the task to when it has been turned around or turn around. Time is as quick as possible
33 00:04:47.990 ⇒ 00:04:49.600 Demilade Agboola: and
34 00:04:49.850 ⇒ 00:04:57.369 Demilade Agboola: good observability. So that means when things break, we are the 1st to know about it. So if anything happens
35 00:04:57.907 ⇒ 00:05:06.150 Demilade Agboola: it’s not the clients, or whatever telling us that things are broken, we are the 1st to know we’re the 1st to be able to respond.
36 00:05:06.689 ⇒ 00:05:11.079 Demilade Agboola: And I think that’s what a win just generally looks like
37 00:05:11.720 ⇒ 00:05:12.340 Amber Lin: Hmm!
38 00:05:15.760 ⇒ 00:05:18.650 Amber Lin: Sounds good. And what about this one
39 00:05:20.192 ⇒ 00:05:27.640 Demilade Agboola: So so that was more of like from a. So the 1st one was from a technical perspective.
40 00:05:27.740 ⇒ 00:05:31.179 Demilade Agboola: The second one was from like a personnel perspective.
41 00:05:32.530 ⇒ 00:05:32.920 Amber Lin: Hmm.
42 00:05:34.360 ⇒ 00:05:35.540 Demilade Agboola: So
43 00:05:35.700 ⇒ 00:05:43.940 Demilade Agboola: a win is more like, you know, we have, like a team of engineers who understand business problems and are able to
44 00:05:44.180 ⇒ 00:05:46.290 Demilade Agboola: give the right answers.
45 00:05:46.430 ⇒ 00:05:58.690 Demilade Agboola: They’re not just like taking on every task because people ask them to do it, but they really understand what the business needs from them and able to prioritize, based off that, and also able to ask the right questions
46 00:05:59.420 ⇒ 00:06:06.790 Demilade Agboola: so that they can get to what is needed of them. They’re not just taking on anything just because people are saying, Do this, do this, do this? No, they’re able to
47 00:06:07.700 ⇒ 00:06:10.280 Demilade Agboola: filter out and prioritize the important thing
48 00:06:10.280 ⇒ 00:06:11.230 Amber Lin: Indeed.
49 00:06:14.660 ⇒ 00:06:18.620 Amber Lin: I see. So you feel like there’s a lot of times
50 00:06:18.940 ⇒ 00:06:27.830 Amber Lin: the teams take on too much tasks because they don’t. They just see the tasks, and they don’t see why they’re really doing it. So then we go over time
51 00:06:28.440 ⇒ 00:06:31.349 Demilade Agboola: Yeah, sometimes we actually, sometimes there’s a y.
52 00:06:31.460 ⇒ 00:06:41.139 Demilade Agboola: but sometimes it’s not a sufficient enough. Why? So sometimes like, if, for instance, we have 5 dashboards to turn around in the next 2 weeks right
53 00:06:41.140 ⇒ 00:06:41.870 Amber Lin: Hmm.
54 00:06:42.590 ⇒ 00:06:46.089 Demilade Agboola: What dashboard should come out first, st second, 3, rd 4, th and 5, th
55 00:06:47.100 ⇒ 00:06:50.670 Demilade Agboola: always going to all come out same time. And so how do we attack that
56 00:06:50.990 ⇒ 00:06:57.960 Demilade Agboola: problem or that phase in the sense of like, hey? If the 2 weeks go over, what’s what’s that dashboard that has to come out first? st
57 00:06:58.960 ⇒ 00:07:06.830 Demilade Agboola: What dashboard can we afford to not turn around? Because sometimes you might. They might say they want 5 dashboards, but potentially they only really need 3 or 4,
58 00:07:07.010 ⇒ 00:07:12.359 Demilade Agboola: and we don’t want situations where we focused on the wrong ones.
59 00:07:12.600 ⇒ 00:07:17.819 Demilade Agboola: And now they need the important one that drives the business every day
60 00:07:17.990 ⇒ 00:07:22.020 Amber Lin: The issue with use every day, and we’ve not done that
61 00:07:22.678 ⇒ 00:07:30.959 Demilade Agboola: So that can be disastrous. It’s just that kind of like ability to be able to prioritize and ensure that we’re always going in the right direction, and the most important direction
62 00:07:33.180 ⇒ 00:07:44.130 Amber Lin: Do you think this is? I know this is very important for the data engineers, right? But it’s also a work, together with the product owner and with the project manager, right
63 00:07:46.180 ⇒ 00:07:47.300 Demilade Agboola: Yeah, definitely.
64 00:07:47.890 ⇒ 00:08:08.309 Demilade Agboola: I think just the ability for the engineers themselves not to just solely rely on the product managers or the product owners to give them the direction, but to be able to by themselves. Which is kind of why, I said, they can filter and prioritize that basis to be able to say, Hey! Or push back and pay like you’re saying. I should do this ticket for this task.
65 00:08:08.310 ⇒ 00:08:18.410 Demilade Agboola: But this is, for you know this dashboard, and I don’t think that’s the most important. This seems like the most important in our conversations with the, you know, C-suite, or conversations with whoever
66 00:08:18.650 ⇒ 00:08:24.800 Demilade Agboola: this has been, you know, one of things that’s come up a lot. And I feel like we should, you know. Look at that.
67 00:08:25.990 ⇒ 00:08:27.610 Demilade Agboola: Yeah. So
68 00:08:27.610 ⇒ 00:08:34.190 Amber Lin: I see. Yeah, that’s very important, because the the more we do the lower lower our rates are.
69 00:08:34.320 ⇒ 00:08:35.390 Amber Lin: So.
70 00:08:37.420 ⇒ 00:08:39.520 Amber Lin: That the
71 00:08:40.440 ⇒ 00:08:47.659 Amber Lin: yeah. So based on these. I think my question was that what would be the metrics for
72 00:08:47.880 ⇒ 00:08:52.489 Amber Lin: like these things? I know we already have turnaround time. So that’s a metric.
73 00:08:56.960 ⇒ 00:09:00.569 Amber Lin: I guess. How would we measure these
74 00:09:01.100 ⇒ 00:09:08.709 Amber Lin: like for all of these things be able to push back? What do you think is a metric that we can use to measure that
75 00:09:11.610 ⇒ 00:09:15.600 Demilade Agboola: like. I think of metrics. I’m not sure how we will measure the metrics, though that’s the real issue.
76 00:09:16.526 ⇒ 00:09:17.659 Amber Lin: So far.
77 00:09:18.030 ⇒ 00:09:21.220 Amber Lin: Let me ask Gpt, give me a second.
78 00:09:23.040 ⇒ 00:09:27.809 Amber Lin: Alright, we have these calls.
79 00:09:28.440 ⇒ 00:09:29.550 Amber Lin: Oh.
80 00:09:36.000 ⇒ 00:09:39.600 Amber Lin: okay, let’s see what they say.
81 00:09:42.940 ⇒ 00:09:45.876 Amber Lin: Huh? There we go.
82 00:09:48.720 ⇒ 00:09:53.790 Amber Lin: Python, build time end to end monitoring.
83 00:09:54.530 ⇒ 00:09:55.640 Amber Lin: Okay?
84 00:09:55.900 ⇒ 00:10:00.930 Amber Lin: Like, mean time to detect mean times. Result, like, yeah, cool.
85 00:10:01.660 ⇒ 00:10:10.059 Amber Lin: What do you guys think about this like what be what from? This is good, and what from this is
86 00:10:10.280 ⇒ 00:10:14.949 Amber Lin: is just words. Where is where is our document?
87 00:10:15.790 ⇒ 00:10:16.810 Amber Lin: Huh?
88 00:10:18.140 ⇒ 00:10:21.520 Amber Lin: Oh, here
89 00:10:26.070 ⇒ 00:10:27.990 Demilade Agboola: Do we need to do that now, or should we
90 00:10:29.610 ⇒ 00:10:37.880 Amber Lin: I mean, we can go back to it. But eventually the point is, if we want to improve this.
91 00:10:38.140 ⇒ 00:10:40.651 Amber Lin: or if you want to improve that.
92 00:10:41.130 ⇒ 00:10:45.949 Amber Lin: we’ll measure these different things. But I think we can come back to it.
93 00:10:46.480 ⇒ 00:10:48.889 Amber Lin: You you pose a good point.
94 00:10:49.590 ⇒ 00:10:56.800 Amber Lin: Yeah, let’s go and look at size indicators.
95 00:10:57.680 ⇒ 00:11:00.500 Amber Lin: Okay, yeah, those are essentially metrics.
96 00:11:00.630 ⇒ 00:11:04.909 Amber Lin: Number data time taken to build models.
97 00:11:05.140 ⇒ 00:11:09.749 Amber Lin: Okay, these are also the metrics. So I bet we’ll just revisit them soon.
98 00:11:10.060 ⇒ 00:11:19.039 Amber Lin: Let’s talk about the problems in workflows, because that’s where we’ll dictate what we do and how we do it.
99 00:11:19.510 ⇒ 00:11:20.310 Amber Lin: So
100 00:11:26.010 ⇒ 00:11:27.760 Amber Lin: while I’m talking over that
101 00:11:29.175 ⇒ 00:11:35.480 Demilade Agboola: So I think the things that take up a lot of time are like debugging and just like requests. So
102 00:11:35.480 ⇒ 00:11:36.160 Amber Lin: Oh!
103 00:11:36.820 ⇒ 00:11:41.020 Demilade Agboola: People who come and like if there’s an issue or something breaks
104 00:11:41.487 ⇒ 00:11:51.270 Demilade Agboola: that process of having to like go over and just kind of figure out where exactly is the break coming from, or where? What, exactly, is causing a bad
105 00:11:53.760 ⇒ 00:11:56.510 Demilade Agboola: a bad like data alpute
106 00:11:57.440 ⇒ 00:11:57.940 Amber Lin: Oh.
107 00:11:58.430 ⇒ 00:12:08.249 Demilade Agboola: That is something that you can, especially if bad details please, can take like sometimes bricks are easy to figure out in sense of like, oh, this isn’t working, Zendesk, or, you know.
108 00:12:08.610 ⇒ 00:12:10.950 Demilade Agboola: polytomic is not working when I ingest in the data
109 00:12:11.205 ⇒ 00:12:11.460 Amber Lin: Okay.
110 00:12:11.460 ⇒ 00:12:25.130 Demilade Agboola: But when it’s bad quality, sometimes it’s like literally a bad joint somewhere, or something like that. And you need to do education or something like that, and you need to go through lines of code to figure out the exact points in which things went bad.
111 00:12:26.860 ⇒ 00:12:27.550 Amber Lin: No.
112 00:12:28.100 ⇒ 00:12:32.100 Demilade Agboola: So that can also happen. Then you have ad hoc requests where?
113 00:12:35.160 ⇒ 00:12:38.650 Demilade Agboola: All of a sudden, someone needs some new data source
114 00:12:42.880 ⇒ 00:12:46.200 Demilade Agboola: or potentially. You know, they want a
115 00:12:46.330 ⇒ 00:12:49.230 Demilade Agboola: serious shifts to an existing data source
116 00:12:49.800 ⇒ 00:12:51.030 Amber Lin: Oh!
117 00:12:51.400 ⇒ 00:12:59.930 Demilade Agboola: So you have to make those changes, and especially if it’s ad hoc, and they need it like asap
118 00:13:00.370 ⇒ 00:13:05.109 Demilade Agboola: that can create some sort of pressure, to turn around.
119 00:13:09.500 ⇒ 00:13:12.860 Demilade Agboola: And so I think those were like the
120 00:13:13.640 ⇒ 00:13:16.840 Demilade Agboola: like. Why, like what I had in mind when I put it, put those things down
121 00:13:16.840 ⇒ 00:13:20.830 Amber Lin: I see so high pressure and short time is
122 00:13:21.500 ⇒ 00:13:27.420 Amber Lin: for all these ad hoc requests right? Or or is it just when they try to shift a new data source
123 00:13:28.446 ⇒ 00:13:34.579 Demilade Agboola: So high pressure. I mean to be fair like when consulting. I think everything’s kind of high pressure short time
124 00:13:34.580 ⇒ 00:13:35.710 Amber Lin: Commercial.
125 00:13:36.030 ⇒ 00:13:41.599 Demilade Agboola: Yeah, but I think usually when there are specific ad hoc requests
126 00:13:41.800 ⇒ 00:13:42.500 Amber Lin: Oh!
127 00:13:42.840 ⇒ 00:13:55.179 Demilade Agboola: Some specific request can even be a bit higher, higher pressure on shorter time, especially when you know the client is a bit antsy, or has been. You know, it’s been disappointed a couple of times. Yeah.
128 00:13:58.920 ⇒ 00:14:00.280 Amber Lin: Bye.
129 00:14:00.660 ⇒ 00:14:07.199 Amber Lin: See this, etc.
130 00:14:07.640 ⇒ 00:14:15.360 Demilade Agboola: Yeah. So we we’re trying to like, please them and make them sure they’re happy. And so in that point, we’re just like, Okay, come, how quickly can we get this? And
131 00:14:15.930 ⇒ 00:14:17.410 Demilade Agboola: yeah.
132 00:14:19.180 ⇒ 00:14:22.450 Amber Lin: I see, I mean ad hoc request is essentially when
133 00:14:22.600 ⇒ 00:14:31.149 Amber Lin: tickets come up inside mid cycle, because usually we try to plan them ahead. But then, when this ad hoc, they just come up mid cycle
134 00:14:31.530 ⇒ 00:14:32.090 Demilade Agboola: Yeah.
135 00:14:34.700 ⇒ 00:14:40.140 Demilade Agboola: And so like, the reason why also that can take time is just cause like it breaks your your flow
136 00:14:40.500 ⇒ 00:14:41.250 Amber Lin: Yeah.
137 00:14:41.400 ⇒ 00:14:57.729 Demilade Agboola: Okay with like this is what I’m I was trying to do. Maybe you’re in that flow, that flow state. And then, now you have to kind of maybe pick up another thing, and that because you’re not like mentally tuned into that. So it’s not. It doesn’t just does just it just doesn’t flow as easily sometimes.
138 00:14:57.910 ⇒ 00:14:58.900 Amber Lin: I see.
139 00:14:58.910 ⇒ 00:15:00.989 Demilade Agboola: We are not in that like headspace
140 00:15:11.300 ⇒ 00:15:13.360 Amber Lin: Okay, sounds good.
141 00:15:13.720 ⇒ 00:15:18.010 Amber Lin: And what is this one lack of clarity
142 00:15:18.420 ⇒ 00:15:20.210 Amber Lin: into what other eighties have done?
143 00:15:20.464 ⇒ 00:15:24.789 Demilade Agboola: So sometimes if you come in on a project like if you swap in on the project
144 00:15:24.790 ⇒ 00:15:25.320 Amber Lin: Oh!
145 00:15:25.650 ⇒ 00:15:34.390 Demilade Agboola: There isn’t necessarily like clarity, sense of like documentation, or like just like it.
146 00:15:34.950 ⇒ 00:15:37.839 Demilade Agboola: Clear definition of things
147 00:15:38.340 ⇒ 00:15:49.820 Demilade Agboola: work with and understand like, hey, this is what this person did. This is why that person did this. This is, you know why this the data only comes in, you know.
148 00:15:50.570 ⇒ 00:16:03.099 Demilade Agboola: once in whatever time period. You know, things like that, like just the understanding of the different sources, the understanding of different models. And just like why certain decisions were made
149 00:16:03.340 ⇒ 00:16:04.390 Demilade Agboola: on
150 00:16:04.590 ⇒ 00:16:11.569 Demilade Agboola: isn’t always clear, and the ability to be able to figure that out. Takes time. So you have to be able to sit down, go through
151 00:16:11.780 ⇒ 00:16:13.899 Demilade Agboola: multiple lines of code.
152 00:16:14.030 ⇒ 00:16:18.579 Demilade Agboola: figure out like, why, you know, did they put this here? Why did they choose that there?
153 00:16:21.330 ⇒ 00:16:27.289 Demilade Agboola: yeah. And just like sometimes figuring out why they are mitigating certain factors in their data. In certain ways
154 00:16:32.040 ⇒ 00:16:36.890 Demilade Agboola: we have a client urban stems, their own internal developer.
155 00:16:37.230 ⇒ 00:16:42.139 Demilade Agboola: She took over from someone that left that quit. I don’t know if he quit or he left, but basically
156 00:16:43.170 ⇒ 00:17:07.269 Demilade Agboola: she’s confused a lot of times. Sometimes she just goes. I’m not exactly sure, or she says, Oh, this source was switched to this source, and because of that we had to add a new like, it’s just it just becomes like you’re doing like you’re patching things together. And if you’re never, or if you come into a project like that, or if you come into a new project, and you’re not aware of every single thing or every single reason why people did certain like patch jobs
157 00:17:08.700 ⇒ 00:17:19.180 Demilade Agboola: You’re not aware of. You have to basically like, read and kind of figure out what exactly is going on, and maybe test things yourself to get to the point that you’re like. Oh, that makes sense. This is why they did that that way.
158 00:17:19.180 ⇒ 00:17:20.239 Amber Lin: No, no, no.
159 00:17:26.970 ⇒ 00:17:47.720 Amber Lin: yeah, that’s always a pain. We’re handing off our stack Blitz to their internal team, and it is taking a long time, and they’re also very confused, and we already have some documentation. But even with that they’re still confused. I can’t imagine where there’s 0 documentation like
160 00:17:48.400 ⇒ 00:17:49.110 Demilade Agboola: Yeah.
161 00:17:49.110 ⇒ 00:17:51.640 Amber Lin: Yeah, wow, okay.
162 00:17:52.080 ⇒ 00:17:56.440 Demilade Agboola: Any recurring manual or repetitive task
163 00:17:56.770 ⇒ 00:17:58.580 Amber Lin: Oh, tableau.
164 00:17:59.360 ⇒ 00:18:02.026 Demilade Agboola: So for me, cause like
165 00:18:03.790 ⇒ 00:18:07.949 Demilade Agboola: right now, especially on like Eden, the Eden project I
166 00:18:08.450 ⇒ 00:18:26.299 Demilade Agboola: I’m coming in with like like a bunch of express that the average au might not be handling this. But for me, like I, I’m also responsible for ensuring that dashboards are ready at the right time, like the the extracts, run all those kind of things like the actual processes.
167 00:18:27.101 ⇒ 00:18:29.910 Demilade Agboola: Right? And as a result.
168 00:18:30.410 ⇒ 00:18:41.597 Demilade Agboola: you know, that is something that is like manual and repetitive, like, I do it like every day, just to go through dashboards and show that the different data extracts run. So we have the latest data.
169 00:18:42.110 ⇒ 00:18:42.840 Demilade Agboola: yeah.
170 00:18:44.190 ⇒ 00:18:59.740 Amber Lin: Oh, so why don’t we go through? Because last time, when we talked, walk me through the whole spectrum of like data, essentially, why don’t we start from the very beginning to the very end? And let’s look at what parts in there might take the longest time
171 00:19:00.610 ⇒ 00:19:03.859 Demilade Agboola: Okay, so we can just say that
172 00:19:04.140 ⇒ 00:19:09.490 Demilade Agboola: high level, there are 3 main parts of data data.
173 00:19:10.293 ⇒ 00:19:11.989 Demilade Agboola: You have the ingestion.
174 00:19:12.650 ⇒ 00:19:14.830 Demilade Agboola: You have the transformation.
175 00:19:15.660 ⇒ 00:19:20.429 Demilade Agboola: And then you have, like, your bi transfer, your bi reporting layer.
176 00:19:26.810 ⇒ 00:19:36.260 Demilade Agboola: So ingestion is just basically like getting data in to the platform like into like your warehouse, and
177 00:19:38.056 ⇒ 00:19:41.249 Miguel de Veyra: Dimilade. Sorry I have a quick question.
178 00:19:41.763 ⇒ 00:19:44.449 Miguel de Veyra: This is not really that related, but
179 00:19:45.140 ⇒ 00:19:50.440 Miguel de Veyra: in your day to day work do you already use like chat, gpt, or maybe what was it? Cursor
180 00:19:51.440 ⇒ 00:19:56.189 Demilade Agboola: Yeah, I use cursor. I’m I’m on the team’s cursor, plan.
181 00:19:56.360 ⇒ 00:19:58.290 Demilade Agboola: And I also use chat. Gpt.
182 00:19:58.520 ⇒ 00:20:00.330 Miguel de Veyra: Okay, okay, okay, that makes sense.
183 00:20:03.740 ⇒ 00:20:12.529 Miguel de Veyra: Because, yeah, we want, because technically, we want to see. First, st if you know what AI tools are you already utilizing it’s core, not cursor, cursor.
184 00:20:13.764 ⇒ 00:20:16.679 Miguel de Veyra: not curse, you know it’s correct. See not
185 00:20:16.680 ⇒ 00:20:17.390 Amber Lin: Her
186 00:20:19.570 ⇒ 00:20:21.160 Miguel de Veyra: oh oh oh yes!
187 00:20:21.560 ⇒ 00:20:26.280 Amber Lin: Oh, every time I search in my terminal it doesn’t.
188 00:20:26.770 ⇒ 00:20:29.039 Amber Lin: Oh, cause I typed cursor
189 00:20:30.140 ⇒ 00:20:31.549 Miguel de Veyra: Yeah, okay, okay, that’s good.
190 00:20:31.870 ⇒ 00:20:35.960 Miguel de Veyra: One of the few things that Uta mentions is that he wants
191 00:20:36.160 ⇒ 00:20:45.080 Miguel de Veyra: basically the entire, especially on the data side, which I think everyone should be using, and I think everyone is. But everyone should be using cursor at least cursor right
192 00:20:45.080 ⇒ 00:20:47.573 Amber Lin: Yeah, I don’t get why you wouldn’t use cursor
193 00:20:49.010 ⇒ 00:20:55.230 Demilade Agboola: Definitely. My cursor does speed up a lot of things, especially when I’m writing things like documentation, which can be a pain in the ass.
194 00:20:55.490 ⇒ 00:20:56.930 Miguel de Veyra: Yeah, definitely.
195 00:20:56.930 ⇒ 00:21:05.449 Demilade Agboola: Just being able to have something that can look at the initial script. And as you’re typing out what you’re trying to explain, it just can help you auto fill it.
196 00:21:05.770 ⇒ 00:21:06.120 Miguel de Veyra: Yeah.
197 00:21:06.120 ⇒ 00:21:08.510 Demilade Agboola: But that makes it that makes it so much faster
198 00:21:08.950 ⇒ 00:21:10.620 Miguel de Veyra: Yeah, it just multiplies you.
199 00:21:10.930 ⇒ 00:21:14.026 Amber Lin: Yeah, okay, so,
200 00:21:14.690 ⇒ 00:21:15.650 Miguel de Veyra: Thank you for that
201 00:21:16.030 ⇒ 00:21:26.239 Amber Lin: I think that definitely ties together with all of these. So maybe like repetitive tasks, and maybe how you’re solving that. So we’ll go by. We’ll go step by step.
202 00:21:27.675 ⇒ 00:21:35.840 Demilade Agboola: So ingestion ideally should not have a repetitive task. You should automate it to be honest. So you should ideally set up the
203 00:21:36.441 ⇒ 00:21:43.139 Demilade Agboola: times in which you should ingest new data and all of that, and you should do it by itself. If you’re going in every time, then there’s something wrong.
204 00:21:45.190 ⇒ 00:21:51.680 Demilade Agboola: So ideally you shouldn’t have repetition there for transformation again.
205 00:21:52.610 ⇒ 00:21:53.880 Demilade Agboola: The
206 00:21:57.020 ⇒ 00:21:58.819 Demilade Agboola: the heavy lift will be
207 00:22:00.210 ⇒ 00:22:04.149 Demilade Agboola: the 1st time you’re building out the everything in there. So you’re building out the models.
208 00:22:05.640 ⇒ 00:22:08.629 Demilade Agboola: Sorry there’s a bit of a give me one second. I have a hiccup
209 00:22:09.130 ⇒ 00:22:10.000 Amber Lin: Okay.
210 00:22:19.370 ⇒ 00:22:20.729 Demilade Agboola: Okay. So I’m back.
211 00:22:21.200 ⇒ 00:22:21.750 Amber Lin: Yeah.
212 00:22:23.110 ⇒ 00:22:32.639 Demilade Agboola: So for the transformation layer. Again, a lot of time. The heavy lift is on building out the the models, the heavy like you build out everything infrastructure.
213 00:22:32.810 ⇒ 00:22:38.580 Demilade Agboola: And then subsequently, what you’re doing is about fixing
214 00:22:38.880 ⇒ 00:22:42.040 Demilade Agboola: or remodeling, based on new information.
215 00:22:43.548 ⇒ 00:22:49.980 Demilade Agboola: So that is probably the more manual part of like in terms of like in a day to day.
216 00:22:50.560 ⇒ 00:22:56.830 Demilade Agboola: In transformation layer. That’s probably the day to day aspect of it. So
217 00:22:57.090 ⇒ 00:23:03.460 Demilade Agboola: sometimes it doesn’t take a lot of time, sometimes takes a a huge chunk of time. So that’s it for
218 00:23:04.578 ⇒ 00:23:09.700 Demilade Agboola: and then for bi reporting kind of similar
219 00:23:10.165 ⇒ 00:23:15.870 Demilade Agboola: but it depends on what tool you use. So, for instance, in tableau, because of like the
220 00:23:16.570 ⇒ 00:23:20.879 Demilade Agboola: how like extracts are done, and all that you also need to just keep an eye on how to
221 00:23:21.606 ⇒ 00:23:27.320 Demilade Agboola: ensure that the extracts are working, everything is fine. Because if, again, if an extract doesn’t load
222 00:23:27.820 ⇒ 00:23:34.289 Demilade Agboola: all of a sudden, people don’t have data, and you know, World War 3 happens that sort of thing.
223 00:23:36.210 ⇒ 00:23:44.490 Demilade Agboola: then. But generally the heavy lift is in ensuring that you get the dashboard to the point that it’s usable for the end users.
224 00:23:44.610 ⇒ 00:23:51.399 Demilade Agboola: And at that point you’re just tweaking into adding new filters. Or you’re like, you know, changing the like
225 00:23:51.710 ⇒ 00:23:56.690 Demilade Agboola: change that in interactions or like things like that. Yeah, just doing like more
226 00:23:56.890 ⇒ 00:24:00.550 Demilade Agboola: like little fine tuning of the dashboard
227 00:24:12.070 ⇒ 00:24:20.670 Amber Lin: I see. Okay, it’s very helpful. And bottlenecks and
228 00:24:22.840 ⇒ 00:24:23.520 Demilade Agboola: Boom.
229 00:24:23.640 ⇒ 00:24:24.500 Amber Lin: Hmm.
230 00:24:24.750 ⇒ 00:24:31.829 Demilade Agboola: Well, bottlenecks. One huge bottleneck is basically like data access the amount of time it takes us to be able to access
231 00:24:31.960 ⇒ 00:24:36.280 Demilade Agboola: the client’s warehouse clients whatever clients dbt.
232 00:24:37.064 ⇒ 00:24:40.569 Demilade Agboola: that obviously delays us. So if
233 00:24:43.710 ⇒ 00:24:56.349 Demilade Agboola: if we don’t have access to like the warehouse, or we don’t have access to certain like things it takes. We like, we can’t really do much. We can have conversations. We can figure out what the problems are. Obviously, we can’t like access to data and build things.
234 00:24:57.730 ⇒ 00:25:11.539 Demilade Agboola: And sometimes it’s like little things like they might have a VPN and we need to send our Ips and get Whitelisted things like that like, so there are multiple. They’re just sometimes there are multiple layers to getting access to everything
235 00:25:18.930 ⇒ 00:25:21.949 Amber Lin: You know, when I’m hearing all these things in my mind is just
236 00:25:22.320 ⇒ 00:25:30.689 Amber Lin: going going through all these ideas of like, Okay, how can we? How can we make that easier? etc? So this is really helpful to hear from you.
237 00:25:31.722 ⇒ 00:25:40.750 Amber Lin: Other bottlenecks. You said business logic. So that’s like client documentation essentially
238 00:25:41.450 ⇒ 00:25:48.350 Demilade Agboola: Not not necessarily. Sometimes it is sometimes it in the
239 00:25:49.440 ⇒ 00:25:54.459 Demilade Agboola: as an ae. When you’re working, sometimes you just don’t have the properly defined
240 00:25:56.590 ⇒ 00:26:00.850 Demilade Agboola: logic, and that is because the in terms of
241 00:26:01.800 ⇒ 00:26:05.439 Demilade Agboola: I don’t see in terms of when
242 00:26:06.240 ⇒ 00:26:13.120 Demilade Agboola: The initial scoping conversations were done, or the initial conversations were done, certain metrics were not clearly gotten.
243 00:26:13.380 ⇒ 00:26:15.890 Demilade Agboola: For instance.
244 00:26:17.920 ⇒ 00:26:19.869 Demilade Agboola: What’s the definition of cogs?
245 00:26:20.260 ⇒ 00:26:30.820 Demilade Agboola: You know we haven’t a rough idea of what cogs is. Yes, but like in this particular use case, how do we know every single cost of goods, like everything that factors into the cost of goods sold
246 00:26:31.300 ⇒ 00:26:32.110 Demilade Agboola: right?
247 00:26:32.280 ⇒ 00:26:51.790 Demilade Agboola: So things like that like there are like those little things that potentially can be a bottleneck, because then, you know, you start. You started on the ground there. Then you finally reach, you know the point where you start doing the more complex, like transformations of like Oh, cogs or this or that. And then you realize that like this is not properly defined. Right, like
248 00:26:53.950 ⇒ 00:27:02.829 Demilade Agboola: So at that point you have to give feedback. Then they get back to you, or they have to get to the clients, get the clients, ideas, and sometimes clients isn’t like readily available.
249 00:27:03.000 ⇒ 00:27:08.590 Demilade Agboola: So when we finally get that information, that’s when you can continue and like, turn out whatever models you need to turn out
250 00:27:09.920 ⇒ 00:27:15.470 Amber Lin: I see. So by logic you mean more metric definition, right?
251 00:27:16.180 ⇒ 00:27:24.790 Demilade Agboola: Yes, yes, more of like the how, the how things flow in the in the client’s business, and how that leads to different metrics
252 00:27:24.790 ⇒ 00:27:34.129 Amber Lin: See? Yeah, I see that across both the data teams I’m managing. So with pool parts and with stack Blitz, they all have. We’re going back and forth on these
253 00:27:34.550 ⇒ 00:27:35.240 Demilade Agboola: Yeah.
254 00:27:35.740 ⇒ 00:27:36.690 Amber Lin: Totally.
255 00:27:37.070 ⇒ 00:27:45.060 Amber Lin: So last part of what to ask. Task asks, okay, I think we talked about
256 00:27:45.770 ⇒ 00:27:49.600 Amber Lin: talked about it a little bit. Oh, but also from the analyst.
257 00:27:52.970 ⇒ 00:27:55.579 Amber Lin: Can you tell me a little bit more about this
258 00:27:57.215 ⇒ 00:28:00.309 Demilade Agboola: So for that, it’s just basically
259 00:28:02.900 ⇒ 00:28:07.300 Demilade Agboola: you could be. And I kind of mentioned the. I kind of mentioned this to Tom yesterday because I had a 1 on one with him.
260 00:28:07.520 ⇒ 00:28:09.479 Demilade Agboola: or we’re just talking about how like
261 00:28:10.047 ⇒ 00:28:17.119 Demilade Agboola: sometimes, that’s right, because when specific from the analyst, sometimes you are in work, mode
262 00:28:18.860 ⇒ 00:28:25.180 Demilade Agboola: You know, you get a ping, that some data is not, you know, some data is
263 00:28:25.760 ⇒ 00:28:32.929 Demilade Agboola: not up to date, or something’s wrong with the data. And then you can check maybe bigquery or whatever data warehouse we’re using.
264 00:28:33.060 ⇒ 00:28:42.383 Demilade Agboola: And we can see that actually, it’s fine from our end. It’s something upstream or sorry downstream, which is the
265 00:28:43.310 ⇒ 00:28:45.400 Demilade Agboola: in maybe tableau, or whatever
266 00:28:45.400 ⇒ 00:28:46.040 Amber Lin: Hmm.
267 00:28:46.210 ⇒ 00:28:48.460 Demilade Agboola: Sometimes it’s a thing of like
268 00:28:48.870 ⇒ 00:28:55.680 Demilade Agboola: we’re put in firefighting mode without it necessarily being a thing that we can do, or we should be in
269 00:28:55.860 ⇒ 00:29:00.599 Demilade Agboola: so that that can drop the flow like it’s there’s nothing wrong per se.
270 00:29:00.850 ⇒ 00:29:09.320 Demilade Agboola: Maybe the person just needed to refresh the extract. Maybe the person like, there’s something they could have done within the environment and solve the problem.
271 00:29:09.450 ⇒ 00:29:12.200 Demilade Agboola: But because, like, they just run into that 1st hiccup
272 00:29:12.430 ⇒ 00:29:17.219 Demilade Agboola: and just go all analysts and the aes, let’s reach out to them, you know, that can disrupt.
273 00:29:18.210 ⇒ 00:29:21.999 Demilade Agboola: that can disrupt the the flow of the A
274 00:29:22.000 ⇒ 00:29:31.549 Amber Lin: I see so so essentially firefighting tasks that shouldn’t even be ours is disrupting the flow essentially
275 00:29:31.820 ⇒ 00:29:34.099 Demilade Agboola: Yeah, sometimes like a misdiagnosis
276 00:29:38.930 ⇒ 00:29:44.840 Demilade Agboola: Sometimes, you know, especially as the Ae, sometimes you’re it kind of feels like you’re responsible for like
277 00:29:45.270 ⇒ 00:29:48.429 Demilade Agboola: understanding, figuring out where the problem’s coming from.
278 00:29:48.670 ⇒ 00:29:53.320 Demilade Agboola: even if it might just be from the analysts. End of things like there may be.
279 00:29:53.730 ⇒ 00:29:59.499 Demilade Agboola: Sometimes it can be a thing of like you look through what you’ve done, and you’re like there’s nothing wrong here.
280 00:30:01.060 ⇒ 00:30:05.119 Demilade Agboola: and then they check on it. Oh, actually, it’s the engineers to do this instead
281 00:30:05.960 ⇒ 00:30:10.840 Amber Lin: I see. So it’s on
282 00:30:13.260 ⇒ 00:30:17.950 Amber Lin: so essentially like the analyst should take more responsibility.
283 00:30:19.490 ⇒ 00:30:20.380 Amber Lin: Yeah.
284 00:30:20.380 ⇒ 00:30:34.619 Demilade Agboola: But just like when they’re when they’re asking certain things, I think that there needs to be like a I’ll say checklist. But like you need to have gone through certain flows before you can go. Oh, this is definitely on the A.
285 00:30:52.360 ⇒ 00:30:53.950 Amber Lin: Me all of it.
286 00:30:54.810 ⇒ 00:30:56.139 Amber Lin: I see
287 00:30:56.290 ⇒ 00:31:10.919 Amber Lin: that’s good. That’s good. Because I to me, I did. I wasn’t that aware of where the line between the Ae. And analysts ends. So I think it will be helpful to define like, where does it? Where does it stop?
288 00:31:12.230 ⇒ 00:31:17.529 Demilade Agboola: Yeah, yeah, I think sometimes the line can be blurry.
289 00:31:18.131 ⇒ 00:31:34.579 Demilade Agboola: But ideally, the line is, is the data available in, is it the necessary data available in the warehouse. If it is, it’s usually on the, on the analyst to handle it all all the way towards like in production, in dashboard
290 00:31:34.700 ⇒ 00:31:38.150 Demilade Agboola: if it’s not like, if you know, it’s not fresh.
291 00:31:38.540 ⇒ 00:31:43.100 Demilade Agboola: If there’s questions about data quality whatever, that’s usually the Ae’s job
292 00:31:43.320 ⇒ 00:31:48.939 Amber Lin: Oh, did you say nested data table? Did I care? Right
293 00:31:48.940 ⇒ 00:31:49.530 Demilade Agboola: Happy.
294 00:31:49.640 ⇒ 00:31:50.940 Demilade Agboola: The needed data
295 00:31:52.160 ⇒ 00:31:52.900 Amber Lin: Oh!
296 00:31:54.350 ⇒ 00:31:55.739 Demilade Agboola: Is the needed like the data
297 00:31:55.740 ⇒ 00:31:56.710 Amber Lin: Oh.
298 00:32:14.700 ⇒ 00:32:20.659 Amber Lin: okay, right? Oh, goodness, okay. Current workflows.
299 00:32:22.240 ⇒ 00:32:27.600 Amber Lin: Let’s see, congestion following tableau.
300 00:32:30.330 ⇒ 00:32:33.909 Amber Lin: Okay, is that that will be everything right?
301 00:32:34.650 ⇒ 00:32:40.130 Demilade Agboola: I mean, obviously, there are still other stuff, you know, we use.
302 00:32:40.830 ⇒ 00:32:47.429 Demilade Agboola: So, for instance, visualization and some products we use Meta base like, I’m not in particular using Meta base. But like Meta Base does come up.
303 00:32:47.690 ⇒ 00:32:48.810 Demilade Agboola: Okay.
304 00:32:49.420 ⇒ 00:32:57.140 Demilade Agboola: The warehouse, you know, we also use like. So that’s like a different thing. So you think that you can put that between ingestion and modeling
305 00:32:59.070 ⇒ 00:33:07.180 Demilade Agboola: So, for, like the warehouse, we use snowflake we use
306 00:33:07.450 ⇒ 00:33:12.429 Demilade Agboola: see? Ideally, we’d like to use snowflake. But sometimes on products where we use redshift or bigquery
307 00:33:12.630 ⇒ 00:33:13.250 Amber Lin: Hmm!
308 00:33:15.760 ⇒ 00:33:18.679 Amber Lin: Is red shipping query the same thing, or just separate
309 00:33:19.002 ⇒ 00:33:24.480 Demilade Agboola: I mean, they’re same concepts. But they’re different providers. So red check is on bigquery is Google
310 00:33:30.600 ⇒ 00:33:34.000 Amber Lin: See? Ingestion.
311 00:33:35.600 ⇒ 00:33:39.910 Amber Lin: That’s everything for ingestion. Right? We will mostly use proatomic
312 00:33:41.243 ⇒ 00:33:49.649 Demilade Agboola: I think again, I like I more recently joined the team so potentially, there may be something else on another project that I have not seen.
313 00:33:49.850 ⇒ 00:33:53.909 Demilade Agboola: but most of what I’ve seen is, or sometimes segments
314 00:33:55.080 ⇒ 00:33:56.739 Amber Lin: All segments.
315 00:33:58.840 ⇒ 00:33:59.770 Amber Lin: Great?
316 00:34:02.743 ⇒ 00:34:05.469 Amber Lin: Clear documentation.
317 00:34:07.130 ⇒ 00:34:09.350 Amber Lin: Okay, it’s currently at work.
318 00:34:10.967 ⇒ 00:34:15.009 Amber Lin: What kind of workflows are we documenting?
319 00:34:16.030 ⇒ 00:34:16.709 Amber Lin: Do you know
320 00:34:16.719 ⇒ 00:34:21.759 Demilade Agboola: Oh, so we’re trying to document things about
321 00:34:25.139 ⇒ 00:34:30.879 Demilade Agboola: how we set up our clients. Warehouse.
322 00:34:31.039 ⇒ 00:34:38.000 Demilade Agboola: So we were creating a template, for instance, where or we have created a modifying it where?
323 00:34:38.999 ⇒ 00:34:41.819 Demilade Agboola: we, the setup of Dvt
324 00:34:41.959 ⇒ 00:34:46.849 Demilade Agboola: is kind of already there, so we can kind of just use it right off the bat.
325 00:34:47.049 ⇒ 00:35:01.719 Demilade Agboola: I’ve added alerts to it so and so we can always set up like the the run actions. So how often it runs and the alerts. So like they are basically anytime, we test bricks that we see it in our slack channel.
326 00:35:02.189 ⇒ 00:35:05.179 Demilade Agboola: So things like that just basically set up of Dvt
327 00:35:05.180 ⇒ 00:35:05.850 Amber Lin: Oh!
328 00:35:06.060 ⇒ 00:35:12.829 Demilade Agboola: Also creating like documentation on like flow workflow for things like Dbt tests.
329 00:35:13.555 ⇒ 00:35:22.330 Demilade Agboola: So that we know how we set up our test to ensure that at the end of the day we’re in in a good spot with our clients. Data, a lot of time.
330 00:35:24.730 ⇒ 00:35:36.169 Demilade Agboola: yeah, things like that. We’re just kind of just like ensuring that. You know the processes that we’re doing that are repeatable. We can get faster with it and ensure that we’re, you know.
331 00:35:36.530 ⇒ 00:35:38.099 Demilade Agboola: you know better with it.
332 00:35:43.160 ⇒ 00:35:47.760 Amber Lin: Biggest challenges in maintaining or scaling.
333 00:35:53.280 ⇒ 00:35:57.560 Demilade Agboola: I would say. Usually it’s documentation and their business transfer ownership
334 00:35:59.540 ⇒ 00:36:09.000 Demilade Agboola: Like. If if it requires you to be the only one who can solve the problem or who can quickly solve the problem, it’s it’s hard to scale that way.
335 00:36:09.290 ⇒ 00:36:10.190 Amber Lin: Hmm.
336 00:36:28.800 ⇒ 00:36:32.269 Amber Lin: let’s see. And different styles. Writing code kind of goes into the
337 00:36:32.690 ⇒ 00:36:37.689 Amber Lin: you can’t really transfer the ownership or help the other person understand
338 00:36:38.080 ⇒ 00:36:38.860 Demilade Agboola: Yeah.
339 00:36:39.880 ⇒ 00:36:40.550 Amber Lin: Let’s see.
340 00:36:40.980 ⇒ 00:36:48.570 Demilade Agboola: Well, well, different. Well, different sometimes is like, yeah, if someone else joins the team or joins like the projects
341 00:36:48.570 ⇒ 00:36:49.400 Amber Lin: No.
342 00:36:49.620 ⇒ 00:36:52.230 Demilade Agboola: It could be hard for them to like just kind of
343 00:36:52.390 ⇒ 00:36:56.550 Amber Lin: I will not only hard, because just it slows down the process of figuring out what you’re doing
344 00:36:57.546 ⇒ 00:36:59.180 Amber Lin: yeah, totally
345 00:37:01.641 ⇒ 00:37:10.800 Amber Lin: right, we’re on to. I think we talked about all of these. You know. We talked about these before friction misunderstandings with the analysts.
346 00:37:12.009 ⇒ 00:37:16.050 Amber Lin: track and files, and you stand up. That’s good.
347 00:37:16.520 ⇒ 00:37:22.569 Amber Lin: Okay, I’ll let Miguel take over for the
348 00:37:24.790 ⇒ 00:37:35.179 Amber Lin: AI part of what we can do and what are some desired outcomes. So I’ll let Miguel take over. You have around 20 min
349 00:37:37.474 ⇒ 00:37:41.199 Miguel de Veyra: Me just. I was looking at the ABC stuff
350 00:37:44.170 ⇒ 00:37:44.710 Amber Lin: Hmm.
351 00:37:45.550 ⇒ 00:37:47.469 Miguel de Veyra: Wait. Let me, can you guys hear me?
352 00:37:48.360 ⇒ 00:37:49.650 Miguel de Veyra: Yeah, go ahead. Okay.
353 00:37:49.960 ⇒ 00:37:57.749 Miguel de Veyra: yeah. Let me just think for a bit. Career generation on this on wait! Let me share my screen. Actually
354 00:38:02.390 ⇒ 00:38:07.090 Miguel de Veyra: on this stuff. Demi, ladi, are they? Are there things that
355 00:38:07.960 ⇒ 00:38:08.439 Amber Lin: Oh! The
356 00:38:08.440 ⇒ 00:38:09.070 Miguel de Veyra: Oh!
357 00:38:09.070 ⇒ 00:38:12.900 Amber Lin: What I created when I asked tragedy, so you can ignore them.
358 00:38:13.540 ⇒ 00:38:16.450 Miguel de Veyra: Yeah, yeah. But is there like something that you know
359 00:38:17.510 ⇒ 00:38:22.079 Miguel de Veyra: that you you find useful? So we could start with that we could start there
360 00:38:29.830 ⇒ 00:38:32.840 Miguel de Veyra: If if there’s none, it’s also fine, we can add a new one
361 00:38:37.580 ⇒ 00:38:43.109 Demilade Agboola: So I mean, query, generation and debugging is pretty useful. It’s kind of why we use cursor.
362 00:38:43.880 ⇒ 00:38:44.890 Demilade Agboola: Yeah.
363 00:38:45.280 ⇒ 00:38:52.920 Demilade Agboola: I mean, it’s not always accurate, obviously. But just that template sometimes just saves you the the time you need to start everything from scratch
364 00:38:58.030 ⇒ 00:38:59.180 Demilade Agboola: auto documentation
365 00:38:59.180 ⇒ 00:39:09.109 Miguel de Veyra: Challenge, for I guess the AI team is you’re already using cursor and chat Gpt for you to move like for us to develop something for you. It has to be something that those 2 doesn’t
366 00:39:09.580 ⇒ 00:39:11.149 Miguel de Veyra: already give you right
367 00:39:13.640 ⇒ 00:39:15.600 Demilade Agboola: Ideally. Yes.
368 00:39:15.600 ⇒ 00:39:16.190 Miguel de Veyra: Yes.
369 00:39:16.360 ⇒ 00:39:17.770 Demilade Agboola: I think that you know
370 00:39:18.550 ⇒ 00:39:21.730 Miguel de Veyra: Cause. Then there’s no point or just making your life complicated
371 00:39:21.730 ⇒ 00:39:22.730 Demilade Agboola: Exactly.
372 00:39:31.640 ⇒ 00:39:36.789 Amber Lin: Yeah, I suppose. Then, what is missing from cursor and chat? Gpt, I think we can start
373 00:39:36.790 ⇒ 00:39:37.980 Miguel de Veyra: Already knows.
374 00:39:41.370 ⇒ 00:39:42.240 Miguel de Veyra: So for
375 00:39:46.370 ⇒ 00:39:52.830 Miguel de Veyra: Code Base, was it? I think it was Yavi coffee, so wouldn’t that be helpful?
376 00:39:54.110 ⇒ 00:39:55.680 Demilade Agboola: Sorry. Can you repeat that? Please.
377 00:39:56.680 ⇒ 00:40:01.000 Miguel de Veyra: So we developed this bot for Javi coffee. One of our clients
378 00:40:01.290 ⇒ 00:40:01.760 Demilade Agboola: Exactly.
379 00:40:01.760 ⇒ 00:40:06.290 Miguel de Veyra: That bot has access or knows the entire code base.
380 00:40:07.281 ⇒ 00:40:11.340 Miguel de Veyra: Yeah, so would that be helpful for for you
381 00:40:15.860 ⇒ 00:40:21.320 Demilade Agboola: I mean, yeah, I can see it being useful in sense of like, if I
382 00:40:21.430 ⇒ 00:40:28.540 Demilade Agboola: was trying to figure out what model to look like if I’m like, where? Where do I start on this project? What model of you know
383 00:40:29.710 ⇒ 00:40:30.300 Miguel de Veyra: Okay.
384 00:40:30.300 ⇒ 00:40:35.690 Demilade Agboola: What model do I need to? Yeah, what model I need to focus on? If I have a an, A question
385 00:40:35.930 ⇒ 00:40:37.170 Miguel de Veyra: Yes, awesome.
386 00:40:37.670 ⇒ 00:40:46.909 Demilade Agboola: Or what is the source of something right, instead of like going through every model bit by bit, like just being able to like, ask that question, and it goes or go to this model first.st That would be helpful
387 00:40:47.360 ⇒ 00:40:47.950 Miguel de Veyra: See ya
388 00:40:48.710 ⇒ 00:40:52.670 Miguel de Veyra: Okay, yeah, I think that’s definitely something we can do.
389 00:40:53.524 ⇒ 00:40:58.000 Miguel de Veyra: So models to focus on source of data. Was that what you mentioned
390 00:40:58.280 ⇒ 00:41:05.049 Demilade Agboola: Yeah, like, what column? What column like? What is the source of this column? So like, where does all of this come from?
391 00:41:05.925 ⇒ 00:41:10.330 Demilade Agboola: There was a query. The lot like, ask questions of the logic, and it can trace
392 00:41:11.760 ⇒ 00:41:14.760 Demilade Agboola: the flow of detail like the flow of you know
393 00:41:15.190 ⇒ 00:41:16.610 Miguel de Veyra: Yeah, absolutely.
394 00:41:16.610 ⇒ 00:41:23.189 Amber Lin: This cursor. Not do that, since it has access to all your databases, I thought it would be able to do that
395 00:41:23.697 ⇒ 00:41:28.449 Miguel de Veyra: No, because you have. Basically, the way cursor worked is you have to actually add files
396 00:41:29.280 ⇒ 00:41:30.430 Amber Lin: Oh!
397 00:41:30.430 ⇒ 00:41:34.380 Miguel de Veyra: Yeah, like, basically, it doesn’t have the entire code
398 00:41:34.380 ⇒ 00:41:40.509 Amber Lin: I see. Have you tried to use Mcp? So if the Mcp has access to snowflake
399 00:41:40.790 ⇒ 00:41:48.845 Amber Lin: it, it should. If you give it access to all the different sources through Mcp. Then it should be
400 00:41:50.100 ⇒ 00:41:51.720 Miguel de Veyra: With Mcp.
401 00:41:52.520 ⇒ 00:41:54.820 Amber Lin: Mcp. Mcp.
402 00:41:55.040 ⇒ 00:41:59.410 Miguel de Veyra: And yeah, okay, okay, I have model control. I have this
403 00:41:59.410 ⇒ 00:42:04.650 Amber Lin: Yeah. So just search cursor, maybe just add cursor.
404 00:42:07.710 ⇒ 00:42:29.140 Amber Lin: It should be able to add different systems. So we probably would work with that to maybe give it access to our notion, and probably also, ideally, I think we could have it have access to our S 3, or maybe other things. So essentially, we just use this as a client interface
405 00:42:29.750 ⇒ 00:42:35.309 Amber Lin: when it has access to our all our files. If you want to scroll down, you can see that you can connect it to
406 00:42:35.460 ⇒ 00:42:37.799 Amber Lin: a lot of different things
407 00:42:38.640 ⇒ 00:42:39.200 Miguel de Veyra: Okay.
408 00:42:39.200 ⇒ 00:42:43.700 Amber Lin: I would recommend that we look into that. I think that would be something that the team can.
409 00:42:44.040 ⇒ 00:42:46.370 Amber Lin: It’ll be pretty fast to do
410 00:42:48.250 ⇒ 00:42:49.483 Miguel de Veyra: Okay, yes,
411 00:42:50.220 ⇒ 00:42:50.649 Miguel de Veyra: But
412 00:42:57.080 ⇒ 00:42:58.760 Miguel de Veyra: how do you add? Link? There we go.
413 00:42:59.430 ⇒ 00:43:00.200 Miguel de Veyra: Okay.
414 00:43:01.943 ⇒ 00:43:06.350 Miguel de Veyra: What about slack messages? And
415 00:43:06.860 ⇒ 00:43:14.659 Miguel de Veyra: was it, was it Zoom Meetings? How usual like, how in your day to day based? How much do you refer to it like
416 00:43:15.980 ⇒ 00:43:18.790 Miguel de Veyra: like how how do you say this?
417 00:43:20.170 ⇒ 00:43:20.660 Demilade Agboola: I mean
418 00:43:20.660 ⇒ 00:43:41.549 Miguel de Veyra: Like cause. Sometimes client has requests in slack, or, you know, says something there. So basically, we’re planning to build an agent that has access to that daily to the even emails, slack messages, Zoom Meetings. So, for example, you forget something you can just ask that agent, hey? What do I do again here, or what’s the best way to approach this
419 00:43:42.630 ⇒ 00:43:55.659 Demilade Agboola: Yeah, I think that could be helpful. I think one of the bigger needs is sometimes we get like a ton of things come through, and we lose track of things. So if, like a summary of like requests.
420 00:43:56.330 ⇒ 00:44:04.640 Demilade Agboola: so like one of the reasons why I like the slack it like the zoom summarizer is at the end of the zoom call like sometimes even calls. I just finished. I just go through and just
421 00:44:05.310 ⇒ 00:44:09.541 Demilade Agboola: look through and see. Oh, this was assigned to me, or this was said to be handled
422 00:44:10.210 ⇒ 00:44:15.729 Demilade Agboola: something like that, just the ability to be able to see. Oh, it’s all the requests that came in today, or you know.
423 00:44:15.940 ⇒ 00:44:16.680 Amber Lin: Yesterday.
424 00:44:16.680 ⇒ 00:44:27.569 Demilade Agboola: But you know, over whatever time period, so that allows us, because sometimes again, these things happen like a bunch of messages, and we lose track of these things, or some things take higher priority.
425 00:44:28.080 ⇒ 00:44:31.049 Demilade Agboola: So, being able to do a couple of things will be quite helpful
426 00:44:31.940 ⇒ 00:44:36.589 Miguel de Veyra: Oh, okay, cause, basically, the the thing that we’re planning to build is
427 00:44:36.720 ⇒ 00:44:41.679 Miguel de Veyra: it? It also has access to the slack zoom and also the linear tickets.
428 00:44:42.260 ⇒ 00:45:02.180 Miguel de Veyra: but I don’t think we’re gonna be creating tickets anytime soon. I don’t think that’s on the pipeline for now, but you know, I I think it would be very useful even for me. Right like, what if I just ask about hey? What should I prioritize today? Based on all the conversations we had yesterday, or like this past week, and then they’ll just give me a list of things
429 00:45:02.380 ⇒ 00:45:04.929 Miguel de Veyra: so I can, you know, start from somewhere
430 00:45:05.410 ⇒ 00:45:06.059 Demilade Agboola: Yeah. No.
431 00:45:06.060 ⇒ 00:45:10.209 Miguel de Veyra: Because I think majority of the problem from what I heard earlier is.
432 00:45:10.480 ⇒ 00:45:13.749 Miguel de Veyra: there’s too much to do and not too much time. Basically
433 00:45:14.620 ⇒ 00:45:15.590 Demilade Agboola: Yeah, definitely.
434 00:45:15.898 ⇒ 00:45:27.629 Miguel de Veyra: Which which I think we all are facing right now. But I think if we can, just, you know, if we have a bot that helps us pri prioritize or give us, gives us which are the priorities.
435 00:45:28.300 ⇒ 00:45:34.709 Miguel de Veyra: or at least you know, which has to be done then. Yes, I think that could be helpful. And yeah.
436 00:45:36.480 ⇒ 00:45:39.149 Miguel de Veyra: oh, wow, we need to connect notion
437 00:45:39.820 ⇒ 00:46:04.840 Amber Lin: Oh, yeah, I’m I’m just listing it out to of different options. I think a lot of 2 things you mentioned here is one with all the requests. Right? One is that things just get for just fall through like you forget them. And 2 you don’t. We tend to lose track of what’s the most important, because there’s just so much things to do. Is that correct?
438 00:46:06.530 ⇒ 00:46:13.980 Demilade Agboola: Yeah, I would say, so. Can we add linear? Because we’re using linear a lot more? I don’t know if it’s possible to just something
439 00:46:17.490 ⇒ 00:46:21.550 Demilade Agboola: as a source. I’m just saying, like, I know, if we’re looking at sources, I know linear also has.
440 00:46:21.850 ⇒ 00:46:23.919 Demilade Agboola: or we’ll have information
441 00:46:23.920 ⇒ 00:46:28.569 Miguel de Veyra: Yes, yes, we’re gonna be adding linear. But right now the priority, I believe, is.
442 00:46:29.650 ⇒ 00:46:41.389 Miguel de Veyra: I think it’s slack, Zoom. And then Github, those 3. Because linear is kinda we’re still starting out. So it doesn’t really contain as much information yet.
443 00:46:41.770 ⇒ 00:46:44.810 Miguel de Veyra: at least from the AI side of stuff. So
444 00:46:44.810 ⇒ 00:46:45.640 Demilade Agboola: Well.
445 00:46:45.640 ⇒ 00:46:47.859 Miguel de Veyra: We’re gonna be prioritizing the 1st 3
446 00:46:48.630 ⇒ 00:46:55.509 Amber Lin: Yeah. And to back on my point of, I think something that might be really helpful is
447 00:46:56.530 ⇒ 00:47:02.160 Amber Lin: when it has context to slack and zoom. The purpose of that is so that
448 00:47:02.180 ⇒ 00:47:26.670 Amber Lin: we know the business context right? So what we know, 1st of all, we know what’s going on right now. And second of all, we can always refer back to. Hey? Why do we even? Why are we even doing this? So once we create that agent, the client agent with all the context in the business, we’ll be able to know what is the priority. So to connect those together when you have a long list
449 00:47:26.670 ⇒ 00:47:34.299 Amber Lin: tasks coming in one. It keeps tracks of it, keeps track of it for you and 2. It will just automatically
450 00:47:34.540 ⇒ 00:47:56.390 Amber Lin: refer back to its context of, hey, this is the business outcome and rank it for you, and maybe it will output. Hey, these are all the requests that came in today. Here is how it relates to the business outcomes. And here is what the bot thinks. Thinks this is most important. Like, does that sound some sound like something that’s interesting to you?
451 00:47:58.140 ⇒ 00:48:06.320 Demilade Agboola: I I think that’s interesting. I also think just even even a a a
452 00:48:06.620 ⇒ 00:48:19.679 Demilade Agboola: quick summary of all the requests would be like tremendous, because just being able to start a new day and see all the requests so we can kind of see all this has been handled. This has been handled. Oh, shit! I forgot about that that
453 00:48:19.680 ⇒ 00:48:29.230 Amber Lin: So per person, you mean, because the Pm. Will ask everyone in the stand up right. But are you talking about having this personal bot system for per person.
454 00:48:29.410 ⇒ 00:48:32.855 Demilade Agboola: No, not no, no, not actually per person, but like
455 00:48:33.500 ⇒ 00:48:37.669 Demilade Agboola: So so imagine our chats with the client right
456 00:48:37.670 ⇒ 00:48:38.300 Amber Lin: Hmm.
457 00:48:38.600 ⇒ 00:48:43.339 Demilade Agboola: The client comes and says, Hey, I looked. I looked at this dashboard.
458 00:48:43.450 ⇒ 00:48:47.660 Demilade Agboola: I’m having issues with it in terms of this, this, this, this, this.
459 00:48:47.770 ⇒ 00:48:52.670 Demilade Agboola: another person comes and says, Hey, my issue with it is, I can’t access the dashboard.
460 00:48:52.830 ⇒ 00:48:54.620 Demilade Agboola: whatever whatever whatever whatever
461 00:48:55.780 ⇒ 00:49:00.760 Demilade Agboola: Something that comes at the end of the day, or, like, you know, the beginning of the new day, and says, Hey.
462 00:49:02.360 ⇒ 00:49:10.850 Demilade Agboola: Pressing one had a request to access the dashboard person. 2 had a request
463 00:49:11.010 ⇒ 00:49:17.319 Demilade Agboola: to change certain things in dashboard. Person 3. But like that sort of thing, even if it’s just that
464 00:49:17.730 ⇒ 00:49:22.980 Demilade Agboola: that ability to be able to see everything and go. Hey, we’ve knocked off one to. Actually, we forgot about this person
465 00:49:24.360 ⇒ 00:49:25.470 Demilade Agboola: could be helpful
466 00:49:25.880 ⇒ 00:49:27.060 Amber Lin: I see.
467 00:49:29.980 ⇒ 00:49:31.460 Demilade Agboola: But you get. Do you get what I mean?
468 00:49:31.460 ⇒ 00:49:41.960 Amber Lin: Yeah. So that’s sort of like an end of day check in. But instead of the Pm. Having to do it manually, the bot remembers everything, and the bot will ask
469 00:49:43.240 ⇒ 00:49:47.849 Demilade Agboola: Exactly so. It kind of just ensures everyone like no requests
470 00:49:48.130 ⇒ 00:49:54.819 Demilade Agboola: forgotten, or at least, if we’re not going to handle it, we can kind of say, like, this is low priority. We’re not, we’re not. We’re not bothering
471 00:49:54.820 ⇒ 00:49:55.200 Amber Lin: Hmm.
472 00:49:55.200 ⇒ 00:50:00.130 Demilade Agboola: But we we have like we are always aware of what’s going on, and we don’t lose track of things
473 00:50:04.910 ⇒ 00:50:10.399 Miguel de Veyra: And then like, for before you start the next day, it kind of sends you a notification at around 9 Am.
474 00:50:10.510 ⇒ 00:50:13.879 Miguel de Veyra: Just, for example, that, hey? These are the stuff you need to work on
475 00:50:14.750 ⇒ 00:50:26.189 Demilade Agboola: Yeah, yeah, it might only be helpful if there could be feedback to the bot. So if you say, Oh, we’ve done numbers 1, 2, 4, and 5. So the next day only sends number 3, for instance.
476 00:50:26.190 ⇒ 00:50:34.879 Amber Lin: Essentially like you reply to the bot on your progress, so that eventually that could take over our daily stand up because you’ll say what you already worked on
477 00:50:37.030 ⇒ 00:50:44.570 Demilade Agboola: Potentially it would work in in in production, and how people interact with it.
478 00:50:44.750 ⇒ 00:50:46.219 Amber Lin: Yeah, I see.
479 00:50:46.660 ⇒ 00:50:52.580 Demilade Agboola: Because the thing about sometimes some requests come in ad hoc. Cause, like slack is way of ad hoc. Requests come in not necessarily
480 00:50:53.300 ⇒ 00:50:56.080 Demilade Agboola: like the business goal requests.
481 00:50:56.210 ⇒ 00:51:01.559 Demilade Agboola: because the high level request that don’t necessarily come in through slack. They’re usually in the Zoom Meetings, or like whatever call
482 00:51:01.560 ⇒ 00:51:01.890 Amber Lin: I’m sure
483 00:51:02.460 ⇒ 00:51:03.150 Demilade Agboola: Yeah.
484 00:51:17.550 ⇒ 00:51:23.670 Amber Lin: Let’s see, that was really helpful to know what kind of different messages, like different requests that will come through these things.
485 00:51:25.830 ⇒ 00:51:33.572 Amber Lin: Hmm! I know that. Oh, we have 8 min. I know that another issue that we brought up is
486 00:51:35.360 ⇒ 00:51:38.289 Amber Lin: like being able to push back or
487 00:51:38.530 ⇒ 00:51:45.480 Amber Lin: based on the business outcome right being aware of the business outcome. Is there something that we can help with that
488 00:51:49.760 ⇒ 00:51:50.910 Miguel de Veyra: I don’t think so.
489 00:51:51.570 ⇒ 00:51:52.030 Demilade Agboola: I think
490 00:51:52.030 ⇒ 00:51:58.399 Miguel de Veyra: Where we can help is in one of the problems here. He mentioned.
491 00:51:59.240 ⇒ 00:52:00.420 Miguel de Veyra: Wait, where is it?
492 00:52:01.370 ⇒ 00:52:13.939 Miguel de Veyra: Where is it? Sorry? Bad quality, most time consuming. So before we even get here, can we have some sort of like a quality assurance agent that basically analyzes, you know
493 00:52:14.450 ⇒ 00:52:18.999 Miguel de Veyra: the code that you did or go that someone did. And then just say, Hey.
494 00:52:19.720 ⇒ 00:52:27.989 Miguel de Veyra: this is a problem. This will cause problems, or this is not up to industry standard, or I don’t know how we want to. Qa it. But what do you think
495 00:52:28.940 ⇒ 00:52:31.440 Demilade Agboola: I think a Qa. Agent will be helpful.
496 00:52:33.230 ⇒ 00:52:42.319 Demilade Agboola: I will see that there are some things that potentially you might need more access than just the code base in the sense of. For instance, if you’re doing a join
497 00:52:43.660 ⇒ 00:52:56.100 Demilade Agboola: and you need to join on area code as well as as well as deeds.
498 00:52:56.330 ⇒ 00:52:57.230 Demilade Agboola: Right?
499 00:52:57.540 ⇒ 00:52:59.580 Demilade Agboola: If you only join on dates.
500 00:52:59.740 ⇒ 00:53:05.559 Demilade Agboola: then you would have duplicates up here because you would associate multiple area codes to the same day.
501 00:53:05.710 ⇒ 00:53:14.859 Demilade Agboola: If you only join on area code, it would associate the same area code. But on multiple days. So you kind of only want to lock it into the same like rows.
502 00:53:15.850 ⇒ 00:53:17.719 Demilade Agboola: Sometimes something like that.
503 00:53:18.170 ⇒ 00:53:27.859 Demilade Agboola: It might not have the necessary context that, like, hey, actually, in the database, there’s you need to. There’s like multiple ways to join
504 00:53:28.150 ⇒ 00:53:30.340 Demilade Agboola: this table and the proper
505 00:53:30.340 ⇒ 00:53:37.799 Miguel de Veyra: Technically, it should have it should have access to the code base and the database basically
506 00:53:38.370 ⇒ 00:53:52.020 Demilade Agboola: Yeah, if you want to get like full contact. So if you want to get like, so he knows that this join you’re doing would only is a 1 to one, or it’s a 1 to many, for instance, and you can say, Hey, do you want to get a 1 to many outputs rather than a 1 to one output
507 00:53:53.040 ⇒ 00:54:03.099 Miguel de Veyra: Okay, yeah, yeah, okay, I understand. I think this is definitely something we can do. Like, I don’t think we need to access the entire database, just some part of it. Just to give it context right?
508 00:54:03.370 ⇒ 00:54:03.970 Demilade Agboola: Yeah.
509 00:54:04.670 ⇒ 00:54:12.670 Miguel de Veyra: Okay, then I think lack of clarity into what other A’s have done is a bit more on the
510 00:54:13.580 ⇒ 00:54:15.610 Miguel de Veyra: documentation side.
511 00:54:17.140 ⇒ 00:54:22.630 Miguel de Veyra: If so, if we have a bot that can that have access to the documentation and the Github that should help. There
512 00:54:22.630 ⇒ 00:54:24.310 Amber Lin: Hmm! I see
513 00:54:26.750 ⇒ 00:54:32.529 Miguel de Veyra: This is more of like a side effect of this. The cure that it’ll be, you know, fixed eventually
514 00:54:33.140 ⇒ 00:54:33.830 Demilade Agboola: Yeah.
515 00:54:34.550 ⇒ 00:54:40.410 Miguel de Veyra: And then for this one, should we automate at the ingestion? How do we do this?
516 00:54:41.000 ⇒ 00:54:41.990 Miguel de Veyra: Currently
517 00:54:44.570 ⇒ 00:54:45.510 Demilade Agboola: So I guess it’s
518 00:54:45.510 ⇒ 00:54:47.050 Miguel de Veyra: Different for each client. Right?
519 00:54:48.900 ⇒ 00:54:55.610 Demilade Agboola: Yeah, it’s different for each client and each tool. But usually it’s a it’s a function of, you know. You need to get the Api keys from the source.
520 00:54:55.840 ⇒ 00:55:00.740 Demilade Agboola: You need to put it into the tool you need to connect to the source.
521 00:55:00.940 ⇒ 00:55:16.260 Demilade Agboola: and then you set up. You know you pick the tables. You want to like the tables, the columns, everything you want to take from the source. And then you put a date like, okay, you put a date. Sorry you put a time, not date. You put a time. So at 6 am. Utc, you know.
522 00:55:16.590 ⇒ 00:55:22.729 Demilade Agboola: get this data from this table so automatically. It does that every single morning.
523 00:55:22.860 ⇒ 00:55:26.830 Demilade Agboola: or you know, middle of the night, or whatever you’ve set it to be. Maybe one am
524 00:55:26.830 ⇒ 00:55:27.430 Miguel de Veyra: Yep.
525 00:55:28.019 ⇒ 00:55:33.170 Demilade Agboola: And then business has data available at 6 am. Or 7 Am.
526 00:55:34.535 ⇒ 00:55:37.680 Miguel de Veyra: And right now, I believe you guys do this manually right?
527 00:55:38.260 ⇒ 00:55:45.350 Demilade Agboola: The setting up the initial setup. Yes, manually, but the daily, like the the daily ingestion, is automatic.
528 00:55:46.341 ⇒ 00:55:50.430 Miguel de Veyra: So I. So technically, this is not really a problem. Since
529 00:55:51.010 ⇒ 00:55:51.430 Demilade Agboola: Oh, yeah.
530 00:55:51.772 ⇒ 00:55:52.800 Miguel de Veyra: Yeah, of course.
531 00:55:52.800 ⇒ 00:55:56.960 Demilade Agboola: Once it’s done, it’s done. It’s only if there’s an that’s what that’s why I said, if there’s an issue.
532 00:55:57.400 ⇒ 00:55:59.290 Demilade Agboola: yeah, you have to start
533 00:55:59.290 ⇒ 00:56:00.640 Miguel de Veyra: Then it becomes a problem
534 00:56:01.060 ⇒ 00:56:01.910 Demilade Agboola: It’s a problem. Yeah.
535 00:56:02.180 ⇒ 00:56:08.150 Miguel de Veyra: Okay, okay, I think that’s pretty much it.
536 00:56:08.870 ⇒ 00:56:16.360 Miguel de Veyra: I think most of the things that we can help is more on either the Qa. Or the documentation side of things
537 00:56:17.670 ⇒ 00:56:18.460 Demilade Agboola: Yeah.
538 00:56:19.940 ⇒ 00:56:24.029 Miguel de Veyra: Basically giving you more context and to what’s happening.
539 00:56:25.190 ⇒ 00:56:32.390 Miguel de Veyra: Cause that was also one of, I guess my worries is, if a person is, you know, using already the AI stuff.
540 00:56:33.330 ⇒ 00:56:36.370 Miguel de Veyra: I think us forcing them to move into
541 00:56:37.150 ⇒ 00:56:45.739 Miguel de Veyra: what we built or even us building something that’s already there. There, let’s say, 80% using 80% effective. I I would say we should build for that
542 00:56:46.240 ⇒ 00:57:14.849 Amber Lin: Yeah. And I think part of the AI team’s goal here is also to see how we can best use the AI tools. Right? Cause. I just like the Mcps. There’s a lot of things that these AI tools will be able to do, and maybe we can have like workshops and download. You have to lead them of how you should best use Chatgpt. Or maybe there’s certain templates for different requests like prompt engineering, or, say, getting the cursor to be more effective. That’s definitely something we want to do.
543 00:57:14.910 ⇒ 00:57:26.970 Amber Lin: And in terms of creating agents from scratch. What is your priority like? What will make the most difference, because we only have so much time. And we want the biggest impact for your team.
544 00:57:27.350 ⇒ 00:57:27.980 Miguel de Veyra: Yeah.
545 00:57:29.850 ⇒ 00:57:30.300 Demilade Agboola: I guess
546 00:57:30.300 ⇒ 00:57:32.269 Miguel de Veyra: Question. Now, Damalade is, which
547 00:57:32.490 ⇒ 00:57:36.970 Miguel de Veyra: client do you want us to work on? Because that’s what basically who they’ve asked us to ask you
548 00:57:38.640 ⇒ 00:57:41.640 Demilade Agboola: Which client has the needs, the most help.
549 00:57:44.014 ⇒ 00:57:50.329 Demilade Agboola: With what? Exactly. That’s the real question, like, what are we working on amongst AI tools?
550 00:57:51.346 ⇒ 00:57:53.010 Miguel de Veyra: We’re gonna build, basically
551 00:57:53.632 ⇒ 00:58:07.379 Miguel de Veyra: an agent that has access to all the clients, Zoom Meetings, slack messages less the slack channels, basically. And then eventually the Github like, which client would you do you think will be most will most benefit from that
552 00:58:08.869 ⇒ 00:58:14.130 Demilade Agboola: I would say urban stems, because their their github, their data and their github is is
553 00:58:14.370 ⇒ 00:58:16.380 Demilade Agboola: pardon my French, it’s a clusterfuck
554 00:58:16.910 ⇒ 00:58:17.330 Miguel de Veyra: It’s time
555 00:58:20.780 ⇒ 00:58:26.150 Demilade Agboola: But you might need to talk to them about that in terms of like how long we we have with them.
556 00:58:26.720 ⇒ 00:58:31.140 Demilade Agboola: And what like the contract length is, because obviously you don’t want to invest some more time with something that is not
557 00:58:31.140 ⇒ 00:58:32.440 Miguel de Veyra: Oh, yeah. True.
558 00:58:32.800 ⇒ 00:58:33.710 Demilade Agboola: Yeah, so.
559 00:58:34.250 ⇒ 00:58:35.539 Miguel de Veyra: Not there no worries cause
560 00:58:35.540 ⇒ 00:58:37.620 Amber Lin: What other ones? Then?
561 00:58:37.620 ⇒ 00:58:38.900 Miguel de Veyra: Yeah. The second one.
562 00:58:39.750 ⇒ 00:58:47.139 Demilade Agboola: I mean, I’m only currently on 2 projects. So I can only compare those 2 you know best, someone, Eden. And I’m on urban stamps
563 00:58:47.890 ⇒ 00:58:50.960 Amber Lin: Okay, I think that’s a question we can ask.
564 00:58:51.345 ⇒ 00:58:51.730 Miguel de Veyra: Yeah.
565 00:58:51.730 ⇒ 00:58:52.080 Amber Lin: Utah.
566 00:58:52.080 ⇒ 00:58:52.490 Demilade Agboola: Yeah.
567 00:58:52.808 ⇒ 00:59:10.979 Amber Lin: I think back to my question of what area would help the most like, regardless of what we’re building right now. I just want to hear from you, because ultimately, whatever we’re building is for you. So just you don’t have to just disregard what we’re working on right now. What will help you? The most
568 00:59:13.543 ⇒ 00:59:14.790 Demilade Agboola: I think
569 00:59:15.600 ⇒ 00:59:22.560 Demilade Agboola: what will help us the most will be the ability to just get context of everything that’s happening in like the database at once will be very.
570 00:59:22.560 ⇒ 00:59:23.430 Amber Lin: Very helpful.
571 00:59:24.272 ⇒ 00:59:31.359 Demilade Agboola: Or the ability to create quickly create like documentation based off what existed in the
572 00:59:33.070 ⇒ 00:59:35.363 Demilade Agboola: And when I mean documentation, I mean, like
573 00:59:36.120 ⇒ 00:59:48.149 Demilade Agboola: either, like reports. So like, if I want to be able to get like the context out or like Dbt documentation. I’m not sure if you’re aware. But Dbt, create like yaml files that you can use for documentation.
574 00:59:48.260 ⇒ 00:59:56.269 Demilade Agboola: But that ability to be able to, you know. Go to that stage quickly, because part of the reasons why we don’t do documentation for clients, and it’s something I’ve been talking to Tom about.
575 00:59:56.420 ⇒ 00:59:57.310 Amber Lin: Hmm.
576 00:59:58.290 ⇒ 01:00:17.240 Demilade Agboola: Ideally, we want to do documentation on clients. We want to be able to put those yaml files in and say, Hey, this are the files. These are the different columns. This is what this column name means. This is a primary key, and we put up tests for primary key. So it must be unique, and there should be no null values in there. Things like that.
577 01:00:17.740 ⇒ 01:00:20.540 Demilade Agboola: But we don’t. The transition takes so much time.
578 01:00:21.830 ⇒ 01:00:24.069 Demilade Agboola: and we want to be as fast as possible.
579 01:00:24.410 ⇒ 01:00:26.209 Demilade Agboola: So what happened
580 01:00:26.556 ⇒ 01:00:27.250 Amber Lin: Go ahead!
581 01:00:27.250 ⇒ 01:00:33.990 Demilade Agboola: What ends up happening is that there are times when bad data or certain things go through.
582 01:00:34.910 ⇒ 01:00:37.379 Demilade Agboola: but we don’t have to test to flag them.
583 01:00:37.510 ⇒ 01:00:39.480 Demilade Agboola: and then we only notice it.
584 01:00:40.790 ⇒ 01:00:49.470 Demilade Agboola: So, for instance, I was looking at a dashboard doing when I was at like just ensuring that things were fine, and I noticed that we hadn’t had dash data in that dashboard for
585 01:00:50.200 ⇒ 01:00:51.999 Demilade Agboola: 10 days or something, and I was like.
586 01:00:52.390 ⇒ 01:00:58.790 Demilade Agboola: what’s going on here. And apparently we had turned off the ingestion thing because we’re trying to settle like
587 01:00:58.900 ⇒ 01:01:02.560 Demilade Agboola: the output was ingested with the client and all that stuff.
588 01:01:03.000 ⇒ 01:01:14.529 Demilade Agboola: But it wasn’t. It wasn’t necessarily something like I was aware of. That’s 1 and 2. It was just something that you know, potentially it wasn’t flagged immediately. It wasn’t anything that like.
589 01:01:14.920 ⇒ 01:01:31.760 Demilade Agboola: If we had tested like we had a test that the latest created should not be more than 24 h from that current day, we would automatically the moment it had gone over 24 h, would immediately know that something’s wrong with our data, and that allows us to be able to go, or the entire team is aware. So just being able to
590 01:01:32.000 ⇒ 01:01:41.039 Demilade Agboola: like constantly be aware of everything like that is happening because there’s a limit to how much you can keep track of everything that’s going on in your in the code base by, as as one person
591 01:01:42.000 ⇒ 01:01:54.989 Amber Lin: I see. So is that the same of get contacts of everything that’s happening in the database? Or is this more of a alert immediately, or alert errors or alert actions? Are they different
592 01:01:55.830 ⇒ 01:02:00.130 Demilade Agboola: Think they’re similar. I think you have enough context. You can then create the alerts necessary
593 01:02:01.020 ⇒ 01:02:19.699 Amber Lin: I see. So I think what I hear is that your priorities, 1st of all, getting contrast with everything that happens in the database which that will enable us to one send alerts and tests. So we don’t end up 10 days without data and 2 to quickly create documentation and reports
594 01:02:19.930 ⇒ 01:02:22.139 Amber Lin: so that we can send that to clients
595 01:02:22.840 ⇒ 01:02:23.610 Demilade Agboola: Yeah.
596 01:02:23.610 ⇒ 01:02:37.930 Amber Lin: I see cause. I want. Essentially, I want our engineers and data analysts to focus their time on technical expertise and not spend so much time on like documentation tech depths. So essentially, that’s what we want to be helping with.
597 01:02:38.460 ⇒ 01:02:39.020 Amber Lin: Okay.
598 01:02:39.020 ⇒ 01:02:39.560 Demilade Agboola: Yeah.
599 01:02:40.180 ⇒ 01:03:00.559 Amber Lin: Great that is, that is really helpful, and me and Miguel, and then we’ll talk over like the different different client teams. How we want to prioritize these things, but thank you so much for letting us know what’s most important. So we don’t end up also, as you mentioned, working on the unimportant dashboards before, and then having to work on the important things again.
600 01:03:00.890 ⇒ 01:03:02.249 Demilade Agboola: Yeah, that’s fair. That’s fair.
601 01:03:02.250 ⇒ 01:03:05.929 Amber Lin: Okay, thank you so much for your time. I need to jump to another meeting
602 01:03:06.050 ⇒ 01:03:07.189 Demilade Agboola: Alright! Thank you. Bye.
603 01:03:07.190 ⇒ 01:03:08.360 Amber Lin: Okay. Bye-bye.