Meeting Title: Brainforge x Billy Cheng Catch-up Date: 2026-03-06 Meeting participants: Robert Tseng, Billy Cheng
WEBVTT
1 00:01:30.590 ⇒ 00:01:32.150 Billy Cheng: Hello!
2 00:01:33.670 ⇒ 00:01:34.810 Robert Tseng: Hey, Billy!
3 00:01:35.200 ⇒ 00:01:36.769 Billy Cheng: How are things?
4 00:01:38.460 ⇒ 00:01:40.190 Robert Tseng: Good.
5 00:01:40.480 ⇒ 00:01:46.360 Robert Tseng: I’m, in… Dallas for a friend’s wedding.
6 00:01:46.410 ⇒ 00:01:47.739 Billy Cheng: Oh my wife.
7 00:01:48.070 ⇒ 00:01:48.610 Robert Tseng: Yeah, I’m not…
8 00:01:48.610 ⇒ 00:01:51.549 Billy Cheng: This one was filed earlier today.
9 00:01:52.470 ⇒ 00:02:03.489 Robert Tseng: Yeah, you know, we just… I just got in early this morning, working a bit, and then have, like, some rooms and duties later on, and then, I’m actually taking next week off, so… so I’ve kind of been…
10 00:02:03.490 ⇒ 00:02:04.420 Billy Cheng: floor… what?
11 00:02:04.420 ⇒ 00:02:07.079 Robert Tseng: Getting ready for vacation mode, but…
12 00:02:07.080 ⇒ 00:02:16.939 Billy Cheng: Nice, cool, cool. Sounds good. You haven’t changed much, you still look the same. Well, in a good way, I don’t know if you want to look the same, but…
13 00:02:19.460 ⇒ 00:02:20.120 Billy Cheng: Yes.
14 00:02:20.120 ⇒ 00:02:24.769 Robert Tseng: I think I’m… I’m not as dark as I used to be, that’s really probably what it is.
15 00:02:25.110 ⇒ 00:02:25.820 Billy Cheng: Don’t get as much time.
16 00:02:25.820 ⇒ 00:02:28.870 Robert Tseng: sunlight anymore because of, I don’t know, New York.
17 00:02:28.870 ⇒ 00:02:29.479 Billy Cheng: There we go.
18 00:02:29.480 ⇒ 00:02:30.700 Robert Tseng: that much time window.
19 00:02:30.700 ⇒ 00:02:32.549 Billy Cheng: When did you move to New York?
20 00:02:33.090 ⇒ 00:02:38.299 Robert Tseng: We moved there, it’s been 2 years, yeah.
21 00:02:38.720 ⇒ 00:02:55.100 Robert Tseng: 2 years ago, so I got married end of 2023, and then… first we went to Asia, kind of visiting family, both in Hong Kong and Taiwan, and then we moved to New York. We first stayed in Jersey City.
22 00:02:55.160 ⇒ 00:03:12.510 Robert Tseng: And then, a year ago, we moved into Manhattan, so… I don’t know if you’ve been to New York, but we live by Central Park, kind of in the southwest corner. It’s called Columbus Circle. Yeah, so we’ve been there for a year now, and we signed a two-year lease, so we’ll be in the same spot for at least another year.
23 00:03:14.360 ⇒ 00:03:20.010 Robert Tseng: Yeah, and then, I hear… I think I mentioned that. We might… we might move to Asia.
24 00:03:20.210 ⇒ 00:03:25.630 Billy Cheng: Yeah, that’s interesting. But yeah, it sounds like you guys,
25 00:03:26.030 ⇒ 00:03:31.910 Billy Cheng: I mean, you’re essentially living in the middle of New York, like, Central Park, I’ve only.
26 00:03:31.910 ⇒ 00:03:32.320 Robert Tseng: Yeah.
27 00:03:32.320 ⇒ 00:03:33.599 Billy Cheng: Yeah, poor.
28 00:03:33.730 ⇒ 00:03:39.100 Billy Cheng: for people’s wedding, or for STB? Is it STB? Yeah.
29 00:03:39.870 ⇒ 00:03:41.419 Robert Tseng: Do you have an SCP in New York?
30 00:03:41.910 ⇒ 00:03:45.569 Billy Cheng: It was in… it was East Coast STP, so we…
31 00:03:45.570 ⇒ 00:03:46.159 Robert Tseng: Right, right, right.
32 00:03:46.160 ⇒ 00:03:52.369 Billy Cheng: New York City, but we’re just in, what’s that place? Harvey Cedars? I don’t know if they’re…
33 00:03:52.680 ⇒ 00:03:53.640 Robert Tseng: Yeah, yeah.
34 00:03:53.940 ⇒ 00:03:56.820 Billy Cheng: Have you been to Javisida? So if I see it?
35 00:03:56.820 ⇒ 00:04:04.180 Robert Tseng: I haven’t been with ICA, but I know that, like, well, they had Rutgers ICA, they would go there, yeah.
36 00:04:04.580 ⇒ 00:04:10.000 Billy Cheng: Yeah, that’s where we went. That was ages ago. That was…
37 00:04:10.000 ⇒ 00:04:11.790 Robert Tseng: Well, you, you look…
38 00:04:11.790 ⇒ 00:04:14.030 Billy Cheng: You still… you are still alive.
39 00:04:14.030 ⇒ 00:04:14.600 Robert Tseng: Looking…
40 00:04:14.600 ⇒ 00:04:18.739 Billy Cheng: Oh my god, that’s 20 years… it was 20 years ago, it was 2006!
41 00:04:19.029 ⇒ 00:04:19.409 Robert Tseng: Wow.
42 00:04:20.290 ⇒ 00:04:21.240 Billy Cheng: Yeah.
43 00:04:21.850 ⇒ 00:04:26.350 Billy Cheng: And I mean, I was in New York probably a few years ago, before COVID.
44 00:04:26.700 ⇒ 00:04:27.360 Robert Tseng: Huh.
45 00:04:27.630 ⇒ 00:04:34.349 Billy Cheng: Yeah, for someone’s wedding, all that stuff. But yeah, I don’t know… to answer your question, I don’t really know New York City that well.
46 00:04:35.900 ⇒ 00:04:41.390 Robert Tseng: Yeah, well, I guess you moved to the UK, like, a couple years ago, you said?
47 00:04:41.760 ⇒ 00:04:46.849 Billy Cheng: Yeah, end of 23, that’s when I moved.
48 00:04:47.250 ⇒ 00:04:57.729 Billy Cheng: I mean, it’s just because, like, Hong Kong people can work in the UK, and I felt like I was in the same company for forever. Yeah. So I felt like, oh, I might as well just give it a try, and, like.
49 00:04:57.730 ⇒ 00:04:58.380 Robert Tseng: Yeah.
50 00:04:59.180 ⇒ 00:05:01.770 Billy Cheng: see how it goes, but then I guess…
51 00:05:01.990 ⇒ 00:05:18.410 Billy Cheng: soon after I moved, then… then the company was doing a layoff, and then I find another job, and, you know, there’s just quite a bit of changes. To be honest, it just feels… it doesn’t feel very secure, it just always seems like something is changing.
52 00:05:19.090 ⇒ 00:05:25.119 Robert Tseng: Yeah, yeah, I mean, this, this is an exciting and also nervous time for a lot of people.
53 00:05:25.550 ⇒ 00:05:25.989 Billy Cheng: A lot of…
54 00:05:25.990 ⇒ 00:05:28.579 Robert Tseng: the way that I think… I mean, I just…
55 00:05:28.900 ⇒ 00:05:37.500 Robert Tseng: I think white-collar work is gonna change a lot. It’s already changing a lot. Yeah, so…
56 00:05:38.220 ⇒ 00:05:45.549 Billy Cheng: Exactly. How do you think, like, maybe for your company, when did you guys start the company? Do you see, like.
57 00:05:45.920 ⇒ 00:05:53.139 Billy Cheng: it’s growing faster because of AI, or do you feel like it comes with its challenges as well?
58 00:05:53.300 ⇒ 00:05:58.329 Robert Tseng: Yeah, so my story with the starting business thing,
59 00:05:59.940 ⇒ 00:06:02.850 Robert Tseng: I’m, like, losing traffic on all the time. When did I start?
60 00:06:05.950 ⇒ 00:06:07.720 Robert Tseng: I guess… okay.
61 00:06:08.680 ⇒ 00:06:10.320 Robert Tseng: Oh, wait. Huh.
62 00:06:11.040 ⇒ 00:06:12.410 Robert Tseng: 202
63 00:06:12.800 ⇒ 00:06:26.099 Robert Tseng: We moved… okay, well anyway, I don’t… I might have gotten the dates a little bit mixed up, but I think, what I… what I… what I think is that, I think 3 years ago from now, from now is when I started, kind of,
64 00:06:26.280 ⇒ 00:06:28.639 Robert Tseng: Just doing my own thing.
65 00:06:28.640 ⇒ 00:06:29.340 Billy Cheng: Yeah.
66 00:06:29.340 ⇒ 00:06:31.040 Robert Tseng: Yeah, I actually was…
67 00:06:31.530 ⇒ 00:06:44.349 Robert Tseng: Well, yeah, I had… I just left the role, I was, like, leading a data team at a CPG brand in LA, it’s called Ruggable. They actually have a UK office, but I don’t… probably never heard of it.
68 00:06:44.350 ⇒ 00:06:45.070 Billy Cheng: Honestly.
69 00:06:45.070 ⇒ 00:06:53.230 Robert Tseng: Yeah, and I was also working for, the, like, a church in LA, it’s called Tapestry.
70 00:06:53.600 ⇒ 00:06:54.510 Billy Cheng: So…
71 00:06:54.510 ⇒ 00:07:00.079 Robert Tseng: Yeah, I ended up not continuing with that church, and then…
72 00:07:00.080 ⇒ 00:07:15.230 Robert Tseng: I, like, yeah, I quit this job, so I was kind of unemployed, like, right before I got engaged. So then, yeah, then I proposed, and then during my engagement period, my wife Rachel says that was a very stressful time, because I was, like, just, like, figuring out what to do.
73 00:07:15.400 ⇒ 00:07:19.589 Robert Tseng: Yeah, so I started a couple different things,
74 00:07:19.790 ⇒ 00:07:38.170 Robert Tseng: Yeah, I was actually trying to launch a product company myself, like, because I had been in the CPG space for long enough that I felt like I wanted to launch some direct-to-consumer e-commerce thing. But, I also needed to, I don’t know, pay my bills, so I was, like, freelancing on the side, so…
75 00:07:38.170 ⇒ 00:07:39.300 Billy Cheng: Yeah.
76 00:07:39.300 ⇒ 00:07:46.389 Robert Tseng: eventually, like, the services business ended up taking off more. I just got enough clients that I was like, oh, okay, I think this is actually…
77 00:07:46.760 ⇒ 00:07:47.600 Billy Cheng: Oh, nice.
78 00:07:47.600 ⇒ 00:07:48.800 Robert Tseng: Working out.
79 00:07:49.750 ⇒ 00:07:53.199 Robert Tseng: It’s faster than the… than the product company.
80 00:07:53.200 ⇒ 00:07:53.760 Billy Cheng: Yes.
81 00:07:53.760 ⇒ 00:08:08.649 Robert Tseng: So yeah, so I started working… I just started doing that, and I just, for the first year, by the time I got married, I maybe just, like, had a small team, just me, and I was, like, I was running everything business-wise, and then had, like, 3 subcontractors.
82 00:08:08.650 ⇒ 00:08:09.730 Billy Cheng: Hmm. Nice.
83 00:08:09.730 ⇒ 00:08:26.989 Robert Tseng: My business partner kind of had a similar timeline. He started doing his freelancing thing around the same time I did, but he had also, like, a small team of 3 people. Probably, like, when we first moved to New York, I got connected to him through a mutual friend, and, we started…
84 00:08:27.160 ⇒ 00:08:31.769 Robert Tseng: working together on the same clients, like, we would share clients and stuff.
85 00:08:32.190 ⇒ 00:08:48.209 Robert Tseng: So, yeah, after working with him for, like, about 6 months, I was like, okay, well, I think it would make more sense for us to kind of bring it all under one house. So yeah, we’ve been working together, kind of, more officially under, like, the Brain Forge brand for almost 2 years now.
86 00:08:48.270 ⇒ 00:08:54.270 Robert Tseng: And, yeah, we’re about, like, 25 people now, and yeah, the business has grown a lot.
87 00:08:54.700 ⇒ 00:09:11.580 Robert Tseng: Yeah, so I think… oh, that’s the story. And then, like, as far as the tech side, yeah, I think it’s been really exciting, because we’re, first and foremost, like, a data… we were, like, a data consultancy. He’s more, like, a pure data engineer. I was more, like, strategy and analytics, and…
88 00:09:11.820 ⇒ 00:09:18.479 Robert Tseng: I think the narrative has really changed in the past 6 months specifically, where,
89 00:09:18.670 ⇒ 00:09:34.129 Robert Tseng: Yeah, I feel like when we first started trying to, like, pitch AI work, like, two years ago, kind of, like, teaching people how to integrate ChatGPT into their work, like, and, like, building, like, these automation demos, like, those are really small projects, and they just kind of just got stuck there.
90 00:09:34.130 ⇒ 00:09:40.889 Robert Tseng: And then, like, 6 months ago, some new breakthroughs in AI, things like, I don’t know how familiar you are with the technical concepts, but, like.
91 00:09:41.000 ⇒ 00:09:44.420 Robert Tseng: MCP and, like,
92 00:09:44.820 ⇒ 00:09:57.870 Robert Tseng: I guess even, like, Claude Co-work and, like, some of these interesting UIs that are coming out, where people are seeing that you can really now plug a bunch of different things into a system, and, like, yeah, like, you’re not…
93 00:09:58.100 ⇒ 00:10:12.289 Robert Tseng: you don’t lose context, it’s able to automatically pull information from multiple sources, like, it’s actually sophisticated enough now that, it can do more complicated workflows. So, yeah, I feel like
94 00:10:12.290 ⇒ 00:10:26.410 Robert Tseng: But it’s also harder to set up, because it does involve, you needing to, like, have clean data in order for these systems to actually, produce the best results. And so I think people started to see that, like.
95 00:10:26.550 ⇒ 00:10:33.130 Robert Tseng: of AI engineering work is really a data problem, and I think, like, the same people that we were…
96 00:10:33.330 ⇒ 00:10:44.640 Robert Tseng: that didn’t move on past, like, some of the demos that we built for them. Like, they called us 6 months ago, they’re like, hey, actually, you know, we want you to come in and help clean up our data as well, because we.
97 00:10:44.640 ⇒ 00:10:45.050 Billy Cheng: Audio.
98 00:10:45.050 ⇒ 00:11:01.310 Robert Tseng: in order to use these things. So, yeah, I feel like the business really picked up the past six months. Yeah, we doubled our headcount and, like, are growing a lot faster now. So, yeah, I think it was really just, like, a timing thing. We’re, like, kind of catching it at a really, exciting time.
99 00:11:01.830 ⇒ 00:11:10.120 Billy Cheng: Oh. That’s really good. So, like, how does that work? Normally, like, when a client comes to you, do they… are they looking for…
100 00:11:10.450 ⇒ 00:11:16.140 Billy Cheng: does kind of AI help to be more productive, or are they looking for… Like…
101 00:11:16.330 ⇒ 00:11:18.250 Billy Cheng: some sort of AI system
102 00:11:18.530 ⇒ 00:11:27.170 Billy Cheng: to kind of work in a ecosystem where they can just essentially work like a CDP of some sort.
103 00:11:27.170 ⇒ 00:11:28.829 Robert Tseng: How does that work?
104 00:11:29.630 ⇒ 00:11:42.250 Robert Tseng: Yeah, okay, so great. So, I mean, I think I saw you’re in lifecycle marketing, so I’m assuming you’re familiar with CDPs. Do you use any other, like, kind of AI systems? Like, I’m just curious, like, how I will adapt my… how I share it with you.
105 00:11:42.490 ⇒ 00:11:43.090 Billy Cheng: I…
106 00:11:43.420 ⇒ 00:11:54.950 Billy Cheng: I don’t think I’m that technical. I think in the past, AS Watson, they tried to build their own CDP, which is crazy, but I think, yeah, different companies have just tried to…
107 00:11:55.340 ⇒ 00:12:06.430 Billy Cheng: I don’t know, pitched some sort of CDP, and then they said, oh, you can connect your transactional data with, like, other data, social media, like, weather, like, all these stuff.
108 00:12:06.590 ⇒ 00:12:21.109 Billy Cheng: So I have some understanding, but I don’t think I really know, so I’m just kind of curious to know, like, the solution that you guys are talking about. Is it more, you know, leveraging on existing
109 00:12:21.280 ⇒ 00:12:28.140 Billy Cheng: maybe AI tools, but, like, connecting the different data points together.
110 00:12:28.360 ⇒ 00:12:33.410 Billy Cheng: Or is it more just, like, you basically build the infrastructure for them?
111 00:12:34.180 ⇒ 00:12:41.540 Robert Tseng: Yeah, so, I mean, I think that’s a great question. So, I think we do try to,
112 00:12:42.690 ⇒ 00:12:52.219 Robert Tseng: we try to build on existing infrastructure. So, like, actually, like, I have a lot of experience with CDPs now. We’re, like, a partner for Segments and Hightouch, and, like, a few of these ones.
113 00:12:52.220 ⇒ 00:12:52.779 Billy Cheng: Oh, I see.
114 00:12:52.780 ⇒ 00:12:53.220 Robert Tseng: So…
115 00:12:53.220 ⇒ 00:12:56.509 Billy Cheng: I touched just, had a call with us or something.
116 00:12:57.130 ⇒ 00:13:09.499 Billy Cheng: They keep trying to pitch it to us in a very shady way as well. They’re like, oh, we work for your U.S. team, but then when we asked the U.S. team, they were like, oh, we don’t know what high touch is.
117 00:13:10.620 ⇒ 00:13:13.170 Billy Cheng: He’s, like, sneaky, but yeah.
118 00:13:13.170 ⇒ 00:13:13.840 Robert Tseng: Yeah.
119 00:13:13.990 ⇒ 00:13:17.099 Robert Tseng: But I think with these tools, I mean.
120 00:13:17.480 ⇒ 00:13:21.939 Robert Tseng: I think traditionally, a CDP was just like a data connector tool, right? It was just kind of
121 00:13:21.940 ⇒ 00:13:41.770 Robert Tseng: like you said, you can plug a bunch of different things together. Maybe there’s, like, some very basic customer profile stitching, in, like, the… in the app, and then you can, like, create, like, some basic segments, and you want… I think there were some opinions about, like, should marketers work in the CDP, or should they work in, like, their customer engagement platform, maybe wherever they’re sending their emails or messages from.
122 00:13:41.780 ⇒ 00:13:44.060 Robert Tseng: And,
123 00:13:44.080 ⇒ 00:14:01.260 Robert Tseng: Yeah, I think, like, yeah, we’re not opinionated about, like, where the UI should be. Like, I think we more have more of an infrastructure perspective of, like, even for these CDPs to work, like, you need to… like, the identity stitching you can do in the CDP is, like, too limited. We need… you need to move into the
124 00:14:01.260 ⇒ 00:14:07.339 Robert Tseng: warehouse. And so, I think our, like, bias is that, like, we try to work with clients that are willing to
125 00:14:07.380 ⇒ 00:14:10.980 Robert Tseng: land all of their data in a data warehouse, or…
126 00:14:11.400 ⇒ 00:14:22.189 Robert Tseng: at least some sort of, like, data lake or data control room, whatever you call it. And, yeah, then, like, the CDP can, like, pull data in and out of that, but we want to maintain that source of truth.
127 00:14:22.260 ⇒ 00:14:42.240 Robert Tseng: And because we have the most flexibility to kind of structure data the way we want to in the warehouse, and we do it in a way where it becomes easy for, like, these LLM-based tools to be able to go in and, like, get context from it. So, I think, like, our strength is really in, like, the context engineering. Like, we work with a lot of
128 00:14:42.240 ⇒ 00:14:56.800 Robert Tseng: it’s not just, like, structured data, but, like, a lot of organizations’ documentation, being able to, like, use all the SOPs that teams are writing, and be able to use, kind of, like, turn those into skills, that, like, AI agents can use to go and, like, actually interact with the data.
129 00:14:56.800 ⇒ 00:15:02.760 Robert Tseng: So, I think, like, the moment that clicks for people is when they realize, like, hey.
130 00:15:02.800 ⇒ 00:15:10.590 Robert Tseng: if I upload a dataset to, like, ChatGBT and ask it a question, it doesn’t give a very nuanced answer, but if you do it with…
131 00:15:10.720 ⇒ 00:15:28.519 Robert Tseng: If you actually invest in this infrastructure, when you ask that question, on top of your own data, the way that your team would want to see it specifically, it, like, gets it right. And, like, I… I think, like, we just have to, like, get people to… to see that, like, oh, wow, this actually… this actually works, and,
132 00:15:28.520 ⇒ 00:15:34.330 Robert Tseng: that that’s, that’s kind of a cool moment that we work with. So.
133 00:15:34.330 ⇒ 00:15:35.800 Billy Cheng: Mmm, I see.
134 00:15:37.260 ⇒ 00:15:56.530 Robert Tseng: So we’re not a product company, we’re not, like, we’re not, like, committing to, like, a specific, like, toolset. Like, I think we know, like, what tools play well with each other, and, like, when we go into an organization, like, there’s usually a starting point. Like, you know, if a team doesn’t have a CDP, and they don’t have a data warehouse, then, like, that’s, like.
135 00:15:56.560 ⇒ 00:16:05.079 Robert Tseng: That’s like a… that’s a different stage of maturity than… I mean, I’m sure your organization has a lot of these, like, systems already. Maybe you have multiple ones, you have, like, a different.
136 00:16:05.480 ⇒ 00:16:12.339 Robert Tseng: for each one, and yeah, like, I think we would approach those two situations very differently.
137 00:16:12.340 ⇒ 00:16:28.350 Billy Cheng: I see. So, like, but then, in terms of, like, the users, essentially, all the data would be embedded in some sort of AI tool already, so then… that’s why they already know the context. So, like, instead of…
138 00:16:28.540 ⇒ 00:16:33.869 Billy Cheng: You know, someone just downloading the data and, like, importing it back to…
139 00:16:34.030 ⇒ 00:16:37.249 Billy Cheng: like, some sort of ChatGPT tool?
140 00:16:37.250 ⇒ 00:16:43.629 Robert Tseng: Like, that was already embedded into… like, the whole warehouse is already embedded with the agent.
141 00:16:44.000 ⇒ 00:16:54.039 Robert Tseng: Yep, yeah, so no more drag and drop, kind of… we do… yeah, all those integrations, that’s part of the work that we do to kind of, like, handle that in the background.
142 00:16:54.580 ⇒ 00:16:59.910 Robert Tseng: Yeah, and I think, like… There’s,
143 00:17:01.300 ⇒ 00:17:06.660 Robert Tseng: So we work with, like, some healthcare and fintech clients as well, who are, like,
144 00:17:10.410 ⇒ 00:17:18.500 Robert Tseng: if they’re asking a question to an LLM sometimes, they’ll get, like… if you ask it 10 different… 10 times, it’ll give you different answers every time, so…
145 00:17:18.500 ⇒ 00:17:18.880 Billy Cheng: Yeah.
146 00:17:18.880 ⇒ 00:17:30.310 Robert Tseng: Yeah, there’s just, like, some basic, like, like, engineering best practices where we do this type of, like, unit testing and adding CICD and all this kind of, like, all these guardrails so that the output is consistent.
147 00:17:30.340 ⇒ 00:17:43.719 Robert Tseng: So that, like, a team can, like, rely on what they’re seeing. So, yeah, ultimately, like, data problems are about, like, oh, people are always like, I don’t trust the data, right? And it’s like, I say, I think it looks…
148 00:17:43.780 ⇒ 00:17:58.399 Robert Tseng: I define it this way, someone else defines it a different way, and I… and I think that problem has… previously, like, it took me… and, like, I had a team of, like, 4 analysts, like, if I were running an analysis at my previous company,
149 00:17:58.890 ⇒ 00:18:08.710 Robert Tseng: yeah, I’d basically have 3 people run the same thing, and then we would get in a room and we’d talk about, like, which approach is better, and then we kind of decide that, like, okay, this is the way the company should do it.
150 00:18:08.720 ⇒ 00:18:26.049 Robert Tseng: But that would, like, be a really long feedback loop. Now I can basically have the same question, run it with, like, multiple agents, basically different LLN models, maybe kind of find two different ways, and run, like, 5 to 10 analyses… run 5 to 10 times the same way.
151 00:18:26.050 ⇒ 00:18:40.599 Robert Tseng: And then I can just go and be a reviewer, and that all happens, you know, that all happens within the same day, rather than, like, me having to wait, like, two to four weeks for everyone to, like, be ready to, like, kind of have that conversation. So, I feel like it’s just dramatically sped up,
152 00:18:40.600 ⇒ 00:18:54.789 Robert Tseng: the work that we do, so, like, we’re able to do, like, BI migrations in, like, a month, CDP migrations in 3 months, like, the speed at which I can do the data work that I knew how to do before is just, like, so much faster. It’s like…
153 00:18:54.790 ⇒ 00:19:07.509 Billy Cheng: Yeah, I see. Because, I mean, in the past, I feel like even right now, I feel like we’re still very much relying on, like, data analysts to write SQL and, like, pull the data, and someone needs to…
154 00:19:07.850 ⇒ 00:19:17.840 Billy Cheng: Someone needs to, like, do the Excel and the, like, charting. It’s still very much manual, like, I get that there are dashboards, but, like, I think if you ask people, they’re still doing…
155 00:19:18.030 ⇒ 00:19:35.410 Billy Cheng: like, what they were doing 10 years ago. Like, people are still just, like, writing SQL, pooling data, putting it in an Excel, and then that Excel isn’t even shared with their teammates. Each team member has their own Excel, and then they do their own PowerPoint. So I feel like it’s still…
156 00:19:35.450 ⇒ 00:19:45.859 Billy Cheng: it’s still silly, but I think what you’re describing is that, like, oh, hopefully we don’t need, like, so many analysts. Okay, it sounds quite scary, but, like, in a way.
157 00:19:46.460 ⇒ 00:19:48.099 Billy Cheng: Like, with the agent?
158 00:19:48.630 ⇒ 00:20:02.960 Billy Cheng: you know, hopefully you get those basic answers right away. So, like, the business users can just interact with some sort of generative AI tool to know, like, oh, how many customers are shopping this brand?
159 00:20:03.130 ⇒ 00:20:07.960 Billy Cheng: What’s my market share? What’s my penetration? So, like, everything is…
160 00:20:08.430 ⇒ 00:20:14.420 Billy Cheng: It’s… is able to be answered by… The agent, somehow?
161 00:20:14.820 ⇒ 00:20:15.310 Robert Tseng: Yeah.
162 00:20:16.810 ⇒ 00:20:18.180 Billy Cheng: Hmm…
163 00:20:18.180 ⇒ 00:20:33.639 Robert Tseng: Yeah, and then all… and all of the outputs, like, whether it’s a chart, slide, document, it’s all, like, you don’t need multiple applications for that. It all comes… it can… the model can kind of put out all… put out all of those things, and you can have it all in the same workspace.
164 00:20:34.210 ⇒ 00:20:35.050 Billy Cheng: Yeah.
165 00:20:35.380 ⇒ 00:20:35.970 Robert Tseng: Yeah.
166 00:20:36.260 ⇒ 00:20:37.739 Robert Tseng: So, I think Slash…
167 00:20:37.740 ⇒ 00:20:38.480 Billy Cheng: This is a.
168 00:20:38.480 ⇒ 00:20:50.529 Robert Tseng: not there yet, but, like, everything else, I think, is… is… is… is there. So we use all… we built, like, an internal system for our teams. I have my analysts be working out of, like, our internal tool instead of, like.
169 00:20:50.730 ⇒ 00:20:53.650 Robert Tseng: pair with, like, I don’t have them using,
170 00:20:54.060 ⇒ 00:21:05.580 Robert Tseng: like, spreadsheet, PowerPoint, and, like, they’re not directly using a command line to query from, like, the SQL database either, like, everything could just be in the same workspace.
171 00:21:06.050 ⇒ 00:21:09.660 Billy Cheng: So everyone just interact with some sort of generative AI,
172 00:21:10.180 ⇒ 00:21:15.350 Billy Cheng: tool, essentially. They are just, like, asking business questions.
173 00:21:15.780 ⇒ 00:21:21.280 Billy Cheng: And then, oh, can you give me a chart on, I don’t know, market share in the past 4 months?
174 00:21:22.150 ⇒ 00:21:26.189 Billy Cheng: And so they can’t get the answer that way. I see. Wow.
175 00:21:26.340 ⇒ 00:21:40.290 Robert Tseng: Yeah, from the business intelligence side, I think that’s kind of what it’s become, and we partner with a tool called Omni that does a really good job, so we would implement that, but I think there’s… yeah, I mean, I think the use cases are, like.
176 00:21:40.600 ⇒ 00:21:49.790 Robert Tseng: There’s so… there’s so many, and yeah, pretty much have turned everybody on our team into an engineer. Everybody’s pushing code, even, like, our marketing people, they can, like.
177 00:21:50.250 ⇒ 00:22:01.560 Robert Tseng: directly from Slack, if they need to, spin up a new landing page, they can give the instructions in Slack, and they change the copy because they just got off a call at the lead.
178 00:22:01.560 ⇒ 00:22:16.719 Robert Tseng: And then it will, like, we have agents running the background that are basically using our design framework, changing the copy, and then putting… spitting it out directly, and we can send it to the client within, like, within a few minutes. Or if they wanted to look at a demo of, like.
179 00:22:16.720 ⇒ 00:22:21.940 Robert Tseng: Because we do custom software as a service, pretty much. So, they wanted to see, like, how do I, like.
180 00:22:22.480 ⇒ 00:22:39.320 Robert Tseng: is there, like, a demo environment that they can play with that’s, like, kind of with this, like, AI workspace kind of envision? Like, we just… they can also trigger that agent from… from Slack directly. So, yeah, like, I think that’s… that was just, like.
181 00:22:40.210 ⇒ 00:22:48.720 Robert Tseng: I mean, I never thought I could, like, build an app or a web page from sending a Slack message, you know? It’s just, like, it’s not crazy. Yeah, yeah.
182 00:22:48.720 ⇒ 00:22:57.569 Billy Cheng: See, I feel like I’m unfamiliar with, like, the different tools. Like, would companies have, like, concern in terms of, like, governance?
183 00:22:57.720 ⇒ 00:23:06.020 Billy Cheng: Because I think for L’Oreal, like, they basically built, like, so-called L’Oreal GBT, essentially, they just pull it offline.
184 00:23:06.120 ⇒ 00:23:08.740 Robert Tseng: Yeah. And then they control…
185 00:23:09.080 ⇒ 00:23:15.369 Billy Cheng: Kind of, like, what sort of data could stay in that, like, environment?
186 00:23:15.380 ⇒ 00:23:18.120 Robert Tseng: Yeah. So hopefully the data wouldn’t go out.
187 00:23:18.150 ⇒ 00:23:25.580 Billy Cheng: So I don’t know, like, like, how does that work for companies with certain sizes, and then they would have concerns on…
188 00:23:25.700 ⇒ 00:23:27.980 Billy Cheng: You know, who has access to…
189 00:23:28.210 ⇒ 00:23:34.230 Billy Cheng: to their consumer database? Like, how does that work? Like, for those tools to not be…
190 00:23:34.570 ⇒ 00:23:40.279 Billy Cheng: Share, like, in the public domain, and being super private and secure, and stuff like that.
191 00:23:40.710 ⇒ 00:23:45.710 Robert Tseng: Yeah, yeah, I think that’s totally valid, and I think that’s why… Like…
192 00:23:45.980 ⇒ 00:23:56.519 Robert Tseng: controlling your data in a… in, like, a day warehouse is probably the safest, like, that’s kind of where we start. One of, like, my… one of my partners is, he’s like.
193 00:23:57.270 ⇒ 00:24:05.629 Robert Tseng: he’s, like, an ex-military, like, cyber intelligence kind of guy, so, like, yeah, one of our clients, I mean, we work with some…
194 00:24:06.310 ⇒ 00:24:17.229 Robert Tseng: government agencies, fintech, healthcare, so they all need it to be very, containerized, and, like, you need to, like, have a very high privacy, bar, but, like.
195 00:24:17.260 ⇒ 00:24:28.849 Robert Tseng: Yeah, what we… the system that we would custom build for, like, a client, like, if we were to work with, like, a… like, with that type of specification, like, it would be… it would be a closed-loop environment, like, there…
196 00:24:28.850 ⇒ 00:24:38.730 Robert Tseng: The model is just trained locally, and it doesn’t actually go back out to… yeah, and it never leaves, it never leaves your infrastructure, basically.
197 00:24:38.730 ⇒ 00:24:39.090 Billy Cheng: Hmm.
198 00:24:39.090 ⇒ 00:24:43.439 Robert Tseng: Kind of the simple way of putting it. Oh, I see. So, yeah.
199 00:24:44.170 ⇒ 00:24:53.049 Robert Tseng: sacrifice a little bit of performance, because you can’t use, like, the most frontier models, but it’s not… the performance is, like, not much. Like, I…
200 00:24:53.150 ⇒ 00:24:58.449 Robert Tseng: I don’t know how much you follow closely the model developments, but, like,
201 00:24:58.950 ⇒ 00:25:07.229 Robert Tseng: well, I… the open source models are, like, they’re, like, the… I forgot which Chinese company released, like, the open source model recently.
202 00:25:07.230 ⇒ 00:25:24.500 Robert Tseng: Well, they’ve taken it down since, because, like, it basically is… it performed… it performed at the same benchmark as, like, GPT-4.0, so, like, it’s like, you know, and that’s something that you can literally host on your phone. It only takes, like, 16 gigabytes to run that… that model, so… I see.
203 00:25:24.500 ⇒ 00:25:26.500 Billy Cheng: Yeah, like, I think.
204 00:25:26.700 ⇒ 00:25:28.210 Robert Tseng: It’s only gonna get…
205 00:25:28.960 ⇒ 00:25:48.560 Robert Tseng: easier to host these models in your own environment. And that’s, yeah, I think that’s why, companies, like, the barrier to build your own kind of, like, custom AI environment and application is just, like, very low now. Yeah.
206 00:25:48.560 ⇒ 00:25:50.530 Billy Cheng: I see, that’s really cool.
207 00:25:51.700 ⇒ 00:26:06.120 Billy Cheng: Sounds like your company is doing quite well, like, getting new clients and stuff like that. And I guess it’s a trend, too, like, everyone just talks about AI, AI adoption, like, how to be more efficient, how to save time, something like that.
208 00:26:06.670 ⇒ 00:26:07.670 Billy Cheng: Yeah.
209 00:26:07.880 ⇒ 00:26:11.200 Robert Tseng: Yeah, I mean, I… it’s kind of like a…
210 00:26:12.510 ⇒ 00:26:18.320 Robert Tseng: good and bad, I think. I think it’s really a moment in time. I think, like, this is a short window.
211 00:26:18.510 ⇒ 00:26:27.000 Robert Tseng: like, I think these models are getting exponentially better every time they have a new release, so…
212 00:26:27.600 ⇒ 00:26:34.350 Robert Tseng: yeah, like, I… I frankly think that they could probably replace our business within, like, a couple years, like, we…
213 00:26:34.350 ⇒ 00:26:34.870 Billy Cheng: Literally.
214 00:26:34.870 ⇒ 00:26:37.749 Robert Tseng: Have, like, a year or two, kind of.
215 00:26:37.750 ⇒ 00:26:38.440 Billy Cheng: Really?
216 00:26:38.440 ⇒ 00:26:44.360 Robert Tseng: do as much as we can, and then I, I, I, yeah, I… I just think that, there, there, we’re…
217 00:26:44.470 ⇒ 00:26:50.859 Robert Tseng: Like, we… we’re… a business like ours is not gonna be… not gonna be necessary anymore.
218 00:26:50.860 ⇒ 00:26:52.190 Billy Cheng: Oh, I… Yeah.
219 00:26:52.650 ⇒ 00:26:53.200 Robert Tseng: Yeah.
220 00:26:53.200 ⇒ 00:26:59.550 Billy Cheng: It’s okay. You can… you can start another company at that point, or… Yeah. Or retire, or something.
221 00:26:59.550 ⇒ 00:27:09.979 Robert Tseng: Yeah. So, but it’s fun, we get to be in conversations, like, yeah, I think we’re… we’re in conversations with clients, or working with clients that I, like.
222 00:27:10.690 ⇒ 00:27:12.240 Robert Tseng: much bigger, and…
223 00:27:12.360 ⇒ 00:27:21.439 Robert Tseng: and it happened all much faster than I thought, thought, than I thought it would. So, it’s… it’s been… it’s been a fun… it’s been a fun, fun journey, so far, yeah.
224 00:27:21.440 ⇒ 00:27:28.810 Billy Cheng: Yeah. I mean, I think I would love to understand more slowly if you have any existing materials?
225 00:27:28.970 ⇒ 00:27:43.589 Billy Cheng: Yeah. Maybe I can just kind of learn more as well. I feel like I just need more education in this area. Totally. Yeah, because I feel like even in the companies I’ve worked for, like, even though they talk about CDP, they talk about
226 00:27:43.880 ⇒ 00:27:50.279 Billy Cheng: like, maybe new tools that they’re adopting. I feel like at the end of the day, I’m still just writing SQL.
227 00:27:50.750 ⇒ 00:28:09.579 Billy Cheng: Yeah. And I saw people just doing Excel still, and still doing PowerPoint. They’re still doing what they were doing 10 years ago. So I’m sure there’s a much better way, a much smarter way to do things, but I just don’t think we are really adopting it.
228 00:28:10.080 ⇒ 00:28:10.610 Robert Tseng: Yeah.
229 00:28:10.950 ⇒ 00:28:11.850 Billy Cheng: Yeah.
230 00:28:12.950 ⇒ 00:28:18.359 Robert Tseng: Well, I mean, I get it, it’s hard to really… like, it fundamentally, like, changes the way
231 00:28:19.460 ⇒ 00:28:28.800 Robert Tseng: you think about the work, so it’s really hard to kind of, like, drive. It’s really the education piece that has been the missing, like, link, I think.
232 00:28:28.990 ⇒ 00:28:47.810 Robert Tseng: Well, yeah, what we’re… what we’re saying now is pretty much the same thing we were saying two years ago, it’s just that I think people are… the market is more educated now, and people are… are seeing some of the… seeing some more use cases. But yeah, I mean, I think I… I mean, we’ve worked with a bunch of… I think lifecycle teams are actually a good use case for us, because,
233 00:28:48.120 ⇒ 00:29:07.119 Robert Tseng: Yeah, I think, the outcome is very clear. It’s usually something like, oh, we want to improve reorder rates, like, help us optimize, like, our campaigns. So they just need better segmentation, better messaging strategy, and so if you could speed that up by helping them run faster experiments, or giving them better cuts of data.
234 00:29:07.120 ⇒ 00:29:25.429 Robert Tseng: then, like, it has, like, an immediate impact. So, in the marketing world, like, that’s typically the, the, the stakeholder that we, that we work with the best. We also work with product and ops teams. So, on operations, it’s typically, like, a customer success kind of person who’s,
235 00:29:25.430 ⇒ 00:29:42.249 Robert Tseng: yeah, like, I… I think just, like, handling, like, high… high ticket volume, and, you know, there’s a lot of data, call center, and, like, emails, tickets, all this kind of stuff. So, yeah, I think it’s… it… for… for teams who are, like.
236 00:29:42.490 ⇒ 00:29:59.870 Robert Tseng: it’s… it’s a lot clearer, because you have, like, a very specific metric that you can hold your team accountable to. It’s… it’s easier to prove, like, the… the value there, compared to, like, what my wife does. Like, she’s kind of in lifecycle, but kind of more creative as well. But their stuff is, like, very much, like…
237 00:30:00.920 ⇒ 00:30:06.749 Robert Tseng: I don’t really think they really know how to measure their performance very, very clearly, very quickly. So.
238 00:30:06.750 ⇒ 00:30:07.340 Billy Cheng: soon.
239 00:30:07.650 ⇒ 00:30:13.710 Billy Cheng: No, I think same for us. Well, I think, in a way, like, even my last job, I was with,
240 00:30:14.260 ⇒ 00:30:18.190 Billy Cheng: a company called Costa Coffee, so, like, they sell coffee, I was part of.
241 00:30:18.190 ⇒ 00:30:20.080 Robert Tseng: Yeah, I know cost of coffee. Yeah.
242 00:30:20.670 ⇒ 00:30:30.060 Billy Cheng: But I felt like what I was trying to describe is that, like, there would still be a data team, like, they would try to, you know, write SQL, you know, have the data, put it into a PowerPoint.
243 00:30:30.320 ⇒ 00:30:30.660 Robert Tseng: Yeah.
244 00:30:30.660 ⇒ 00:30:37.930 Billy Cheng: But then, maybe you will put down 10 insights into the PowerPoint. Maybe, like, if you’re lucky, one or two insights we can…
245 00:30:38.130 ⇒ 00:30:40.369 Billy Cheng: be used by the marketing team.
246 00:30:41.060 ⇒ 00:30:46.780 Billy Cheng: So, like, the whole process is very manual, like, it goes through so many different meetings.
247 00:30:47.080 ⇒ 00:30:47.420 Robert Tseng: Yeah.
248 00:30:47.420 ⇒ 00:30:49.199 Billy Cheng: In order to action.
249 00:30:49.340 ⇒ 00:30:50.280 Billy Cheng: But, like…
250 00:30:50.570 ⇒ 00:31:06.940 Billy Cheng: if I understand what your vision is, like, essentially, well, you don’t need all these meetings or people, like, if you just empower the marketing person to go directly into this database, which is already connected to some sort of AI,
251 00:31:07.620 ⇒ 00:31:10.070 Billy Cheng: And then they can just ask the question.
252 00:31:10.240 ⇒ 00:31:13.729 Robert Tseng: Yep. And, like, the insight’s already there, so they hopefully can…
253 00:31:13.840 ⇒ 00:31:19.510 Billy Cheng: somehow action it through Salesforce or whatever CRM tool.
254 00:31:19.920 ⇒ 00:31:25.170 Billy Cheng: Rather than, you know, going through all these meetings and asking for measurements.
255 00:31:26.000 ⇒ 00:31:29.830 Billy Cheng: And then they might just cherry-pick one or two things to action.
256 00:31:29.850 ⇒ 00:31:30.720 Robert Tseng: Yeah.
257 00:31:32.800 ⇒ 00:31:37.309 Robert Tseng: Yeah, no, I think that… I think your intuition is totally spot on, like, I think…
258 00:31:37.630 ⇒ 00:31:55.439 Robert Tseng: data people, like, in-house data teams are, like, they gatekeep too much. Like, I think governance is important, but once the definitions are pretty defined and agreed upon, like, the power needs to shift over to the operators, or, like, the marketers, or whatever. Like, I think they need… they need to, like, they’re the ones that are actually gonna…
259 00:31:55.440 ⇒ 00:32:01.309 Robert Tseng: pick, like, decide on the strategy, and so giving them the tools, like, is… is better, so…
260 00:32:01.310 ⇒ 00:32:14.409 Robert Tseng: I think when I first started, I would like selling to, like, heads of data, telling them, hey, instead of you hiring 5 data engineers, we can just, like, you can hire us, we’re basically, like, a fractional data team, where we can go much faster.
261 00:32:14.410 ⇒ 00:32:14.860 Billy Cheng: Yeah.
262 00:32:14.860 ⇒ 00:32:34.299 Robert Tseng: I think I was, like, trying to think… I was thinking that, like, the ideal customer for me would be more of, like, a head of data engineering, but then it turns out, it’s like, well, they actually don’t want that. Like, they don’t care, like, they don’t feel the pain. They’re like, well, they already got this budget from the company, they’re just gonna hire the people. Going faster just puts the pressure on themselves, so, like, I don’t really necessarily want that.
263 00:32:34.300 ⇒ 00:32:36.130 Billy Cheng: So…
264 00:32:36.130 ⇒ 00:32:54.200 Robert Tseng: Yeah, so, like, well, then who can… who in the team that I… in the organization can I actually work with, where making them look good and their performance is actually, like, a win-win, where, like, they… they… they’re happy that, like, they’re… they… they look better because they’re more productive, and then, like, we get to kind of showcase, kind of, like, the…
265 00:32:54.200 ⇒ 00:32:54.690 Billy Cheng: the capability.
266 00:32:54.690 ⇒ 00:33:01.189 Robert Tseng: And it turns out it’s more of, like, the business people. It’s the marketers, it’s the product people.
267 00:33:01.410 ⇒ 00:33:01.810 Robert Tseng: Yeah.
268 00:33:02.190 ⇒ 00:33:04.620 Robert Tseng: engineers after that.
269 00:33:04.740 ⇒ 00:33:05.650 Robert Tseng: Yeah.
270 00:33:05.720 ⇒ 00:33:10.889 Billy Cheng: I mean, I think I thought about… I think we did explore, oh, maybe we can just…
271 00:33:11.180 ⇒ 00:33:18.860 Billy Cheng: like, you know, on Databricks, you can grant access to more people, and then there’s a… some sort of AI function.
272 00:33:18.970 ⇒ 00:33:23.100 Billy Cheng: Where you can, you know, the business person can raise a question.
273 00:33:23.480 ⇒ 00:33:33.130 Billy Cheng: and then within Databricks, then they can give you an answer. But then I was told that, oh, what about governance? Like, how do you ensure that, you know.
274 00:33:33.250 ⇒ 00:33:47.850 Billy Cheng: the answer given by AI is… is… is approved, and it’s accurate. I think that was the challenge. So we didn’t really go down that route, but then we did explore, oh, you know, why don’t we just use Databricks since we already are paying for Datarex?
275 00:33:47.850 ⇒ 00:33:48.430 Robert Tseng: Yeah.
276 00:33:49.550 ⇒ 00:33:55.580 Billy Cheng: And, like, essentially the database is already, you know, connected to Databricks.
277 00:33:56.550 ⇒ 00:33:57.130 Robert Tseng: Yeah.
278 00:33:58.170 ⇒ 00:34:03.499 Robert Tseng: No, I think that’s… I think that’s right. I mean, we’re… we’re actually a Snowflake partner, so, on the other
279 00:34:03.890 ⇒ 00:34:05.099 Robert Tseng: So, like.
280 00:34:05.100 ⇒ 00:34:05.570 Billy Cheng: That’s, yeah.
281 00:34:05.570 ⇒ 00:34:09.480 Robert Tseng: But, like, we, we, we…
282 00:34:10.560 ⇒ 00:34:15.189 Robert Tseng: I guess their whole system is called Cortex, but it’s very… it’s very similar. I think,
283 00:34:15.360 ⇒ 00:34:19.800 Robert Tseng: to Databricks in terms of capabilities. So, yeah, I think,
284 00:34:20.080 ⇒ 00:34:31.050 Robert Tseng: both Databricks and Snowflake have had a hard time, like, really reaching to… out to, like, non-technical audience, because it just… it’s kind of an intimidating environment. Typically, it’s just been only engineers that are using it.
285 00:34:31.080 ⇒ 00:34:42.900 Robert Tseng: But now that they both have their own marketplaces, they have other developers building apps on top of it, and, like, the barrier to building an app is, like, so… so low now. I think their whole… they’re promoting, like.
286 00:34:42.940 ⇒ 00:34:52.680 Robert Tseng: Creating, like, a more, like, non-technical, friendly, like, like, interface for people to be able to go and actually, like, get the answers they want.
287 00:34:52.719 ⇒ 00:35:02.430 Robert Tseng: But, yeah, so I think ultimately the governance thing, it does require people to, like, kind of enforce that, and to show that it’s trustworthy, and so I think that’s why, like.
288 00:35:02.430 ⇒ 00:35:13.340 Robert Tseng: you know, and I think a lot of data teams are not good at doing that, because they’re used to just, like, we’ll go get the answer ourselves, and, like, we’ll be the gatekeepers, rather than, like, building, like.
289 00:35:13.340 ⇒ 00:35:22.580 Robert Tseng: A true governance framework and system that, like, is able to just, like, kind of self-regulate, have the right guardrails, so that people understand, like.
290 00:35:22.670 ⇒ 00:35:47.650 Robert Tseng: and, like, even in the systems, when you’re asking a question, like, it will kind of adjust your… it’ll adjust your prompt, or it’ll kind of, like, give you… give you the… a more nuanced answer. It won’t let you kind of, like, just, like, get something random out of it, right? So, I think that is part of, like… that is… that is kind of, like, an engineering problem that needs to be solved for… for self-service to really be adopted. And it’s obviously still a challenge, like, we see
291 00:35:47.650 ⇒ 00:35:48.480 Robert Tseng: that across.
292 00:35:48.480 ⇒ 00:35:51.329 Robert Tseng: our, our, our, our clients as well.
293 00:35:51.330 ⇒ 00:35:51.899 Billy Cheng: on stage.
294 00:35:52.090 ⇒ 00:35:57.150 Robert Tseng: Yeah, I mean, one more thing I’ll say is, like, so, like, we work with the…
295 00:35:57.460 ⇒ 00:36:02.499 Robert Tseng: a big… a big brand. They have, like, it’s like a team of four,
296 00:36:02.740 ⇒ 00:36:09.560 Robert Tseng: people, retail, kind of working with wholesale partners. They have, like, about 10,000 plus wholesale partners. Yeah.
297 00:36:09.620 ⇒ 00:36:22.010 Robert Tseng: And, I think, like, last year, they were maybe working, like, 4,000, but they knew they were gonna double their wholesale certain account, and so that when they worked with us, they’re like, okay, how do we, like, basically keep the same number of people, but able to, like.
298 00:36:22.010 ⇒ 00:36:41.400 Robert Tseng: they show that they’re going to be two times as productive. And, yeah, a lot of that ended up becoming, like, a data governance project, just to make sure as you’re adding new partners, there’s going to be new categories, the segments are going to be… it’s always a segmentation problem. So, making sure that, like, we’re there to kind of help… help them
299 00:36:41.680 ⇒ 00:36:46.550 Robert Tseng: Grow, and manage the taxonomy of, like, options that they have.
300 00:36:46.550 ⇒ 00:37:03.500 Robert Tseng: As they’re, like, trying… as they’re… as they’re building their… their… their network. So, I think that’s pretty much, like, what we’ve become, which is, like, we’re just… we stay on as governors, but then the actual, like, updates to… to any of the data models, any of the routine reports, that’s all already kind of, like.
301 00:37:03.500 ⇒ 00:37:10.880 Robert Tseng: code… codified, and that agents just run that at this point. So, we’re just there just to continue to assess the changing definitions.
302 00:37:11.450 ⇒ 00:37:16.909 Billy Cheng: Yeah. No, I think I’m… as you’re saying, I was just thinking about, like.
303 00:37:17.130 ⇒ 00:37:36.210 Billy Cheng: these initiatives my boss bosses asked us to do. I think they were just saying that, oh, for different teams, how can we use the Lorel GBT more? How can we use the AI tool to create more visuals? But I don’t think anyone has really explored connecting the consumer database with some sort of AI tool.
304 00:37:37.240 ⇒ 00:37:44.260 Billy Cheng: because, like, even the use cases that we kind of explore, I don’t think we really explore…
305 00:37:44.540 ⇒ 00:37:53.390 Billy Cheng: like, connecting, let’s say, the L’Oreal GBT with the consumer database? Because that’s what you’re saying, right? Essentially.
306 00:37:53.500 ⇒ 00:37:58.390 Billy Cheng: connecting the database directly with some sort of AI tool.
307 00:37:59.140 ⇒ 00:38:02.069 Billy Cheng: Yeah, I’m just thinking, like.
308 00:38:02.650 ⇒ 00:38:15.159 Billy Cheng: what does it take for this to happen, rather than still relying on us to, I don’t know, export the data, import back to some sort of AI, like, internal AI tool? I feel like it’s kind of silly.
309 00:38:15.440 ⇒ 00:38:21.380 Billy Cheng: Because you don’t… You need someone from Data Team, because they have access.
310 00:38:21.600 ⇒ 00:38:29.379 Billy Cheng: to, like, run SQL and, like, pull the data, and, like, import back to the GBT tool. That’s… that’s really silly to me.
311 00:38:29.380 ⇒ 00:38:29.960 Robert Tseng: Yeah.
312 00:38:32.760 ⇒ 00:38:36.340 Robert Tseng: Because your internal GPT tool is not connected to your data, is basically what you’re saying.
313 00:38:36.340 ⇒ 00:38:37.980 Billy Cheng: Yeah, I don’t think so.
314 00:38:37.980 ⇒ 00:38:38.350 Robert Tseng: It’s just…
315 00:38:38.350 ⇒ 00:38:38.840 Billy Cheng: I’m getting closer.
316 00:38:38.840 ⇒ 00:38:44.720 Robert Tseng: loop LLM with, like, your own documents, but, like, you just… you have to still, like, kind of drag and drop, like, the pieces that you want to.
317 00:38:44.720 ⇒ 00:38:49.690 Billy Cheng: It’s a closed loop, Gemini, or whatever other tools that they’re connecting to, yeah.
318 00:38:49.690 ⇒ 00:38:50.260 Robert Tseng: Yeah.
319 00:38:51.480 ⇒ 00:38:56.150 Billy Cheng: Hmm… but yeah, I don’t know if you have another meeting,
320 00:38:56.380 ⇒ 00:39:05.200 Billy Cheng: I’m basically done for the day, sorry, the conversation, we just kind of start talking about data stuff, which is very interesting.
321 00:39:05.200 ⇒ 00:39:06.630 Robert Tseng: I’m glad you’re interested.
322 00:39:06.630 ⇒ 00:39:11.669 Billy Cheng: I haven’t asked you more about your life, or something like that, I just get too excited for what you.
323 00:39:11.670 ⇒ 00:39:20.040 Robert Tseng: I know, yeah. I do have another call I have to jump to now, but, like, I’ll… we should, we should do another longer… we should do another longer call.
324 00:39:20.040 ⇒ 00:39:21.140 Billy Cheng: Yeah. Yeah.
325 00:39:21.140 ⇒ 00:39:21.920 Robert Tseng: I’ll…
326 00:39:22.350 ⇒ 00:39:29.879 Robert Tseng: Yeah, we’ll share more… I mean, I’m gonna be off next week, so I’m happy to take personal calls, yeah, so I don’t have to be a work call.
327 00:39:29.880 ⇒ 00:39:31.099 Billy Cheng: That’s good.
328 00:39:31.100 ⇒ 00:39:40.709 Robert Tseng: Yeah, so I’ll, we’ll set up another time. I’m a lot more flexible next week, since I’m pretty much just gonna be chillin’ on a beach and stuff, so… Cool.
329 00:39:40.710 ⇒ 00:39:44.890 Billy Cheng: Alright, sounds good. Yeah, in that case, we’ll find another time to catch up, but yeah.
330 00:39:44.890 ⇒ 00:39:50.910 Robert Tseng: Yeah, I’ll give you a different thing, like, I know the link I gave you only has, like, 30 minutes or whatever, we’ll do, like, we’ll set up a longer call. Yeah.
331 00:39:50.910 ⇒ 00:39:56.029 Billy Cheng: Yeah, no problem. Google, I see ya, take care, have fun and balance.
332 00:39:56.030 ⇒ 00:39:57.539 Robert Tseng: Thank you. Bye.