Meeting Title: Eden AI Kick-Off Call Date: 2026-03-20 Meeting participants: Pranav Narahari, Robert Tseng, Adam P, Daniel
WEBVTT
1 00:00:18.960 ⇒ 00:00:20.240 Pranav Narahari: Oh, hello!
2 00:00:23.390 ⇒ 00:00:24.170 Robert Tseng: Hey, dude.
3 00:01:05.170 ⇒ 00:01:06.250 Robert Tseng: Hey, Adam?
4 00:01:07.930 ⇒ 00:01:08.780 Adam P: Hey, guys.
5 00:01:10.310 ⇒ 00:01:11.879 Pranav Narahari: How’s it going, Adam? Nice to meet you.
6 00:01:11.880 ⇒ 00:01:19.649 Adam P: He was… he was well. Yeah, Danny, I was talking to Danny earlier, and he’s like, you want to join this one? I said, sure, so I’m here to support Danny.
7 00:01:19.830 ⇒ 00:01:20.530 Robert Tseng: Right.
8 00:01:21.090 ⇒ 00:01:21.730 Pranav Narahari: Nice.
9 00:01:22.770 ⇒ 00:01:25.409 Robert Tseng: Is he coming? We’re gonna wait for him to join?
10 00:01:25.410 ⇒ 00:01:30.950 Adam P: Yeah, yeah, he’s… he’s not… he’s always busy. He’s a busy dude. He’d probably run… his last meeting’s probably running over.
11 00:01:31.190 ⇒ 00:01:32.320 Robert Tseng: Okay, cool.
12 00:01:32.760 ⇒ 00:01:34.830 Robert Tseng: Traffic in Hangouts for a bit.
13 00:01:34.830 ⇒ 00:01:35.420 Adam P: Yeah.
14 00:01:40.360 ⇒ 00:01:43.430 Robert Tseng: So, Adam, do you do… you like IT now? Like, I don’t understand your role.
15 00:01:43.430 ⇒ 00:01:44.090 Adam P: Yeah.
16 00:01:44.090 ⇒ 00:01:47.660 Robert Tseng: from the first time I talked to you, which was probably a couple months ago.
17 00:01:47.660 ⇒ 00:02:00.800 Adam P: bounced around a bunch, but now I’m kind of in this, like… yeah, I guess we call it IT. It’s not… we don’t really do IT, right, because we’re a fully remote company kind of deal, but any sort of IT realm thing is kind of getting thrown in my lap now.
18 00:02:01.120 ⇒ 00:02:09.089 Robert Tseng: Okay, so when we’re, like, gonna ask for, like, access to things, that would pretty much be through you. I know you reset my, my Google account, Chris. Okay, cool.
19 00:02:09.259 ⇒ 00:02:14.059 Adam P: Yep, yep, any access, any sort of, yeah, integration pieces and stuff like that, yeah, just…
20 00:02:14.060 ⇒ 00:02:14.680 Robert Tseng: Sure.
21 00:02:14.680 ⇒ 00:02:16.360 Adam P: Feel free to filter through me first.
22 00:02:16.620 ⇒ 00:02:18.529 Robert Tseng: Okay, that is good to know.
23 00:02:18.920 ⇒ 00:02:19.470 Pranav Narahari: Yeah.
24 00:02:20.480 ⇒ 00:02:22.150 Robert Tseng: Ayyy!
25 00:02:22.150 ⇒ 00:02:23.520 Daniel: What’s up, gang gang?
26 00:02:25.930 ⇒ 00:02:27.320 Daniel: Happy Friday!
27 00:02:28.020 ⇒ 00:02:29.670 Daniel: How are you guys?
28 00:02:29.890 ⇒ 00:02:30.690 Robert Tseng: Good.
29 00:02:31.210 ⇒ 00:02:36.199 Robert Tseng: I didn’t know there were, Eden… Eden Sports tees.
30 00:02:36.510 ⇒ 00:02:38.400 Daniel: Yeah, man, we got some cool shit.
31 00:02:38.400 ⇒ 00:02:41.840 Robert Tseng: Is that, like, Lulu, or Under Armour, or what is that?
32 00:02:42.190 ⇒ 00:02:50.509 Daniel: I think these versions are Lululemon, but now we’re partnered with that, I keep calling them Kensei… Quince Company?
33 00:02:50.980 ⇒ 00:02:53.629 Robert Tseng: Oh, okay. Nice.
34 00:02:53.630 ⇒ 00:02:54.920 Daniel: Got some cool new gear.
35 00:02:55.840 ⇒ 00:02:56.640 Robert Tseng: Yeah.
36 00:02:57.100 ⇒ 00:03:03.139 Daniel: Okay, context here. Kind of dropping Adam… Adam Poma. Adam P.
37 00:03:03.140 ⇒ 00:03:04.289 Adam P: Yeah, other Adam.
38 00:03:04.480 ⇒ 00:03:05.500 Daniel: Another Adam, I don’t know.
39 00:03:06.150 ⇒ 00:03:13.879 Daniel: kind of nomenclature there. For a couple of reasons. So Adam’s basically taken over most of our… what do you call them?
40 00:03:14.070 ⇒ 00:03:15.040 Daniel: Man, babe.
41 00:03:15.720 ⇒ 00:03:16.960 Daniel: technology infrastructure…
42 00:03:16.960 ⇒ 00:03:17.410 Adam P: We’re just talking.
43 00:03:17.410 ⇒ 00:03:19.300 Robert Tseng: I knew about it before you came on, yeah.
44 00:03:19.300 ⇒ 00:03:24.539 Daniel: Yeah, okay, cool. So I realize that part of this should probably be relevant,
45 00:03:24.600 ⇒ 00:03:41.679 Daniel: Adam may be one of the champions in rolling out some of these opportunities for our staff. Great. And so, wanted to get him in the meeting here. Background context for Adam, we’ve been working with Brainforge around a concept to try and expose
46 00:03:41.880 ⇒ 00:03:43.840 Daniel: A higher level of…
47 00:03:44.010 ⇒ 00:03:59.670 Daniel: database access, through Gemini, particularly. We’re already in Google Workspace. Gemini’s not my most favorite tool, but Google does a damn good job of putting together workspace integrations together, and it provides a lot of functionality with access between teams to…
48 00:03:59.750 ⇒ 00:04:07.009 Daniel: calendar sharing, to pulling up old Slack notes, to summarizing what I did this week’s kind of, kind of access points.
49 00:04:07.110 ⇒ 00:04:20.450 Daniel: And so Brainforge has proposed a plan to try and up-level our ability to use Gemini as sort of a command center, and get more visibility on the leadership side, more functionality for our actual teams.
50 00:04:20.600 ⇒ 00:04:22.360 Daniel: So kind of twofold there.
51 00:04:22.720 ⇒ 00:04:26.360 Daniel: And yeah, we’re kicking off sort of early high level here.
52 00:04:27.070 ⇒ 00:04:28.080 Adam P: Heck yeah.
53 00:04:28.830 ⇒ 00:04:29.570 Adam P: I don’t know, man, I’m.
54 00:04:29.570 ⇒ 00:04:30.230 Robert Tseng: Yeah.
55 00:04:30.480 ⇒ 00:04:34.010 Adam P: the little bit I’ve used of Gemini, I’m curious…
56 00:04:34.390 ⇒ 00:04:38.470 Adam P: what… what are the… what are… what can it do that I don’t know it can do, sort of thing.
57 00:04:38.640 ⇒ 00:04:40.330 Adam P: That’s where I’m at.
58 00:04:40.980 ⇒ 00:04:52.880 Daniel: So, there’s a lot there, right? And Adam, I think you and I will have to explore, sort of, the actual use case functionality and workflow. Right. A little deep dive on… on…
59 00:04:52.880 ⇒ 00:05:01.929 Daniel: what… but all of it starts with the same thing, which is, if you don’t have the information living in your Google Workspace repository, Gemini won’t be able to seamlessly access it.
60 00:05:02.460 ⇒ 00:05:12.009 Daniel: So if we can explore all the art of the possible in terms of how to use these tools, we need to make sure the tools actually have access to the way we work, which includes things from
61 00:05:12.010 ⇒ 00:05:26.019 Daniel: getting Slack in there to, you know, ultimately getting stuff like Paycom in there. Like, all of that’ll have to be integrated at data later. So that’s what Brainforge is proposing now, and then we’re gonna go through, basically, a huddle process on, okay.
62 00:05:26.030 ⇒ 00:05:34.070 Daniel: What can we do with Danny’s Gemini access now, and is this something we want to scale? How do we change stuff, and kind of go from there in a fluid motion?
63 00:05:34.410 ⇒ 00:05:35.839 Adam P: Yeah, yeah, makes sense.
64 00:05:37.510 ⇒ 00:05:38.960 Daniel: Is that fair at Brainforge?
65 00:05:39.200 ⇒ 00:05:48.459 Robert Tseng: Yeah, yeah, so, I mean, I put together a couple slides, I mean, just… they’re just, like, milestones. I know we’re, like, really just kind of… we’re gonna keep it a little bit more flexible than we originally scoped out.
66 00:05:48.460 ⇒ 00:06:12.470 Robert Tseng: Like, I think we’ve already kind of talked kind of at a high… at the highest level that we can for kind of the different phases, but, yeah, I mean, we’ve kind of basically split it up into two phases. One is just getting all the integrations, making sure that we’ve listed everything that we want to get connected in this first kind of sprint, really. We may not get everything, like, Paycom, like, may not be in the first one, but, like, Caliber Slack, Gmail Drive will be, and so I’ll have to not run through the list that we have here.
67 00:06:12.520 ⇒ 00:06:14.610 Robert Tseng: And then, you know, we talked about
68 00:06:14.690 ⇒ 00:06:33.459 Robert Tseng: being what we want to be able to include, you know, some historical backfill of data, so, like, to what extent is it… is it talking one month back, two months back, you know, whatever that is. So I’ll let him kind of gather the requirements that he needs to kind of start… basically start, ASAP. And then, yeah, I mean, if…
69 00:06:33.460 ⇒ 00:06:39.040 Robert Tseng: We talked about… You know, the current workflows and Adam’s question of, like.
70 00:06:39.300 ⇒ 00:06:50.590 Robert Tseng: what can it do that we don’t already do, or, like, what do we know? Like, I think that’s a… that’s a great question, and let’s start with, like, what it does for you already, and then we can… we can kind of, like, build off of that.
71 00:06:50.590 ⇒ 00:07:04.410 Robert Tseng: Danny already gave us a few ideas of what his wish list would be, but I think, you know, if we’re gonna spend time with you, kind of, as the champion for this project, and you use it similarly to how Danny envisions, then maybe we can just really get, like, the cool workflows from you as well.
72 00:07:05.330 ⇒ 00:07:16.310 Adam P: Yeah, makes sense. Yeah, and my use… I’ve… I basically treated it kind of like a chat GPT, you know, like, here’s a query, tell me, give me a response. I’ve never used it for any sort of…
73 00:07:16.510 ⇒ 00:07:21.220 Adam P: literal workflows yet, like, do this for me, and nothing like that yet.
74 00:07:21.720 ⇒ 00:07:33.329 Robert Tseng: Yeah. Okay, well, as Pranav is kind of pulling up the, the deck, I’ll let you go through it, but, you know, just to kind of give you a little hint of, like, where you should be thinking about this, how you should be thinking about this.
75 00:07:33.330 ⇒ 00:07:42.320 Robert Tseng: you know, at Brainforge, like, at this point, we are using AI to generate all the tickets that would come out of the transcripts of calls like this.
76 00:07:42.320 ⇒ 00:07:57.389 Robert Tseng: We have, like, an AI agent that’s reviewing the tickets, making sure that they’re fully, like, they have fully rich context. We’re even assigning, like, almost 20% of our tickets to AI agents to run on cloud runners. And so, you know, I think you can just think about it, like.
77 00:07:57.680 ⇒ 00:08:09.580 Robert Tseng: you know, I think there’s a lot more than just kind of interacting with a singular chat interface, and we want to be able to kind of introduce those capabilities to you, knowing that this is kind of the starting point.
78 00:08:09.820 ⇒ 00:08:11.070 Adam P: Yeah. Yeah.
79 00:08:11.070 ⇒ 00:08:11.570 Pranav Narahari: Yeah.
80 00:08:12.970 ⇒ 00:08:13.810 Pranav Narahari: Cool.
81 00:08:13.810 ⇒ 00:08:14.619 Daniel: Take it away, bro.
82 00:08:15.430 ⇒ 00:08:35.150 Pranav Narahari: Perfect, perfect. Yeah, so, Danny, what we talked about kind of in that first conversation, I think it was last week or something like that, was, how can we scale this slowly? You know, and we were talking about, for you personally, like, how you’re already kind of maximizing your usage of, Gemini, and it’s, you know, it’s querying from Calendar, it’s querying from
83 00:08:35.150 ⇒ 00:08:53.899 Pranav Narahari: Google Drive, like, that whole workspace. But we saw, like, some things that were not fully fleshed out, and they’re… you’re kind of creeping towards the outer bounds of, like, what that looks like. And so that’s where, like, these first data milestones come into play, because this is where we pull in all the data from all of the company.
84 00:08:53.900 ⇒ 00:08:57.080 Pranav Narahari: So, like, interactions maybe that you are not
85 00:08:57.150 ⇒ 00:09:03.820 Pranav Narahari: fully, like, looped into. Specifically, I remember talking about Slack, and so with Slack, like.
86 00:09:04.140 ⇒ 00:09:13.320 Pranav Narahari: it’s great if you can find the themes of the conversations and the channels that you’re a part of, but what about the private DMs? What about the other…
87 00:09:13.340 ⇒ 00:09:22.550 Pranav Narahari: channels that you are not looped into, the specific threads that you are not in, like, part of, not, you know, added on. And so, building this…
88 00:09:22.560 ⇒ 00:09:40.020 Pranav Narahari: for the… for the first phase, yeah, kind of getting the connection with Calendar, Slack, Gmail, Google Drive, all these different three places, it’s not going to be just for you, even though we’ll… we’ll pilot with just you. We’ll build the infrastructure in a way so that it will scale to everybody across Eden, so then
89 00:09:40.040 ⇒ 00:09:47.370 Pranav Narahari: when we actually, like, give you these themes, it’s not gonna just be, like, when you’re in the loop, it’s when you’re not in the loop, which I think is gonna be the most valuable.
90 00:09:49.580 ⇒ 00:09:50.810 Pranav Narahari: Yeah.
91 00:09:50.810 ⇒ 00:10:02.679 Daniel: So, and Adam, just to… just for context in there, the first use case to kind of delivered to them is, I want Gemini to be able to drop into our workspace with our Slack connection and tell me
92 00:10:02.970 ⇒ 00:10:05.450 Daniel: The name of a project, name it.
93 00:10:05.600 ⇒ 00:10:18.680 Daniel: oh, they’re working on, removing doctor ID photo, right? Recognize that from context, and then tell me the status of that project. The last communications were currently… this is with Ryan in terms of discussions with Beluga, right?
94 00:10:18.860 ⇒ 00:10:27.979 Daniel: Right. So not only will it be able to help categorize a branded project, right, okay, we can see documents related to Eden Pharmacy, you know, kickoff.
95 00:10:28.170 ⇒ 00:10:39.359 Daniel: Phase 1, or whatever, but it’ll also spot themes and recognize workflow motions through that context, and give me a picture of what the team is working on and where it’s at.
96 00:10:40.180 ⇒ 00:10:40.800 Adam P: Wow, yeah.
97 00:10:40.800 ⇒ 00:10:45.029 Daniel: or do that and expose a private layer so that I can see, okay.
98 00:10:45.100 ⇒ 00:11:02.430 Daniel: Adam, Poma, and Mitesh have a side conversation related to the onboarding of this, you know, other marketing team, but it’ll recognize that, anonymize it, won’t tell me, Adam said this last in the chat, it’ll say, so far, it’s now with Adam in discussions with getting them access to workspace credentials.
99 00:11:02.530 ⇒ 00:11:03.430 Adam P: Right.
100 00:11:04.050 ⇒ 00:11:13.910 Adam P: Yeah, I’d be curious to… so, I mean, I know it’s the very first slide or whatever, but, you know, from the tech, IT, architecture side of things, like, how does it
101 00:11:13.910 ⇒ 00:11:25.599 Adam P: dude, like, where does the data get extrapolated from, and how… where does it store in transit, sort of, to, like, munch through, and, like, what is… I will get there, I’m sure, but that’s, like, where my mind’s headed.
102 00:11:25.600 ⇒ 00:11:38.869 Pranav Narahari: Yeah, so I actually did want to touch upon that on this. So, the first milestone here is kind of the data connections, so the different sources, and, setting up the fields in such a way that we’re pulling in the relevant information, and then
103 00:11:38.870 ⇒ 00:11:53.549 Pranav Narahari: not pulling in the information that we don’t want to assess themes on top of. But the part that you just talked about is like, okay, where is this data living? Where, then, can we create analysis on top of? So that’s where this data ingestion and backfill comes into play.
104 00:11:53.550 ⇒ 00:12:06.999 Pranav Narahari: And so this is where we… we have, like, a warehouse of the data, and then in the future, not gonna be just Danny’s data, it’s gonna be across the company. And so that’s really where you need to have a data warehouse.
105 00:12:07.460 ⇒ 00:12:15.379 Pranav Narahari: to really just organize all the different pieces of data, and then on top of that, do the analysis. And then one thing, because, you know.
106 00:12:15.490 ⇒ 00:12:18.180 Pranav Narahari: With, the nature of the data that you guys maybe
107 00:12:18.220 ⇒ 00:12:31.409 Pranav Narahari: working with via Slack specifically, if there’s certain PII data in there, Google Workspace is great because they have, like, a way to redact some of this, PII stuff, and it’s built in.
108 00:12:31.410 ⇒ 00:12:42.390 Pranav Narahari: Now, we will need to do a little bit of a spike to see, like, are they extracting the exact data that you guys want to be, like, redacted? And so, that’s kind of what this piece is right here.
109 00:12:42.790 ⇒ 00:12:55.159 Pranav Narahari: But then at that point, Adam, to, like, to your question, like, that’s when we have that data layer figured out. So everything that we need as context to make the insights on, that will be completed after this Phase 1.
110 00:12:55.670 ⇒ 00:13:00.019 Adam P: Gotcha. Yeah, starting in GCP, that makes sense. And, like…
111 00:13:00.300 ⇒ 00:13:03.399 Adam P: you know, format style. Does Gemini have a, sort of.
112 00:13:03.590 ⇒ 00:13:10.900 Adam P: SOP of sorts of, hey, you should store data like this, or is that all part of this design that is figuring out how to
113 00:13:11.060 ⇒ 00:13:13.049 Adam P: Architect the data itself.
114 00:13:13.660 ⇒ 00:13:14.310 Pranav Narahari: Yeah.
115 00:13:14.310 ⇒ 00:13:15.890 Robert Tseng: I need to give it Yeah.
116 00:13:15.890 ⇒ 00:13:16.360 Adam P: Yeah, okay.
117 00:13:16.360 ⇒ 00:13:17.200 Pranav Narahari: Yes.
118 00:13:17.200 ⇒ 00:13:37.969 Daniel: So, for example, we have some built-out functionality. Look, Workspace is not, like… like, I would build this thing out in Claude if it was just a general basis, right? But we have a situation where we have a BAA with Google, already maintaining our HIPAA compliance requirements, regardless of what data is piped in, right? That already exists with our BQ functionality and whatever else. So, the BAA is an extreme component there.
119 00:13:37.970 ⇒ 00:13:48.849 Daniel: The second thing is, this is already native integrations. We know the calendar integration isn’t gonna break, or our email, or workspace credentials. The third thing is, it layers over the credentialing from our drive.
120 00:13:49.010 ⇒ 00:14:08.829 Daniel: So, employees only have access to the drive files that they have access to. So, if we say, accesses to everybody, everybody will have access to that drive file. But we can have a separate credentialed legal folder that is only accessible by people with access to the legal folder, and that credentialing is already native in Google Drive.
121 00:14:09.820 ⇒ 00:14:10.500 Adam P: Right.
122 00:14:10.500 ⇒ 00:14:11.160 Pranav Narahari: Yes.
123 00:14:11.160 ⇒ 00:14:14.840 Daniel: Jonas in the background, working with Jared’s team, to
124 00:14:14.970 ⇒ 00:14:25.199 Daniel: identify and categorize legal, HR, you know, marketing into folders that allow us to give credentialed access and a natural function in this.
125 00:14:27.430 ⇒ 00:14:28.320 Adam P: Right, right.
126 00:14:30.120 ⇒ 00:14:30.940 Adam P: Yeah, sweet.
127 00:14:32.490 ⇒ 00:14:33.339 Pranav Narahari: Yeah, I think…
128 00:14:33.340 ⇒ 00:14:33.890 Robert Tseng: Thank you.
129 00:14:33.930 ⇒ 00:14:44.499 Robert Tseng: Yeah. Well, if I can say one more thing, so, like, as far as, like, estimated hours or whatever, like, even these are just kind of, you know, we have a certain pace that we’ve kind of agreed to that we’re going to be working at, so we’ve kind of split it up this way.
130 00:14:44.500 ⇒ 00:14:56.060 Robert Tseng: So on the ingestion backfill, I want to kind of be on the technical requirements, even, like, Danny’s example of, like, one, he kind of… the way you framed it was, like, just looking in Slack specifically, getting all the context out of the Slack messages.
131 00:14:56.060 ⇒ 00:15:00.039 Robert Tseng: Well, I just want to kind of give you perspective on, like, internally for us, when we do it.
132 00:15:00.040 ⇒ 00:15:07.240 Robert Tseng: when I’m, like, querying something that’s, like, the theme-based from Slack, and I want to get context from, like, a particular thread.
133 00:15:07.240 ⇒ 00:15:31.689 Robert Tseng: not only am I looking at that, I’m hap… you know, we have our system set up where it’s going to go into, kind of, the master kind of project file, or whatever. So it’ll… there’s some mapping, some source of truth that kind of keeps things organized, because, sometimes, like, you know, Slack threads itself, like, people have different biases in the way they talk about a project, and, like, that can get the model, confused. And so, that’s kind of, like, in function, like a staging layer, where
134 00:15:31.690 ⇒ 00:15:46.619 Robert Tseng: in the data world, we basically have these different staging models already for your transactional data that we built for your data warehouse. But also, like, you just need to pull context from different places. So, if it is, like, whatever project A that Daniel… that Danny was specifying.
135 00:15:46.620 ⇒ 00:16:10.080 Robert Tseng: That people are talking about. You can also go and check through your calendar to see how many meetings have been booked, or have already taken place regarding that. So if, like, Slack is telling you people have been discussing this project for, like, I don’t know, like, only 5 days, and they’ve only met twice, there’s, like, you’re also checking that against, like, Google Calendar to make sure, like, hey, like, actually, maybe they met 5 times over the course of 2 or 3 weeks. So.
136 00:16:10.080 ⇒ 00:16:33.900 Robert Tseng: these different sources start to have, like, different layers of redundancy, and you need to be able to, like, kind of do that… to instruct the model to do the matching across… these are multiple agents running in the background at this point, making sure that it’s all kind of coming up with a uniform story that’s not limited to just the Slack context. I’m assuming that if you do have an MCP set up for Slack to do G drive immediately right now.
137 00:16:33.900 ⇒ 00:16:52.549 Robert Tseng: and you’re asking a question in Slack, it’s only asking it in Slack, and it’s not really doing this kind of matching thing that we’re describing out of the box. So, you know, these are some of the functionality that we have to build in, which is kind of really gonna be… it’s all… this is all the background stuff that’s happening in this first phase.
138 00:16:53.640 ⇒ 00:16:55.620 Robert Tseng: Does that make sense? Yeah, okay.
139 00:16:56.180 ⇒ 00:17:01.220 Daniel: And just to gut check me, I mean, seems cool.
140 00:17:01.450 ⇒ 00:17:05.940 Daniel: From, like, blue sky thinking, yeah. Is this gonna be, like, maintainable?
141 00:17:08.250 ⇒ 00:17:14.929 Daniel: are we just gonna build a monster here that, like, is gonna false signal and nobody uses it? Like, those are some of the things I’m thinking in my head.
142 00:17:14.930 ⇒ 00:17:15.290 Pranav Narahari: Yeah.
143 00:17:15.380 ⇒ 00:17:16.500 Daniel: That’s myself.
144 00:17:17.319 ⇒ 00:17:35.029 Pranav Narahari: Well, I’m actually really glad you asked that, because I think that’s the core point of this project. It’s like, how do we make themes that are actually actionable, and it’s not just noise, right? And so, after we get this data layer figured out, the next part is actually extracting the themes and what we call as clustering within the data.
145 00:17:35.029 ⇒ 00:17:45.929 Pranav Narahari: And so, yeah, maybe this is just… I should hop into the next slide, but that’s the very next thing that we start working on. And so it’s extracting the themes across all these different data sources.
146 00:17:45.929 ⇒ 00:18:01.329 Pranav Narahari: And then what we want to do in the beginning is get you guys into the loop, too, of what we’re seeing in the data, because what’s going to happen is that there is going to be some noise. And so, how do we differentiate the noise, from what is actual themes that can be actionable?
147 00:18:01.399 ⇒ 00:18:03.459 Pranav Narahari: And so…
148 00:18:03.909 ⇒ 00:18:18.189 Pranav Narahari: that is, like, what Robert was talking about, like, you know, not just focusing on one source, but, like, painting an entire picture, like, a unified, like, this is actually what’s going on, because we see this happening in Slack, we see this happening on calendars, etc.
149 00:18:18.249 ⇒ 00:18:27.119 Pranav Narahari: This is where that analysis happens, and then this is where the initial report is where… this is the milestone where we’ll provide that initial report.
150 00:18:29.760 ⇒ 00:18:47.659 Adam P: Yeah, that’s… in my head, so, like, the example you gave, Danny, where you’re, you know, get a holistic view of how Project A is kind of going, like, that use case is awesome for you. I’m curious what use cases there are that we can identify for, like, I don’t know.
151 00:18:47.750 ⇒ 00:18:52.589 Adam P: care team member A, you know? Like, what would they utilize something like this for? Or even, like.
152 00:18:52.840 ⇒ 00:18:56.249 Adam P: The marketing team, like, how would they leverage this sort of thing?
153 00:18:56.470 ⇒ 00:18:58.569 Adam P: Identifying what those use cases are.
154 00:18:58.570 ⇒ 00:19:01.029 Daniel: Thought a couple of use cases. Number one is…
155 00:19:01.240 ⇒ 00:19:06.929 Daniel: we aren’t going to be able to functionally touch workflow, necessarily, through this, right? So this is…
156 00:19:07.040 ⇒ 00:19:26.330 Daniel: Josh has a ton of initiatives right now on building out, like, Claude agent interactions for functional workflow. How could we have Zendesk do preemption and pre-type responses based on KBs? Zendesk has integrations with some Claudevax tools that do that. This is not about functional workflow. The pharmacy will have a
157 00:19:26.330 ⇒ 00:19:29.360 Daniel: functional workflow that’s built out across Pharmetica, right?
158 00:19:29.360 ⇒ 00:19:29.970 Adam P: Huh.
159 00:19:29.970 ⇒ 00:19:33.180 Daniel: The AI strategy here is internal workflow.
160 00:19:33.590 ⇒ 00:19:51.789 Daniel: So you could, for example, Adam, ask, what are the termination requirements with our contract with B-side marketing, or whatever, right? It would pull from a database, if you’re credentialed to see that contract, it will tell you, it will parse out and show the termination provision written in that contract in Gemini.
161 00:19:52.120 ⇒ 00:19:56.440 Daniel: You can ask it, can you set… can you… can you make sure that,
162 00:19:56.600 ⇒ 00:20:07.190 Daniel: Can you set up a Friday weekly with Mitesh and I to go over this? Boom, done. It’ll have access to calendar and be able to do some of those agenda flows. That’s not the data layer, that’s more functionality discussion.
163 00:20:07.890 ⇒ 00:20:09.400 Daniel: Some of the other things we talked about.
164 00:20:09.940 ⇒ 00:20:17.349 Daniel: as a user, you’ll get access to basically Eden’s tools in one place, and access to the databases.
165 00:20:17.430 ⇒ 00:20:26.630 Daniel: as a management team, we’ll be able to go see discovery of major themes. How much is Eden Pharmacy working with Eden Technologies?
166 00:20:26.660 ⇒ 00:20:44.409 Daniel: If it has context, that’ll say, these people are in Eden Technologies, these people are in Eden Pharmacy, there’s an interaction workflow there, right? You could spot a project. We could have weekly pulse reports, which is one of the coolest things I want as a COO, to see what is overall productivity and where are my teams allocating their time and resources.
167 00:20:44.410 ⇒ 00:20:47.670 Daniel: That gives high-level bird’s-eye views of what’s going on.
168 00:20:47.930 ⇒ 00:20:57.930 Daniel: So, there are some… this is very internal-focused, but could significantly unlock where we may need to install direct workflow in-production AI tools, and that.
169 00:20:57.930 ⇒ 00:20:59.629 Adam P: Yeah, that makes sense.
170 00:20:59.900 ⇒ 00:21:01.299 Adam P: I get it. Cool.
171 00:21:03.150 ⇒ 00:21:04.139 Pranav Narahari: And so…
172 00:21:04.410 ⇒ 00:21:18.179 Pranav Narahari: That makes total sense to me, too, and it’s like, with these insights, it’s then, okay, unlocking the workflow, like you just said, Danny, like, to do the actual action can happen next, too. But yeah, this is the initial, kind of.
173 00:21:18.300 ⇒ 00:21:23.589 Pranav Narahari: extraction of… Okay, and one thing I wanted to mention, too, is, like, with
174 00:21:23.610 ⇒ 00:21:33.239 Pranav Narahari: you guys are gonna have a certain understanding of the organization that we’re gonna have to… we’re gonna need to learn. And so, with this theme, Discovery Baseline, like, Delivered Milestone.
175 00:21:33.240 ⇒ 00:21:43.670 Pranav Narahari: That’s where we get more of the inner workings of, like, what are the clusterings that you guys have already recognized? So then we can then bake that into the analysis that we drive.
176 00:21:43.730 ⇒ 00:21:45.660 Pranav Narahari: And so…
177 00:21:46.090 ⇒ 00:21:59.670 Pranav Narahari: this is kind of… this is really the meat of the project, in terms of, like, where the real value add is. The rest of it is, like, we’re building around it so we can actually drive the insights, but this is where, like, the real…
178 00:21:59.790 ⇒ 00:22:07.300 Pranav Narahari: the real analysis out of this. And then, same with just this part two, like, I wanted to split this up into two, just because
179 00:22:07.300 ⇒ 00:22:23.979 Pranav Narahari: I need there to be, like, a formalized, like, and how I’ve seen this be successful in the past, is if there’s a formal, like, hey, we present something to you guys, and you guys let us know where things are good, where things are just noise. And so, if we can bake that in, and just, like, call it, like, a real milestone, I think it’ll help us out.
180 00:22:24.140 ⇒ 00:22:29.170 Pranav Narahari: And then with, bandwidth friction signals, this is, like.
181 00:22:29.640 ⇒ 00:22:41.850 Pranav Narahari: how do we actually turn it into something actionable, or an actual, like, alerting mechanism? And so, we’ll work together to talk about how do we actually want to… how do we portray, or…
182 00:22:42.150 ⇒ 00:22:49.819 Pranav Narahari: how do we describe these clusterings into a way where… into a… into a format that actually makes sense to you guys? So…
183 00:22:50.130 ⇒ 00:23:02.860 Pranav Narahari: Whether it’s a dashboard, whether it’s, like, certain, like, like, five… top five, like, things to take a look at, these are different things that we can, like, discuss as well.
184 00:23:03.940 ⇒ 00:23:15.639 Pranav Narahari: And then, yeah, you mentioned, like, the weekly pulse report, too. Like, so this, why we need to have a data warehouse is because we’re gonna be consolidating weekly information, and so to…
185 00:23:15.670 ⇒ 00:23:24.679 Pranav Narahari: we can’t just do, like, one query on this. We could maybe do it for just you, Danny, but then when it comes to the point of, like, the entire organization,
186 00:23:24.850 ⇒ 00:23:30.129 Pranav Narahari: It’s just not a really… like, you talked about, is this gonna be scalable? Is this gonna be a mess after we get to a certain point?
187 00:23:30.350 ⇒ 00:23:42.190 Pranav Narahari: the infrastructure we’re talking about is… we’re building it in a way where it’s not going to be a mess, because we’re… we’re seeing it down the line where it could be utilized for everybody, not just for you.
188 00:23:42.770 ⇒ 00:23:48.609 Pranav Narahari: And then, yeah, so at this point, too, like, we’re generating a report, we want to create some refinements on that.
189 00:23:49.930 ⇒ 00:24:07.349 Pranav Narahari: and then just kind of having some documentation to just, like, finalize what we see, like, this deliverable being. Like Robert mentioned in the beginning, too, like, we want to be a little bit more flexible. I think that’s what everybody kind of on this call, like, we’ve decided to be at.
190 00:24:07.820 ⇒ 00:24:14.769 Pranav Narahari: But this is kind of, like, the roadmap that we set out, and we can kind of go forward and, like, make any changes as necessary.
191 00:24:18.830 ⇒ 00:24:24.940 Daniel: I’ve got a gut feeling here that… Getting the tools in.
192 00:24:26.450 ⇒ 00:24:30.979 Daniel: and just trying to see what Gemini does with it, I know that’s… that’s unplanned.
193 00:24:31.760 ⇒ 00:24:38.199 Daniel: But, like, it might do some interesting things that help us understand what this workflow could look like, right?
194 00:24:38.200 ⇒ 00:24:38.860 Robert Tseng: Yeah.
195 00:24:39.130 ⇒ 00:24:46.969 Daniel: So it’s hard for me to tell, like… I also don’t want to sell this short, either. I don’t want to, you know, just say, oh, it’s about the weekly post report. I was just trying to find some, like…
196 00:24:47.220 ⇒ 00:24:50.580 Daniel: Future state visualizations that we could drive towards?
197 00:24:51.240 ⇒ 00:24:51.620 Robert Tseng: Yeah.
198 00:24:51.700 ⇒ 00:25:03.319 Daniel: The two were, like, project tracking and this weekly poll support. Like, to me, that would be, okay, that’s kind of capstone on this. It saw the context, saw the visibility, and was able to deliver insights that are actionable, right?
199 00:25:03.520 ⇒ 00:25:11.160 Robert Tseng: Yeah. No, I think capstone is a good way of framing it. It’s like, when you work on a capstone project in academia, like, there’s a lot of other sub…
200 00:25:11.160 ⇒ 00:25:34.370 Robert Tseng: like, side quests that you take to get there. So, like, and maybe the side quests become more valuable than the capstone itself, right? But I think, like, just kind of giving us something to aim at, like, just allows us to, like, have a line of sight of, like, where… where does this iteration stop, and it’s, like, good enough for a handoff. So, if I were to just kind of, like, abstract a little bit from these milestones, I think the first two, this is, like, where we’re really fine-tuning things, right?
201 00:25:34.370 ⇒ 00:25:39.299 Robert Tseng: Pranav calls them clusters, we can call them themes, whatever you want, but really what this is, is, like.
202 00:25:39.300 ⇒ 00:25:47.450 Robert Tseng: do we, like, when you talk about internal Eden operations, and what all those different, kind of, like, pockets are, are…
203 00:25:47.450 ⇒ 00:26:12.299 Robert Tseng: is the data being summarized correctly into the same concepts that you want to see, like, reused around the organization? So we can call them whatever we want. They’re really vector embedding clusters, like, that’s engineering-wise, like, that’s what it is. But, like, you know, you guys may have different ways of talking about it, and this will be coming out in your SOPs, we go in, we watch your workflows, we’ll start to kind of piece together what that is, too, very much like how
204 00:26:12.340 ⇒ 00:26:31.599 Robert Tseng: the same exercise we do on the data side, where when we’re talking about, like, pharmacy operations, end-to-end of an order journey, like, it took a while for us to kind of nail each of the different steps for the order, and then, like, kind of build that out into a data model. So, I kind of view this as a similar exercise to what we’re doing, but just, like, for ops.
205 00:26:31.600 ⇒ 00:26:33.249 Robert Tseng: Kind of, like, workflows.
206 00:26:33.250 ⇒ 00:26:46.299 Robert Tseng: And then, like, the bandwidth and friction signals piece, this is, like, the… the format that you want it presented to you. If you want Slack alerts, you want email digest, like, you know, do you want dashboard, visuals, whatever it is, like, that’s… that’s to me, like.
207 00:26:46.300 ⇒ 00:27:10.959 Robert Tseng: just the… just the format that which you will consume the information. Obviously, you can access it through… through chat, but what… if you access it through chat, and you don’t kind of, like, pre-configure this, every time you query AskGemini, it may give you a different format, and that inconsistency may just, like, not… that just doesn’t scale well, because people are still ultimately not looking at the same thing if, you know, which you do want some flexibility, but, you know.
208 00:27:10.960 ⇒ 00:27:12.330 Robert Tseng: That’s why that’s there.
209 00:27:12.430 ⇒ 00:27:28.119 Robert Tseng: And then, yeah, the rest of the stuff is really just, like, a capstone, plus, like, a point where we say, okay, Danny, this is good enough for you to use now. Maybe you introduce it to the rest of the C-suite, or, like, some small pilot of users, and then we can keep building on it, with you after that.
210 00:27:31.450 ⇒ 00:27:34.720 Robert Tseng: Okay, cool. Anything, any other questions on this?
211 00:27:34.940 ⇒ 00:27:50.520 Daniel: No, I think that was a really good summary on this. I don’t want to put too much in the box, but we have to have something that we know is going to be actionable and deliverable in order to demonstrate the use case, right? From there, we can build different tools, or different visualizations, or whatever it may be, so…
212 00:27:50.520 ⇒ 00:28:03.389 Daniel: I love exploring the possibilities. I do also understand the value of narrowing this and saying, okay, if Danny could get a project, context-based project pulse report on, you know, the week of work.
213 00:28:03.430 ⇒ 00:28:06.129 Daniel: That’s a pretty good thing to aim for here.
214 00:28:06.130 ⇒ 00:28:06.730 Robert Tseng: Yeah.
215 00:28:09.180 ⇒ 00:28:13.349 Daniel: So, what do we need to help you guys with to start driving this?
216 00:28:14.140 ⇒ 00:28:21.490 Robert Tseng: Well, I’m assuming you’re already attempting this in your way, like, I either, you know, I know you’re busy, we don’t have to get you on calls, but if you can send
217 00:28:21.490 ⇒ 00:28:46.409 Robert Tseng: recordings, report, direct us to resources, like, I guess my role in this project is really to kind of, like, put my… myself in, like, the Danny shoes, and then, like, learn that as quickly as I can, turn that into engineering requirements, and give it to Bernal, so you can kind of see me as, like, kind of the PM on… on… on this… on this, on this project. So, whether it’s easier for you to show me, or, like, we can just talk through it, I’ll just interview you through, like, a call
218 00:28:46.410 ⇒ 00:28:48.459 Robert Tseng: like, I’m… that could be a starting point.
219 00:28:48.460 ⇒ 00:29:01.040 Daniel: I think… I think probably some of our best starting points would likely be our strategic foundational documents, because we have some things like, you know, Q1 goals, Q2 goals, across these verticals. They may not…
220 00:29:01.380 ⇒ 00:29:16.439 Daniel: encompass all the projects, but at least this could be, like, compared to strategic plan, based on that one meeting that we had with Rebecca on the pharmacy team, they noted that the SLAs are at this level, and our goal is this level, so they’re inbound, right?
221 00:29:16.440 ⇒ 00:29:16.790 Robert Tseng: Yep.
222 00:29:16.790 ⇒ 00:29:22.960 Daniel: Like, like, I think maybe, maybe tying this directly into some of those strategic plans would likely make the most sense.
223 00:29:23.200 ⇒ 00:29:23.870 Robert Tseng: Okay.
224 00:29:24.070 ⇒ 00:29:34.280 Daniel: So, what I can do is share those vertical plans with you. We actually have them, right? So, that’s something I can just share with you. You can kind of get a download, and then let’s have a…
225 00:29:34.750 ⇒ 00:29:43.690 Daniel: a call to go through that after you’ve had a chance to digest, sort of, how those verticals are operating, and look at more of the whole picture, or something, I guess.
226 00:29:43.850 ⇒ 00:29:45.880 Robert Tseng: Okay, yeah, that works.
227 00:29:46.310 ⇒ 00:29:57.969 Daniel: Cool. So, that’s a follow-up I’ll take, is to get you everything I’ve got on those. We have, like, an OKR dashboard, for example, on some things that we’re working with the teams on. That should give us some good context.
228 00:29:58.490 ⇒ 00:29:59.450 Robert Tseng: Okay.
229 00:29:59.530 ⇒ 00:30:10.219 Daniel: And then the other thing that we can sort of use to test this is pick a couple, like, current in-flight projects, and just kind of see what we can capture through it, and see if that.
230 00:30:10.220 ⇒ 00:30:10.690 Robert Tseng: Yeah.
231 00:30:10.690 ⇒ 00:30:18.829 Daniel: So we had one that was, like, we just wrapped up a project for the beluga test ID requirement removal project, right? I’d love.
232 00:30:18.830 ⇒ 00:30:19.190 Robert Tseng: Yeah.
233 00:30:19.190 ⇒ 00:30:33.789 Daniel: to see how Gemini can try and pick apart the context, and describe what happened there, because we know, end-to-end, when it started, when it finished, what the result was, and I’d love to see if we can pick up those contexts. So there’s probably 5 of those use cases we can go through.
234 00:30:34.330 ⇒ 00:30:46.770 Robert Tseng: Okay, yeah, no, those are… those are great, great materials to give us. And then for Nav, obviously, you want to get access to things so you can start to build out these connectors, so, maybe him and Adam can kind of work together on getting that.
235 00:30:47.140 ⇒ 00:30:47.850 Adam P: Oh, yeah.
236 00:30:47.850 ⇒ 00:30:49.360 Pranav Narahari: Definitely. Perfect.
237 00:30:49.540 ⇒ 00:30:53.269 Daniel: Sweet, I’m stoked about this, I think this will be a big unlock, so… Okay.
238 00:30:53.590 ⇒ 00:31:01.559 Robert Tseng: Cool. Cool. Alright, would it be helpful to just set a weekly with you guys, or do you want to do it more ad hoc? Okay.
239 00:31:01.560 ⇒ 00:31:05.069 Daniel: The only wrench I’m gonna throw in that is I’m out basically the next 2 weeks.
240 00:31:05.070 ⇒ 00:31:05.480 Robert Tseng: Okay.
241 00:31:05.480 ⇒ 00:31:06.430 Daniel: I’m a hunting man, so…
242 00:31:06.430 ⇒ 00:31:07.180 Adam P: How convenient.
243 00:31:07.180 ⇒ 00:31:09.789 Daniel: Yeah, we leave on Tuesday, and…
244 00:31:09.790 ⇒ 00:31:11.829 Adam P: Here’s this project. See you later, guys.
245 00:31:11.830 ⇒ 00:31:25.200 Daniel: But what I hope is a lot of this is, like, we’re just implementing the data layer here, so it’s… Yeah. …access, I mean, get all this stuff in, and then we’ll work through, you know, how we can start to draw some signals from it.
246 00:31:25.460 ⇒ 00:31:26.110 Robert Tseng: Sure.
247 00:31:26.270 ⇒ 00:31:41.210 Robert Tseng: Okay, that sounds good. Yeah, so we’ll do ad hoc for now until you’re fully back, but otherwise, we’ll obviously keep sending updates to that channel that we set up, probably add Adam to that channel as well. And if we don’t see you, then I guess, yeah, have a good honeymoon. Where are you headed?
248 00:31:41.490 ⇒ 00:31:43.219 Daniel: We’re going to Turkey and Greece.
249 00:31:43.640 ⇒ 00:31:44.870 Robert Tseng: Oh, nice.
250 00:31:44.870 ⇒ 00:31:45.350 Pranav Narahari: Nice.
251 00:31:45.820 ⇒ 00:31:46.960 Adam P: Way over there.
252 00:31:46.960 ⇒ 00:31:49.919 Robert Tseng: Yeah, hope the flights and everything are smooth.
253 00:31:50.240 ⇒ 00:31:51.220 Robert Tseng: Good luck.
254 00:31:51.220 ⇒ 00:31:51.550 Pranav Narahari: Yeah.
255 00:31:51.550 ⇒ 00:31:55.230 Daniel: We see war in the Middle East, so that’s where we’re gonna fly, so…
256 00:31:55.560 ⇒ 00:31:56.080 Adam P: Right into it.
257 00:31:56.080 ⇒ 00:31:56.570 Robert Tseng: Yeah.
258 00:31:56.910 ⇒ 00:31:57.820 Daniel: Thank y’all.
259 00:31:58.250 ⇒ 00:31:59.659 Robert Tseng: Okay, alright, thanks guys.
260 00:31:59.660 ⇒ 00:32:00.240 Adam P: Thanks, guys.
261 00:32:00.240 ⇒ 00:32:00.860 Pranav Narahari: discuss.