Meeting Title: PP2G | Next phase planning Date: 2025-05-27 Meeting participants: Bencohen, Daniel Schonfeld, Amber Lin
WEBVTT
1 00:01:21.480 ⇒ 00:01:22.730 bencohen: Hey! Amber.
2 00:01:23.830 ⇒ 00:01:25.010 Amber Lin: Hello!
3 00:01:25.440 ⇒ 00:01:26.389 bencohen: How are you?
4 00:01:26.760 ⇒ 00:01:42.349 Amber Lin: I’m pretty good. I came back from Austin, you know. I was staying Atum’s house when he was he was on vacation the last week, so I stayed at his house in Austin. I was. I got the tour around Austin the whole week.
5 00:01:42.350 ⇒ 00:01:43.450 bencohen: Did you like it?
6 00:01:43.630 ⇒ 00:01:49.919 Amber Lin: I like. It’s very hot. I liked it. It was cool. I like la weather.
7 00:01:51.110 ⇒ 00:01:57.830 bencohen: Oh, nothing compares to Los Angeles except your June is kind of weird in La June gloom.
8 00:01:57.830 ⇒ 00:02:06.410 Amber Lin: True that is coming up like the heat is coming up it. It was kind of cold last night, and then sometimes gets really hot in the day, so I.
9 00:02:06.410 ⇒ 00:02:06.920 bencohen: Yeah.
10 00:02:06.920 ⇒ 00:02:09.839 Amber Lin: I have no clue. What about you? Where are you based.
11 00:02:10.229 ⇒ 00:02:11.289 bencohen: Connecticut.
12 00:02:11.290 ⇒ 00:02:12.780 Amber Lin: Connecticut, Wow.
13 00:02:13.730 ⇒ 00:02:18.279 bencohen: Near the city. Dan’s on the way. Utam’s coming right.
14 00:02:19.217 ⇒ 00:02:26.010 Amber Lin: I believe won’t be able to make this meeting. He’s at a lunch with one of our
15 00:02:26.260 ⇒ 00:02:32.810 Amber Lin: partners, I believe, but, he said, I have access to him via text.
16 00:02:33.090 ⇒ 00:02:38.780 Amber Lin: So if there’s anything very, very urgent, I will let him know. But Hi, Dan.
17 00:02:43.560 ⇒ 00:02:44.650 Daniel Schonfeld: Can you hear me?
18 00:02:45.090 ⇒ 00:02:45.579 Amber Lin: Yeah, I mean.
19 00:02:45.580 ⇒ 00:02:46.520 bencohen: Here you go!
20 00:02:46.720 ⇒ 00:02:48.520 Daniel Schonfeld: See me. There we go.
21 00:02:49.520 ⇒ 00:02:50.450 Daniel Schonfeld: Hello!
22 00:02:52.470 ⇒ 00:02:53.380 bencohen: How’s it going.
23 00:02:53.680 ⇒ 00:02:55.580 Daniel Schonfeld: Good ready to rock.
24 00:02:55.580 ⇒ 00:02:56.210 Amber Lin: Game.
25 00:02:56.520 ⇒ 00:02:59.910 Amber Lin: Yeah, I think this 1st time actually seeing your face.
26 00:03:00.030 ⇒ 00:03:00.940 bencohen: He’s handsome.
27 00:03:00.940 ⇒ 00:03:04.250 Amber Lin: Isn’t he got those eyes?
28 00:03:06.420 ⇒ 00:03:09.050 Amber Lin: I bet that’s the 1st thing people notice about you.
29 00:03:09.050 ⇒ 00:03:11.179 Daniel Schonfeld: Yeah, right? Usually, it’s my nose.
30 00:03:13.630 ⇒ 00:03:20.050 Amber Lin: Well and on the video I can’t see any. But both of you gotta look the other way for me to actually see anything.
31 00:03:21.490 ⇒ 00:03:44.909 Amber Lin: Okay, my agenda for today is, I want us to go over. What of the work we’re currently doing? Because there’s the AI stuff. There’s a dashboard stuff. There’s some analysis stuff. So we kind of have a lot going on. And then I wanted to see what you guys feel about the current progress. And if you have anything down the line that you want to continue investigating.
32 00:03:45.630 ⇒ 00:03:57.500 Amber Lin: So if you’re fine with that, I’m gonna start to run you through a few things that we’re working on. And you tell me, hey, do we want to continue? How’s this going? And then we’ll talk about what we want to do? And the next.
33 00:03:57.500 ⇒ 00:03:57.820 bencohen: See.
34 00:03:59.010 ⇒ 00:04:16.750 Amber Lin: Yeah. So 1st of all, we completed the 2 weeks for the AI poc, so we wanted a ui that you can use to test. We wanted to complete it in a short time period of 2 weeks, and we were able to deliver that pretty well, pretty pretty fast
35 00:04:16.810 ⇒ 00:04:35.149 Amber Lin: in that time period, with all the features that we discussed in the document with all the tickets. So we have the web app. We it answers all the questions pretty well, and we have been able to start the internal Qa. So we have
36 00:04:35.180 ⇒ 00:04:52.940 Amber Lin: complete the 1st phase for the aipoc. So if there’s anything moving forward that we want to do. I know we might want image abilities. We might want to shift it to a different domain. Or maybe you want voice capabilities. Like all of that, we can plan out. But I want us to have a
37 00:04:53.300 ⇒ 00:05:17.099 Amber Lin: something set in writing when we move forward. So that’s the 1st thing in the AI poc that we did second part is, you know, we have maintenance of the dashboard, as Ben is more involved with that, especially of the Daily Kpi dashboard. We’re still meeting with Kim daily to make sure that she is able to use it. If there’s any maintenance issues that needs to be done. So there’s work on that side as well.
38 00:05:17.300 ⇒ 00:05:43.060 Amber Lin: And thirdly, we have the analysis that mostly I was doing with helping Robert. Of how do we do the price drops? Has it impacted our sales? What has it? What is working on different channels? And looking at that dashboard, pulling all that data to see. Okay, how have things been doing? I don’t know if you guys had a chance to look at those decks that I created. So
39 00:05:43.180 ⇒ 00:05:48.749 Amber Lin: overall, these are the 3 main points that we’re working on. And I want to see how you guys feel on each of them.
40 00:05:49.970 ⇒ 00:05:59.309 bencohen: Sure. We really want to focus on AI on this call. But I’ll I’ll just just since you asked. The dashboard, I think, is a lot better.
41 00:06:00.200 ⇒ 00:06:04.989 bencohen: I think you guys have done a good job of of tightening it up
42 00:06:05.460 ⇒ 00:06:08.639 bencohen: on the sales stuff. I saw both decks.
43 00:06:08.870 ⇒ 00:06:14.739 bencohen: I mean, I made the decision on on the price drop, whatever it was to
44 00:06:15.200 ⇒ 00:06:17.329 bencohen: maybe almost 3 weeks ago.
45 00:06:19.570 ⇒ 00:06:29.810 bencohen: I think we probably need to dig a little deeper. Because I basically just kind of was like, you know what I think. We we feel $40 too expensive or whatever it was. And we just did it based on that
46 00:06:30.640 ⇒ 00:06:34.190 bencohen: when we were too high before. Okay? So we knew that. But
47 00:06:34.560 ⇒ 00:06:42.200 bencohen: I think we need like a layer deeper, because I don’t really need someone to tell me that we lowered the price $40. And we went up 20% like
48 00:06:42.490 ⇒ 00:06:45.100 bencohen: that. That’s very easy. That’s not.
49 00:06:45.890 ⇒ 00:06:49.189 bencohen: A layman like me can figure that one out. So I think
50 00:06:49.360 ⇒ 00:06:52.519 bencohen: if we’re going to be doing analysis on performance.
51 00:06:53.120 ⇒ 00:06:56.510 bencohen: we need you to find some, some stuff that really can move the needle.
52 00:06:56.610 ⇒ 00:06:59.460 bencohen: I wanted to set that goalpost.
53 00:06:59.590 ⇒ 00:07:03.879 bencohen: that if if we’re going to do it at all, we need to be like looking to
54 00:07:04.140 ⇒ 00:07:06.379 bencohen: find things that are going to be really, really.
55 00:07:06.380 ⇒ 00:07:33.760 Amber Lin: Yeah. So finding those actual steps? Okay, it’s actually going to be 80? Or can we actually raise the price, as we found out on Amazon. And maybe they’re not that responsive. So we can maybe actually increase the price without affecting quality. Or maybe we dive deeper on shopify or on Amazon, and pay. Not just pricing. Can we do it without playing with the prices? There are other things we can do. So. You want those tactical implementation steps right?
56 00:07:34.940 ⇒ 00:07:46.050 bencohen: All of those are good, I think, from for what I’d like to see on, that is like an actual plan. Because now, analyzing what I my basic, pretty simple idea, which is what we’ve been doing all these years.
57 00:07:46.680 ⇒ 00:07:55.889 bencohen: I think we’re we’re looking for like a layer, deeper. So if you have a plan for that, you say. You know, for the next month what we’d like to find out is our coupons more valuable than.
58 00:07:56.560 ⇒ 00:08:07.930 bencohen: Whatever. You know, there’s a lot of ways, you guys know. But but an actual, you know. Make, you know. Do this on shopify. Lower the price. Here’s the coupon. Give us
59 00:08:08.430 ⇒ 00:08:12.360 bencohen: a week. Give us 20 transactions, whatever you need.
60 00:08:12.360 ⇒ 00:08:22.289 Amber Lin: Yeah, cool. So like a complete testing plan to figure out, these are all things we want to go through. Want to isolate. The variables want to make sure to see the impact of all of them.
61 00:08:22.570 ⇒ 00:08:24.889 bencohen: Yeah, yeah, that would be of
62 00:08:25.050 ⇒ 00:08:29.600 bencohen: that would be of some value. But on the AI, I think that
63 00:08:29.720 ⇒ 00:08:33.750 bencohen: that’s really where we need to be focused because this is a big opportunity that
64 00:08:34.000 ⇒ 00:08:41.849 bencohen: Dan wants to spearhead. And obviously we’re going to work together to do this. I think the the proof of concept is a past proof of concept.
65 00:08:42.049 ⇒ 00:08:48.239 bencohen: I don’t know what happened between your developer and ours, but we we bought a bunch of domains.
66 00:08:48.930 ⇒ 00:09:00.730 bencohen: It just needs to happen needs, we need to move this off of this heroku instance. Well, it’s always going to be on Heroku, but it needs to be on our domain. I told him to just do it. So I think it’s hung up on
67 00:09:01.220 ⇒ 00:09:12.980 bencohen: you guys. So if you can just see to make sure it gets done like today tomorrow. That’d be good if anything is needed from our end, let me know. But our developer has full access to the domain.
68 00:09:13.870 ⇒ 00:09:15.950 bencohen: I think the issue is we.
69 00:09:16.060 ⇒ 00:09:25.379 bencohen: If this is a huge opportunity which we, we assess that it is, we need to think of all the ways that we can
70 00:09:25.640 ⇒ 00:09:32.089 bencohen: add to this, I think our 1st idea was put it on a better domain, so we could share it with. Suppose.
71 00:09:32.380 ⇒ 00:09:32.740 Amber Lin: Yeah.
72 00:09:32.740 ⇒ 00:09:33.790 bencohen: And friends.
73 00:09:33.990 ⇒ 00:09:36.500 Amber Lin: See what feedback, cool parts.
74 00:09:36.500 ⇒ 00:09:38.540 bencohen: Yeah, we want to see what people
75 00:09:38.900 ⇒ 00:09:43.429 bencohen: I think before we we give you a you know, a mandate on the next piece of work. I’d like.
76 00:09:43.590 ⇒ 00:09:48.259 bencohen: you know, some friends, to just describe how they interact. I don’t even I don’t even want to leave.
77 00:09:48.260 ⇒ 00:09:48.640 Amber Lin: Keep the money.
78 00:09:48.640 ⇒ 00:09:52.981 bencohen: I just want to send it to them and just see what happens. And I’m sure, Dan, the same
79 00:09:53.610 ⇒ 00:09:59.056 bencohen: I think that could inform what we might be missing or what we’ve done. Well,
80 00:09:59.790 ⇒ 00:10:01.279 Daniel Schonfeld: And have you sent it to anyone yet.
81 00:10:01.750 ⇒ 00:10:05.749 bencohen: No, I’ve been waiting for the for the to be on the domain.
82 00:10:06.220 ⇒ 00:10:09.209 Daniel Schonfeld: But I think the pptg.ai was fine.
83 00:10:09.640 ⇒ 00:10:12.120 bencohen: Yeah, of course, just waiting for them to to point.
84 00:10:12.120 ⇒ 00:10:28.259 Amber Lin: Yeah. I remember we were waiting for I think his name was Nas to give us a I think a dns got what it’s called. I think we need that to enable it on our side. I would say, if if I.
85 00:10:28.260 ⇒ 00:10:28.920 bencohen: We sent it.
86 00:10:28.920 ⇒ 00:10:39.229 Amber Lin: But I would say, you just send the Heroku one to some close friends to test it out while we can work on connecting it to the domain, because feedback and time is really valuable.
87 00:10:40.470 ⇒ 00:10:50.379 bencohen: The domain takes 10 min Nazral sent the C name and all of that stuff. So if it got lost in a thread or something in slack that happens. But
88 00:10:50.580 ⇒ 00:10:55.430 bencohen: I’d like to put that on the correct domain. Looks much better since sharing it with people that way.
89 00:10:55.650 ⇒ 00:10:59.490 bencohen: and it’s quick, there’s no there’s no reason to. There’s no reason to not do it.
90 00:10:59.590 ⇒ 00:11:04.420 Amber Lin: Okay, yeah. Let me let me go check that once. That’s where.
91 00:11:04.420 ⇒ 00:11:07.870 bencohen: We’ll just. We’ll just. We’ll we can follow up on that. After the call.
92 00:11:08.350 ⇒ 00:11:09.470 Amber Lin: Yeah. Totally.
93 00:11:09.890 ⇒ 00:11:14.540 bencohen: From your guys perspective, what do you think would be.
94 00:11:15.380 ⇒ 00:11:18.740 bencohen: guess what? What do you guys feel about this? So far like what?
95 00:11:18.990 ⇒ 00:11:23.690 bencohen: Based on your knowledge because you guys are more AI focused than we are.
96 00:11:25.100 ⇒ 00:11:28.370 bencohen: What are the opportunities that you think? Make sense here.
97 00:11:29.840 ⇒ 00:11:44.760 Amber Lin: What is I have a lot of thoughts because I I also lead our other AI client projects. I was wondering what is your goal for having this? AI. Is it for you guys to seem like your tech forward? Is it? For it.
98 00:11:44.760 ⇒ 00:11:45.360 bencohen: No, no, no.
99 00:11:45.360 ⇒ 00:11:47.039 Amber Lin: Us? Or what is it for.
100 00:11:47.200 ⇒ 00:11:49.540 bencohen: No, no, no, it it we don’t wanna
101 00:11:50.520 ⇒ 00:11:53.749 bencohen: do this as like a stunt. We actually see value.
102 00:11:54.400 ⇒ 00:11:56.499 bencohen: And and giving people value
103 00:11:56.690 ⇒ 00:12:04.549 bencohen: for free. Obviously, our, let’s say, compensation from all of this effort is, they will be in our community.
104 00:12:05.220 ⇒ 00:12:07.249 bencohen: Perhaps get their email address.
105 00:12:08.170 ⇒ 00:12:15.610 bencohen: We could maybe turn them into a customer one day. But this is a way of of looking at customer acquisition.
106 00:12:15.880 ⇒ 00:12:21.929 bencohen: you know, Pr, controlling the narrative and and getting data ultimately.
107 00:12:22.730 ⇒ 00:12:27.350 Amber Lin: Sounds good. So it sounds like it’s a lot of top of funnel action.
108 00:12:27.500 ⇒ 00:12:28.140 bencohen: Yeah.
109 00:12:28.320 ⇒ 00:12:36.609 Amber Lin: Yeah. Okay, let’s see, I think a really quick value add would be
110 00:12:36.900 ⇒ 00:12:42.220 Amber Lin: because these customers ask all these questions. There’s really an opportunity to say.
111 00:12:42.330 ⇒ 00:12:44.809 Amber Lin: Hey, we got this on our website.
112 00:12:44.960 ⇒ 00:13:03.750 Amber Lin: and we’re currently going through a discount of like 10% off or etc. So if they ask any questions that that’s related to a certain product. Then you can say, by the way, this is this is something we provide, or, by the way, we have certain services, if that’s something that you would like.
113 00:13:03.870 ⇒ 00:13:10.490 bencohen: Yeah, that’s a great place to end up. But first, st we need to have a really well rounded product that answers.
114 00:13:11.650 ⇒ 00:13:18.489 bencohen: All people’s curiosities relating to their pool. So I think chemistry, we thought, was universal, because everybody that has a pool has
115 00:13:18.850 ⇒ 00:13:20.539 bencohen: sort of to deal with that.
116 00:13:20.700 ⇒ 00:13:26.029 bencohen: Some people don’t deal with anything with regard to their pool, and they might never ask a question ever.
117 00:13:26.200 ⇒ 00:13:28.059 bencohen: but for a lot of America.
118 00:13:28.400 ⇒ 00:13:36.239 bencohen: which is where we’re focused right now. Water. Chemistry is kind of like the. It would be the top of the funnel, you know.
119 00:13:36.900 ⇒ 00:13:37.470 bencohen: and then.
120 00:13:37.470 ⇒ 00:13:39.280 Daniel Schonfeld: Let me let me jump in real quick, guys.
121 00:13:43.880 ⇒ 00:13:45.370 Daniel Schonfeld: I asked.
122 00:13:46.000 ⇒ 00:13:51.400 Amber Lin: So I wanna I I did 2 things in Chat. Gbt, I said, I basically.
123 00:13:51.870 ⇒ 00:13:53.589 Daniel Schonfeld: Told it. What I’m trying to do.
124 00:13:53.760 ⇒ 00:13:55.910 Daniel Schonfeld: I said, look, I’m trying to build a pool app.
125 00:13:57.870 ⇒ 00:14:00.709 Daniel Schonfeld: How would I do that if I know nothing about coding.
126 00:14:00.830 ⇒ 00:14:05.900 Daniel Schonfeld: And then how would I build one? That’s scalable, because the idea here is
127 00:14:06.600 ⇒ 00:14:10.730 Daniel Schonfeld: to build something rapidly, and we can iterate on the fly
128 00:14:10.960 ⇒ 00:14:29.109 Daniel Schonfeld: that we keep adding data to to make it smarter and better than what’s out there, because there are. If you go on chat, Gpt, and you just put in. Go in the explorer. You’ll find 20 apps, one made by the community, one made and some made by tech guys that do everything that we’re trying to do.
129 00:14:29.930 ⇒ 00:14:34.870 Daniel Schonfeld: So I just went in and reverse engineered it. I asked Chatgpt, how did this person build this
130 00:14:35.040 ⇒ 00:14:36.350 Daniel Schonfeld: pool chat app?
131 00:14:36.900 ⇒ 00:14:39.870 Daniel Schonfeld: And it basically told me everything that they did for each one of them.
132 00:14:40.945 ⇒ 00:14:44.030 Daniel Schonfeld: And it said exactly how they trained it.
133 00:14:44.360 ⇒ 00:14:47.360 Daniel Schonfeld: what what data sources they’re pulling from.
134 00:14:48.040 ⇒ 00:14:51.869 Daniel Schonfeld: So my, my thought process was and I’m just gonna share my screen with you.
135 00:14:51.870 ⇒ 00:14:52.400 Amber Lin: Yeah.
136 00:14:53.590 ⇒ 00:14:55.529 Daniel Schonfeld: How do I do that? Here we go.
137 00:14:57.940 ⇒ 00:14:59.420 Daniel Schonfeld: Can you see this.
138 00:14:59.420 ⇒ 00:15:00.549 Amber Lin: Yeah, I can see your screen.
139 00:15:00.550 ⇒ 00:15:05.690 Daniel Schonfeld: This is what Chat Gpt told me to do. If I don’t know how to code, or if I do know.
140 00:15:07.050 ⇒ 00:15:08.120 Daniel Schonfeld: Code.
141 00:15:09.170 ⇒ 00:15:28.569 Daniel Schonfeld: So it said it said, If I want to build out my own app and do it on my own, and I told it. I don’t want to rely on you for all the answers I want to build my own front end, and then you should kind of color in the rest of it. This is the this is kind of the protocol it gave me to do this myself on the left side.
142 00:15:29.210 ⇒ 00:15:31.990 Daniel Schonfeld: And then basically, for, like you guys to do it on the right side.
143 00:15:32.590 ⇒ 00:15:40.969 Daniel Schonfeld: but you can kind of see the different things. I was asking it to do. And this is kind of a roadmap, for how I look at what we want this thing to look like
144 00:15:41.140 ⇒ 00:15:43.190 Daniel Schonfeld: in in a short period of time.
145 00:15:43.770 ⇒ 00:15:46.220 Daniel Schonfeld: So we want to be able to upload documents.
146 00:15:46.690 ⇒ 00:15:52.949 Daniel Schonfeld: And it told me these these different programs to store those documents or to use it as a database
147 00:15:53.440 ⇒ 00:16:07.770 Daniel Schonfeld: technician input so if we have service guys or even our own text internally, and they want to add new data in this is a place, they can do it the front end again. This is what it told me to just do. If I have no coding experience.
148 00:16:08.740 ⇒ 00:16:14.519 Daniel Schonfeld: Admin dashboard Llm. Integration. Just use Zapier open AI plugin, chat, base, etc.
149 00:16:14.900 ⇒ 00:16:17.129 Daniel Schonfeld: Anyways, I don’t want to go through each one, but
150 00:16:17.360 ⇒ 00:16:26.975 Daniel Schonfeld: what I’m trying to figure out is, can we reverse engineer the 1st 20 pool chat gpt I don’t know what you call them.
151 00:16:27.980 ⇒ 00:16:29.469 Amber Lin: Yeah. The custom gpts.
152 00:16:29.470 ⇒ 00:16:33.910 Daniel Schonfeld: The custom Gpts like really quickly. I could probably do it in in an hour.
153 00:16:35.260 ⇒ 00:16:41.700 Daniel Schonfeld: take all those data sources, dump them in and tell our front end app to go pull from those areas.
154 00:16:42.220 ⇒ 00:16:45.045 Daniel Schonfeld: Take all of our manuals.
155 00:16:45.840 ⇒ 00:17:08.910 Daniel Schonfeld: internally warranty zendesk tickets to look through those and now have something a bit proprietary that we actually have a front end that’s pulling in all of this data again. I’m not. I know nothing about code or anything. I just doing this kind of logically, it’s like, how do we start pulling in proprietary data sources plus publicly available
156 00:17:09.160 ⇒ 00:17:16.659 Daniel Schonfeld: to get a rich set of data that we can answer a lot of questions quickly, or at least as good as everyone else out there.
157 00:17:17.140 ⇒ 00:17:24.550 Daniel Schonfeld: because we also have our own internal proprietary data from our own business, plus our retail business.
158 00:17:24.730 ⇒ 00:17:33.130 Amber Lin: And have something where it can answer a lot of questions pretty quickly, and then I’d rather be spending more time.
159 00:17:33.640 ⇒ 00:17:35.459 Daniel Schonfeld: Kind of iterating on the fly.
160 00:17:35.980 ⇒ 00:17:55.769 Daniel Schonfeld: you know. Wow! It’s pulling from 5 different data sources. It’s too. It’s it’s not concise, like I feel like we can kind of fine tune it as we go. But at least we can get out in front of these people. It’s gonna be so boring like I said it to a friend of mine I was like, Hey, can you use this to just test chemistry? She’s like, can I upload a test strip. No.
161 00:17:56.070 ⇒ 00:18:18.079 Daniel Schonfeld: Can I show it? Pictures? No. All I can do is tell it the Ph. So I know what’s going to happen. It’s going to tell the Ph. Oh, your pool’s green, and add some chlorine that could take weeks for Ben and I to send out to a hundred people. And they’re like, yeah, this is great. It’s like sending a text message and getting it back and saying, the Ph is 6. By the way, this is what we asked for.
162 00:18:18.080 ⇒ 00:18:19.570 Amber Lin: Cool, exciting feature.
163 00:18:19.570 ⇒ 00:18:29.099 Daniel Schonfeld: Yeah, I’m not. I’m not putting anyone down on Brainforge team. This is exactly what I asked for. So we got what we asked for. But is there now a way.
164 00:18:29.790 ⇒ 00:18:34.699 Daniel Schonfeld: That we can set something like this up, maybe like the Mvp. Model that’s on the left side.
165 00:18:34.910 ⇒ 00:18:36.989 Daniel Schonfeld: And or can I just interject.
166 00:18:37.850 ⇒ 00:18:44.829 bencohen: Here’s the thing with the app development piece of this is gonna is gonna come later because a lot of this stuff now is built
167 00:18:45.180 ⇒ 00:19:04.909 bencohen: was no js, so that could be wrapped in from, you know, a web app to a mobile app relatively, simply, I think the bigger thing is, are you suggesting? We try to ingest all of what goes on in all the other custom gpts, so that we’re like the super Gpt, and that in itself is a proprietary idea. And then from there we layer in
168 00:19:05.340 ⇒ 00:19:08.000 bencohen: what’s actually very proprietary from our
169 00:19:08.000 ⇒ 00:19:10.720 bencohen: trying to open source a bit.
170 00:19:10.920 ⇒ 00:19:17.850 Daniel Schonfeld: The data that’s going into this thing from our end. So I said, Ben, me
171 00:19:17.950 ⇒ 00:19:22.710 Daniel Schonfeld: and Cody are the we can upload our own data. We can upload data sources.
172 00:19:23.798 ⇒ 00:19:35.230 Daniel Schonfeld: Let’s just say I found something really cool out there. That lists all the different service guys in America. And I’m like, oh, that’d be neat to if anybody ever asked for a service guy in their area. We have that, and I can just go do it.
173 00:19:35.410 ⇒ 00:19:58.899 Daniel Schonfeld: Cody figures out how to fix a Hayward heater, and he’s like shit. I’m going to put that into A into a document, and I’m going to upload that today. Ben speaks with his service Guy, and he says, Oh, there’s this new thing that fixes water flow and it makes it work faster. It’s called this, and all you got to do is that? And he asked him, can you just write it in an email? Ben uploads that rather than doing it
174 00:19:59.080 ⇒ 00:20:18.150 Daniel Schonfeld: one at a time. What if we kind of allowed ourselves to mass upload data whenever we want? We we would document and say, Hey, we uploaded this to air table. Whatever the hell it’s called. You know. And we we keep adding more, more content and data every day if we want to.
175 00:20:18.150 ⇒ 00:20:19.570 Daniel Schonfeld: Yeah, can I?
176 00:20:19.570 ⇒ 00:20:20.539 Daniel Schonfeld: This massive deal.
177 00:20:21.390 ⇒ 00:20:24.552 Amber Lin: Can I actually tell you what we’re doing for other clients?
178 00:20:24.840 ⇒ 00:20:25.250 bencohen: Yes.
179 00:20:25.520 ⇒ 00:20:35.589 Amber Lin: Type of agents. So for one of our clients, this is for their customer service team, and they’re a very big almost national company. And they.
180 00:20:35.590 ⇒ 00:20:38.140 bencohen: Is this the the pest? Is this like the pest? One.
181 00:20:38.300 ⇒ 00:21:01.099 Amber Lin: Yes, this is, I was just visiting them in Austin, and they have a lot of documents, and every day to spend like on the call, spend time looking through all those documents, and what we do for them is one build a custom Gpt, essentially answering all those questions that customers have but 2. The most the more interesting part is, we’re building a
182 00:21:01.110 ⇒ 00:21:22.370 Amber Lin: AI links to a database for their trainers, for people who want to upload those documents, update training documents, making sure that everything is covered and those 2 are going to be linked together. So for you guys, you want to be able to easily maintain all the documents. Make sure that there’s no
183 00:21:22.370 ⇒ 00:21:37.520 Amber Lin: contradicting errors, and make sure that it’s in an easy, readable format for the bot for the AI agent to ingest. So what I would propose is that we actually do the same here we do something that’s customer facing.
184 00:21:37.670 ⇒ 00:21:41.819 Amber Lin: And we do something that’s internally facing, because I wanted to take
185 00:21:42.230 ⇒ 00:21:57.399 Amber Lin: as minimal of your time as necessary because uploading documents formatting them will take time. And ideally, we have something that automates that and makes that easier. So I would say, having that structure
186 00:21:57.460 ⇒ 00:22:12.409 Amber Lin: will be really helpful. And ultimately it’s a knowledge base, right? I don’t know if you’ve seen those knowledge graphs with different tags, and it’s kind of like a big web of information. So tagging those documents will also help AI to retrieve them
187 00:22:13.083 ⇒ 00:22:18.680 Amber Lin: versus it, going through a whole long list and saying, Oh, what! Where is everything?
188 00:22:19.950 ⇒ 00:22:37.450 Daniel Schonfeld: That makes sense. Yeah, whatever format or process or system you guys have set up great, it doesn’t have to go automatically in. But if we if I can wake up. I read a bunch of articles I see manuals, things I know we don’t have in there, and I can check a list, even if it was a Google sheet that said, Here’s everything about heaters.
189 00:22:37.940 ⇒ 00:22:44.059 Daniel Schonfeld: Specifically Hayward heaters, and it lists out each the knowledge base. That’s a good good word.
190 00:22:44.405 ⇒ 00:22:52.380 Daniel Schonfeld: And I see that it’s not there. I could just put it in there, and every week someone from your end can just go in. Do whatever you gotta do, scrape it in or format it whatever.
191 00:22:52.490 ⇒ 00:23:00.859 Daniel Schonfeld: But we can constantly add to this thing, and it can grow without having to do a new like statement of work or something. Something. That’s kinda
192 00:23:01.160 ⇒ 00:23:05.032 Daniel Schonfeld: you know what I’m talking about. I know you. I know you understand what I’m saying. So
193 00:23:05.460 ⇒ 00:23:09.980 Daniel Schonfeld: rather than do one off builds. It’s a constant living, breathing data set.
194 00:23:10.800 ⇒ 00:23:23.399 Amber Lin: Sounds good. Is that is that your main focus of making sure that it has access to all as much as possible crucial and pull information, or is there? Is that the main focus.
195 00:23:23.400 ⇒ 00:23:30.870 Daniel Schonfeld: I I think right now we don’t exactly know. That’s the that’s the true answer. But what I kind of see is.
196 00:23:31.590 ⇒ 00:23:34.220 Daniel Schonfeld: there’s not a lot of people doing this respect
197 00:23:34.220 ⇒ 00:23:43.960 Daniel Schonfeld: for swimming pools like, there’s not a lot of people who would want to do this, or would have a need to do it. We do, because we’re one of the only direct to consumer pool companies.
198 00:23:43.960 ⇒ 00:23:44.440 Amber Lin: Hello!
199 00:23:44.440 ⇒ 00:23:55.679 Daniel Schonfeld: Anywhere. And so there, it’s a multifunctional tool in that number one. It could be kind of a lead generation or customer acquisition tool, because if we had all the data in the, in the
200 00:23:55.840 ⇒ 00:23:59.770 Daniel Schonfeld: theoretically, in the world about pools, everyone can use it.
201 00:24:00.110 ⇒ 00:24:00.870 Amber Lin: Yeah.
202 00:24:00.870 ⇒ 00:24:01.769 Daniel Schonfeld: Diagnose their problem.
203 00:24:01.770 ⇒ 00:24:03.240 Amber Lin: 2 for pool problems.
204 00:24:03.240 ⇒ 00:24:26.009 Daniel Schonfeld: Yeah, and it would fix it instantly, or tell you what the problem is, or tell you how to fix it, because it’s a mystery. It’s like fixing your car. Nobody knows how to do that, and when you bring it in, you know you get hosed on the price pools kind of the same thing. Some guy shows up and says, you own $3,000 for this piece of equipment you never heard of. Nobody has any way to to fact, check it, or to understand it, or to diagnose.
205 00:24:26.600 ⇒ 00:24:27.190 Amber Lin: Yeah.
206 00:24:27.190 ⇒ 00:24:34.590 Daniel Schonfeld: So we want to give owners that power to either do it themselves or to keep everyone in check.
207 00:24:35.230 ⇒ 00:24:47.570 Daniel Schonfeld: the the real thing that I. The real value for us is that if every pool owner in the world has it, and we sell pool equipment, we can offer our own pool equipment when that problem arises.
208 00:24:48.170 ⇒ 00:24:51.800 Daniel Schonfeld: and so that the ultimate goal is that, yes, sell pool equipment.
209 00:24:52.355 ⇒ 00:25:04.570 Daniel Schonfeld: But I think this this thing could take a life of its own in many different directions. But it could be a lead Gen. Tool. It could live on our website as an AI assistant to help people find the right product on our own site.
210 00:25:06.580 ⇒ 00:25:13.799 Daniel Schonfeld: I think it could be a subscription service, because it’s like a you know, a a service technician in your pocket for the pool.
211 00:25:13.970 ⇒ 00:25:33.829 Daniel Schonfeld: I think there’s a million things I can think of ideas for how this would be useful. But I think we got to get this core knowledge base or data set in there. So we can start at seeing if people actually give a shit. That’s the that’s the other big question mark is, do people care to have this? And will they use it? And so I don’t know the answer to that yet, until we really.
212 00:25:33.830 ⇒ 00:25:39.169 Amber Lin: I see that sounds like we need to build towards the Mvp. That’s
213 00:25:39.690 ⇒ 00:25:57.529 Amber Lin: bit more than what we have right now is a proof of concept and we need to make sure that before we send it out to people that it actually answers pretty accurately. Then we can get pretty decent and accurate feedback on. Okay, is this helpful? Because if the answers are not correct, of course it’s not going to be helpful.
214 00:25:57.720 ⇒ 00:26:06.550 Daniel Schonfeld: Yeah, let’s just say even. And another idea, even if none of this worked for the consumer, the consumer never cared to use an app like this.
215 00:26:07.150 ⇒ 00:26:22.159 Daniel Schonfeld: It would be good for us internally for marketing purposes, just to look up stuff and understand things for warranty. Have a QR code on the box that links to this thing just for basic questions. So we don’t have to hire additional customer service people for marketing.
216 00:26:22.160 ⇒ 00:26:23.300 Amber Lin: Good research.
217 00:26:23.560 ⇒ 00:26:28.409 Daniel Schonfeld: There. I just I can keep going on different applications for this. But I know it’s gonna be
218 00:26:28.610 ⇒ 00:26:31.155 Daniel Schonfeld: so important that we have this
219 00:26:32.130 ⇒ 00:26:34.369 Daniel Schonfeld: But obviously we we can’t just up
220 00:26:34.610 ⇒ 00:26:43.899 Daniel Schonfeld: throw and throw a million data sets in there and try to figure it out later, when it doesn’t come back accurate. So I I get the need for it to be systematic.
221 00:26:44.620 ⇒ 00:26:45.140 Amber Lin: Yeah.
222 00:26:45.140 ⇒ 00:26:56.610 Daniel Schonfeld: But that’s what I’m looking for is not so much the the end result. But like building the the right structure. Now that we can start rapidly testing this thing to see what people care about or why they use it.
223 00:26:56.610 ⇒ 00:26:57.190 Amber Lin: Hmm.
224 00:26:57.930 ⇒ 00:27:03.789 Daniel Schonfeld: And then we’ll then we’ll double triple, quadruple down on those things that they find the most value of. So we can.
225 00:27:04.550 ⇒ 00:27:06.400 Daniel Schonfeld: you know, get more adoption out of it.
226 00:27:06.400 ⇒ 00:27:09.589 bencohen: My my my assumption.
227 00:27:10.240 ⇒ 00:27:18.809 bencohen: I don’t know that it’s going to be true, but I think we can work we can. We can. We can find out is, I think, things that are going to save pool owners. Money are going to be what.
228 00:27:19.200 ⇒ 00:27:21.780 Amber Lin: Has the highest utility here.
229 00:27:22.240 ⇒ 00:27:23.290 bencohen: So.
230 00:27:24.950 ⇒ 00:27:31.360 bencohen: you know, I think if that’s our guiding light, it makes a lot of sense. Right now for people to
231 00:27:31.610 ⇒ 00:27:35.289 bencohen: find money saving information. They have to do a lot of work.
232 00:27:35.490 ⇒ 00:27:36.010 bencohen: You, too.
233 00:27:36.010 ⇒ 00:27:37.680 Amber Lin: Even know where to save money.
234 00:27:37.980 ⇒ 00:27:46.619 bencohen: And they they can. There’s great creators. There’s people that do it, but they really have got to dig, and they might run out of energy. It’d be interesting.
235 00:27:46.620 ⇒ 00:27:49.049 Daniel Schonfeld: They don’t. They don’t even know they can save money.
236 00:27:50.000 ⇒ 00:27:50.910 bencohen: A lot of them. Yeah.
237 00:27:51.310 ⇒ 00:28:02.359 bencohen: a lot of our customers, you know. I think this is right down their their alley, you know, when they find our pumps. We are like the 1st time they’ve seen the possibility of saving anything for their pool.
238 00:28:03.120 ⇒ 00:28:14.159 bencohen: Right so, and we’re a little company compared to the to the industry. So like if if you take that frame of mind where you’ve you’ve opened somebody’s eyes to a pump. That’s not $2,000. It’s only 820.
239 00:28:14.960 ⇒ 00:28:23.079 bencohen: All of a sudden the possibilities, they say. Oh, well, what else could be an area we can save now that we’ve optimized the pump piece of the puzzle. So
240 00:28:23.200 ⇒ 00:28:31.809 bencohen: I think that’s where we’re gonna find a lot of traction, I think what you guys have with that pest company is that natural escalation to a human agent?
241 00:28:32.160 ⇒ 00:28:35.810 bencohen: Right? When things, when things get to a point where all of a sudden
242 00:28:36.150 ⇒ 00:28:50.550 bencohen: the machine may be out of answers, and it could get frustrating. So you know when to escalate. And then, you know, that’s when we have, I think, a good paywall to move people into something, but that we’re getting ahead, I think, for now Dan’s idea is to create.
243 00:28:51.350 ⇒ 00:28:54.270 bencohen: you know, a moat of information.
244 00:28:55.120 ⇒ 00:29:06.989 Daniel Schonfeld: I use chat gpt to find I had an issue with my Hayward heater. It broke. I didn’t even know if it was under warranty I use chatgpt to figure out if it was under warranty, and to find me the service guy which it found.
245 00:29:07.170 ⇒ 00:29:10.129 Amber Lin: We don’t make any money on it or referral on it.
246 00:29:10.330 ⇒ 00:29:21.452 Daniel Schonfeld: It’s it’s that’s like even the simplest form. It’s a an affiliate like a publisher and an affiliate network. It’s literally there’s so many use cases for it?
247 00:29:22.230 ⇒ 00:29:44.779 Daniel Schonfeld: and so we just have to think about. Do we use something like I have on my screen here? Or do we just go for it and say, you know what we’ll we’ll build you this thing on the right side. I have no idea if these are the right things. But you know, it’s gonna take 2, 3 weeks. But once you build that, you can start building all sorts of different types of knowledge bases within it, and use it for whatever you want.
248 00:29:46.780 ⇒ 00:29:51.340 Amber Lin: Yeah, sounds like we’re all very excited to keep building forward.
249 00:29:51.763 ⇒ 00:30:14.229 Amber Lin: I do want to talk a little bit about how we’re gonna move forward with the commitment. You guys are very, very special to us, especially I. I joined late. But you are Brain Forge’s 1st client. And you guys are very, very important, because we usually don’t do 2 weeks and 2 week engagements anymore. Most of our engagements are 3 months to 6 months. So I kind of wanted to hear
250 00:30:14.230 ⇒ 00:30:33.329 Amber Lin: how you want this to move forward. If we would have a stable commitment to be able to give you that roadmap. Tell you, okay, these are the things we need to do and clearly commit our engineers to a more longer term project. So I want to just hear what you feel about that.
251 00:30:33.920 ⇒ 00:30:44.380 Daniel Schonfeld: Yeah, we’re we’re open to it. I think this is like the future of what we’re doing. I don’t know what I don’t know. I don’t know how many people, how much time I see this being an ongoing thing in perpetuity, I I don’t see.
252 00:30:44.380 ⇒ 00:30:44.770 Amber Lin: Good evening.
253 00:30:45.430 ⇒ 00:30:57.369 Daniel Schonfeld: So I I mean, maybe it takes a a handful of people to build the initial the initial thing. But then I could see someone being dedicated almost full time, part whatever, just to working on this.
254 00:30:57.550 ⇒ 00:30:58.890 Amber Lin: Yeah, totally. And.
255 00:30:58.890 ⇒ 00:30:59.500 Daniel Schonfeld: Just growing it.
256 00:30:59.500 ⇒ 00:31:28.459 Amber Lin: Also help scale your internal people up into knowing how to operate these things like, we won’t tie you to us wherever we want to lift all our clients up to enable you to do things as well, so it will be great if you’re open to engagement that we scope it out, say maybe 10 to 20 h of engineering per week over the next 3 months. If we’re as an example engagement and we will strap upon contract, and we can look over it.
257 00:31:28.620 ⇒ 00:31:31.190 Daniel Schonfeld: Can I say one thing? I I don’t care about the hours I care about.
258 00:31:31.190 ⇒ 00:31:31.750 Amber Lin: The.
259 00:31:31.750 ⇒ 00:31:35.910 Daniel Schonfeld: Salt. So like, I think we should figure out what’s how do we do?
260 00:31:36.150 ⇒ 00:31:39.400 Daniel Schonfeld: What we just said? And how long will that take? Maybe it’s.
261 00:31:39.400 ⇒ 00:31:39.740 Amber Lin: Okay.
262 00:31:39.740 ⇒ 00:31:42.939 Daniel Schonfeld: Month, 2 months to get to like the 1st iteration
263 00:31:44.300 ⇒ 00:31:54.230 Daniel Schonfeld: and then, yeah, I don’t care about the hours. If you say, look, it’s gonna cost. I’m just making us up 10 grand. I’m just making up to to do this, and we’re gonna give you 20 HA week, and it’s gonna take 30 weeks.
264 00:31:54.460 ⇒ 00:31:59.139 Daniel Schonfeld: I’m not interested like, I rather say, how do we? How do we just get to that 1st phase? 10 grand.
265 00:31:59.790 ⇒ 00:32:00.950 Daniel Schonfeld: We could do it in a week.
266 00:32:01.180 ⇒ 00:32:15.789 Amber Lin: This is awesome because we actually prefer fixed fixed monthly contracts, and then we can just do how many hours within that budget without without the need to penny pinch. Oh, we did this extra hour on this so.
267 00:32:16.070 ⇒ 00:32:19.339 bencohen: I think. Let me. I think we’re going in the wrong.
268 00:32:19.470 ⇒ 00:32:27.559 bencohen: I think you guys need to spend a little bit of time specing this out because this is an unusual. This isn’t quite an agent. It’s not. It’s not quite a.
269 00:32:27.710 ⇒ 00:32:31.010 bencohen: This is more of a system of how we’re going to ingest.
270 00:32:31.330 ⇒ 00:32:32.090 bencohen: So
271 00:32:32.260 ⇒ 00:32:46.879 bencohen: I think we need to see from you guys how you imagine building that. And you know and give us like a Chinese menu sort of situation. You know, we think building the ingestion system is, gonna take this much time. I know you need to quantify things in hours. So
272 00:32:47.350 ⇒ 00:32:52.709 bencohen: but we don’t we really only care about doing the right thing? So like
273 00:32:52.930 ⇒ 00:33:03.489 bencohen: it’s show us the plan for the ingestion how that can grow. And then, if if it’s a fixed thing. If it’s not a fixed thing, whatever good plan means, we’ll be able to move forward.
274 00:33:03.700 ⇒ 00:33:14.999 Amber Lin: Okay, fantastic. That’s really, really good news for me. And I’ve been thinking about this roadmap constantly. And it’s really great to hear from you guys what you envision, and to know what’s most important.
275 00:33:15.542 ⇒ 00:33:26.429 Amber Lin: I will meet internally to scope it out. I’ll ask you guys if I have any questions. We can review that roadmap, and then we can talk about what kind of engagement we want to do.
276 00:33:26.430 ⇒ 00:33:34.019 Daniel Schonfeld: When do you think you can have? I just want to stay very involved, at least in the beginning, for sure, because I want to make sure we get this off on the right
277 00:33:34.200 ⇒ 00:33:45.399 Daniel Schonfeld: path. And again looking at the what I have on the screen here. You guys can tell us look, I think it’s worthwhile for you to spend less money. Build this. You guys can build this Mvp version.
278 00:33:45.780 ⇒ 00:33:56.300 Daniel Schonfeld: And then, once it’s working, you can scale up. I don’t know what goes into those things. The last thing we wanna do is get, you know, a bunch of people who love it. And then we got to stop for a month to build it out.
279 00:33:56.300 ⇒ 00:33:57.370 Amber Lin: Yeah. Totally.
280 00:33:57.370 ⇒ 00:33:58.769 Daniel Schonfeld: Make sure that we.
281 00:33:58.770 ⇒ 00:34:02.529 Amber Lin: Know what’s getting billed, and we know what’s gonna happen. We’ll give you that plan.
282 00:34:02.720 ⇒ 00:34:05.400 Daniel Schonfeld: Okay, cool, alright, awesome.
283 00:34:05.610 ⇒ 00:34:06.180 Daniel Schonfeld: Great.
284 00:34:06.180 ⇒ 00:34:09.179 Daniel Schonfeld: We’re really excited about this one. And we’re happy to be doing it with you guys.
285 00:34:09.679 ⇒ 00:34:20.389 Amber Lin: Awesome. And that makes me really, really happy because this is a great team to work with, like, I’ve worked with you guys, I work with Kim. I talked to Steven, like, you guys are all really really nice people.
286 00:34:20.909 ⇒ 00:34:22.979 Daniel Schonfeld: Thank you so much. You too.
287 00:34:23.219 ⇒ 00:34:24.279 Amber Lin: Enjoy your day.
288 00:34:25.090 ⇒ 00:34:26.479 Daniel Schonfeld: All right, Amber. We’ll talk to you soon.
289 00:34:26.480 ⇒ 00:34:28.820 Amber Lin: Okay, talk to you soon. Take care. Bye-bye.
290 00:34:29.120 ⇒ 00:34:29.760 Daniel Schonfeld: Bye.