Meeting Title: Brainforge <> Interlude: Kickoff Date: 2025-07-24 Meeting participants: Rafay’s Circleback.ai Notes, Rafay Iqbal, Mustafa Raja, Uttam Kumaran, Matthew Good


WEBVTT

1 00:01:28.340 00:01:29.340 Uttam Kumaran: Hey!

2 00:01:30.760 00:01:32.020 Rafay Iqbal: Yo! How’s it going.

3 00:01:32.020 00:01:33.170 Uttam Kumaran: Hey? Good! How are you?

4 00:01:33.640 00:01:40.220 Rafay Iqbal: Just wait on method. Join just texted in. So I guess give him a quick sec. I think he’s joining up in just a few here. Oh, there he is!

5 00:01:43.990 00:01:44.630 Matthew Good: Nice.

6 00:01:45.650 00:01:46.460 Uttam Kumaran: Hey!

7 00:01:47.290 00:01:48.120 Matthew Good: What’s going on.

8 00:01:49.910 00:01:56.512 Uttam Kumaran: Good. So we have some stuff to share today. Let me just get you guys into this Doc, that we’re started working on.

9 00:01:58.380 00:02:01.159 Uttam Kumaran: I kind of just want to walk you through

10 00:02:01.730 00:02:05.459 Uttam Kumaran: like kind of the overall architecture. It may be

11 00:02:05.600 00:02:12.290 Uttam Kumaran: little bit like technical today, but we do have some questions as we go through this.

12 00:02:14.070 00:02:18.330 Uttam Kumaran: some of these will just kind of dictate, like what the initial proof of concept is. Gonna look like.

13 00:02:18.880 00:02:23.100 Uttam Kumaran: I just invited you to this notion, Doc.

14 00:02:49.250 00:02:51.589 Uttam Kumaran: Let me know when you guys can see that.

15 00:02:52.270 00:02:53.970 Matthew Good: Yeah, yeah, I’m looking. I can see it.

16 00:02:54.270 00:02:59.590 Uttam Kumaran: Okay, cool. So basically, this is like, kind of, we would usually

17 00:02:59.900 00:03:05.859 Uttam Kumaran: work through like a kind of a project management, Doc. But I I think really most of the meat of what I want to talk about is

18 00:03:06.280 00:03:21.549 Uttam Kumaran: part of in this like technical design, Doc. Mainly, I want to just work through. Kind of just give you guys an overlay of like how the stock is gonna work. And then I have some questions. Let me share this.

19 00:03:32.140 00:03:32.930 Uttam Kumaran: heck.

20 00:04:08.440 00:04:10.704 Uttam Kumaran: okay, cool. So

21 00:04:11.990 00:04:20.600 Uttam Kumaran: yeah, I would say, probably the core of this is gonna start at 2, and I’ll just kind of narrate through this and just correct me if there’s more

22 00:04:20.870 00:04:27.679 Uttam Kumaran: things to add, or if we talk through scope, and you’re like, actually, there’s a little bit more details, or we want to add to this. But basically.

23 00:04:27.870 00:04:32.839 Uttam Kumaran: we want to talk about the streamlining of client deck creation.

24 00:04:33.120 00:04:50.669 Uttam Kumaran: you know, roughly trying to achieve something like 20 to 40% in efficiency. This sort of our like baseline. Again, efficiency is kind of a rough overall metric. But we have some metrics that we’ll start to measure basically it. It enables team members to

25 00:04:52.760 00:05:16.159 Uttam Kumaran: rapidly produced sort of content for that. You guys are generating for your clients. Additionally, we kind of wanted to particularly work towards your mission of like, hey, how can how can interlude be sort of one of the innovators in this space, basically using AI to to sort of help, not only with their margins, but maybe the health of the team. So right now I’m I’m assuming it’s mainly

26 00:05:16.250 00:05:24.839 Uttam Kumaran: Matt. It’s just it’s just you guys that are working on this. Are there anyone else on the team? Or is there any plans on like other folks on the team that are going to be like using this system.

27 00:05:26.517 00:05:35.679 Matthew Good: I think it’s just us, for right now we have Vicki, who’s like our our kind of like. She’s our Pm. Not kind of. She’s our Pm. I would eventually like wanna

28 00:05:36.120 00:05:40.300 Matthew Good: loop her in on some of this stuff. But I think for just this deck stuff right now would just be Ralph and I.

29 00:05:40.720 00:05:41.310 Uttam Kumaran: Okay.

30 00:05:41.570 00:05:56.469 Uttam Kumaran: so kind of like, we’re we’re gonna use sort of titles in in lieu of like instead of like your names. Just so we can have some like agnostic user stories. But basically what we’re what we sort of was like, cool as a design team member, I wanna

31 00:05:57.000 00:06:06.100 Uttam Kumaran: basically eliminate this manual copy and pasting they could focus on creative work gonna put related to the deck narrative process.

32 00:06:07.220 00:06:10.579 Uttam Kumaran: Also as a project manager. I want to reduce the draft creation time.

33 00:06:11.052 00:06:13.229 Uttam Kumaran: We don’t. We kind of like.

34 00:06:13.350 00:06:21.430 Uttam Kumaran: I had this like coordinator role. But I think we could probably honestly put this into here.

35 00:06:22.070 00:06:28.440 Uttam Kumaran: Basically, you know the way I think, look at it. Even my company is like, how can get people with lower expertise.

36 00:06:29.357 00:06:33.179 Matthew Good: Plus AI to get close to something like you guys would do right?

37 00:06:33.180 00:06:34.200 Matthew Good: Exactly. Yeah.

38 00:06:34.200 00:06:40.610 Uttam Kumaran: Yeah. So that’s how we think about it as well. In my business like this happens where we have, like coordinators on the sales side.

39 00:06:40.610 00:06:40.939 Matthew Good: And which.

40 00:06:40.940 00:06:46.599 Uttam Kumaran: Them, plus a lot of AI stuff can get closer to like A like a Us based Bdr, or something like that.

41 00:06:47.080 00:06:51.731 Uttam Kumaran: So we have a couple of these things that we’re thinking about. So in terms of like the

42 00:06:52.510 00:06:55.440 Uttam Kumaran: I’ll leave this in terms of like the functional requirements.

43 00:06:55.914 00:07:10.055 Uttam Kumaran: I’ll just list this. Oh, I’ll just go through this. And you can tell me, like what we want to change. So basically, it’s 1 we want to ingest slack, triggered client briefs and file attachments through some bot or some mechanism. We want to trigger a workflow.

44 00:07:10.850 00:07:17.610 Uttam Kumaran: we can optionally reference additional previous documents or questionnaires.

45 00:07:17.770 00:07:28.543 Uttam Kumaran: we return a draft deck. This can either be returned in slack, or I also have a question on where this should be the human in the loop can add notes or revisions.

46 00:07:29.240 00:07:34.660 Uttam Kumaran: and then on approval, this gets published. Into notion. Yeah.

47 00:07:34.980 00:07:41.410 Uttam Kumaran: throughout this process we’re gonna take. We’re gonna sort of log. How long it took the end to end the token usage

48 00:07:41.858 00:07:46.579 Uttam Kumaran: some other metrics that we do, and that will be sort of displayed in a little dashboard, and then

49 00:07:47.025 00:07:56.350 Uttam Kumaran: you’ll also be able to run like if we catch errors and we make fixes, we can rerun the entire workflow to test that they worked again.

50 00:07:57.850 00:08:02.889 Uttam Kumaran: so we have some like metrics that we we would look. But I would say, this is really like.

51 00:08:03.330 00:08:07.460 Uttam Kumaran: Well, I want to spend some time debating about given.

52 00:08:07.669 00:08:07.880 Matthew Good: Right.

53 00:08:07.880 00:08:10.280 Uttam Kumaran: I saw your existing workflow, which is like

54 00:08:10.500 00:08:13.579 Uttam Kumaran: copying into Claude, iterating, and then.

55 00:08:14.140 00:08:15.339 Uttam Kumaran: Going into notion.

56 00:08:16.546 00:08:20.030 Uttam Kumaran: But like, tell me what you think of this so far.

57 00:08:20.230 00:08:24.530 Matthew Good: Yeah, yeah, this is super helpful. I’m just. I’m just reading these bullets again.

58 00:08:25.209 00:08:29.599 Matthew Good: And just slack triggered client briefs and file attachments. Yeah.

59 00:08:29.990 00:08:33.280 Matthew Good: figure the agent workflow. Yeah, I think number 3

60 00:08:33.440 00:08:38.462 Matthew Good: is definitely. I mean, the only thing I would say here is just not optional, like, I definitely

61 00:08:39.570 00:09:05.809 Matthew Good: or kind of like how I’ve evolved. Because I, as my prompting hopefully, has gotten better evolved. I just like start telling Claude that I’m just giving it a knowledge base. And I’m like, I’m gonna build your knowledge base step by step, right? Like, ingest each thing sequentially. And I will give it. You know that questionnaire for sure. Sometimes clients don’t fill it out like actually typed wise, but we talk through it live. So I have. Sometimes it’s just questionnaire. Sometimes it’s questionnaire and call recording like transcript

62 00:09:06.063 00:09:21.539 Matthew Good: but regardless like, we always run a kickoff call, if for no other reason, just to get that recording so that’s like a very key component. And a lot of people suck at writing. So it’s just easier to get the the nuances of what they want to convey or talk about, or the story just like verbally with them.

63 00:09:23.690 00:09:42.180 Matthew Good: so so yeah, definitely, questionnaire prior decks, especially for fundraising. Next, there’s always rarely is like a net. New deck. It’s like, Hey, we have this like initial version, but like it sucks, and we need some help. So there’s often kind of like a current state that we want to build off of, and then I’ll you know. I’ll tell Claude like

64 00:09:43.690 00:09:47.440 Matthew Good: I’ll try to tell it like how to think. I’ll be like, evaluate this from

65 00:09:47.640 00:09:54.520 Matthew Good: not just like as a you know, as an expert pitch copywriter and deck designer, but also about this from the eyes of an lp.

66 00:09:54.660 00:10:07.069 Matthew Good: and then and then I’ll and then I’ll ask it if you saw on the rails. One. But I’ll ask it like, okay, based on the deck you generated. Now put your Lp hat on and identify all the holes in it, and it’ll go back through. And that’s how kind of I like ping Pong back and forth between.

67 00:10:07.290 00:10:07.940 Uttam Kumaran: Okay.

68 00:10:08.140 00:10:09.270 Matthew Good: If that makes sense.

69 00:10:09.460 00:10:17.929 Uttam Kumaran: That makes sense. So yeah, there’s a questionnaire. There’s a transcript. The other question I had was like, so then there’s also like kind of this like system, prompt.

70 00:10:26.010 00:10:37.010 Uttam Kumaran: And then I also was gonna ask if you had considered using like, had you? Have you tried using search mode before? Or is there any benefit you think to having

71 00:10:37.743 00:10:47.280 Uttam Kumaran: the AI go search for information? Either it could be multiple ways. This could be scraping the client site. This could be

72 00:10:47.380 00:10:50.430 Uttam Kumaran: just searching for recent news about the client.

73 00:10:50.900 00:10:51.270 Matthew Good: Okay.

74 00:10:51.270 00:10:52.380 Uttam Kumaran: Think that’s relevant.

75 00:10:53.070 00:10:58.189 Matthew Good: Honestly not. I mean, I’d I’d be like definitely open to push back. But I don’t.

76 00:10:58.847 00:11:00.390 Rafay Iqbal: Think that might be too much.

77 00:11:00.390 00:11:00.810 Uttam Kumaran: Okay.

78 00:11:00.810 00:11:05.540 Matthew Good: Yeah, yeah, I think just because a lot of these funds, too, are these found like they’re not. This is all like

79 00:11:05.660 00:11:14.410 Matthew Good: proprietary confidential information that we’re getting. So we’re signing ndas to get the stuff or not. Not often signing ndas, but like it’s not stuff that’s like widely available on the web.

80 00:11:15.790 00:11:16.450 Uttam Kumaran: Got it.

81 00:11:17.150 00:11:17.770 Matthew Good: Yeah.

82 00:11:18.220 00:11:18.920 Uttam Kumaran: Okay.

83 00:11:22.410 00:11:23.610 Uttam Kumaran: don’t you?

84 00:11:29.250 00:11:30.020 Uttam Kumaran: Okay,

85 00:11:33.200 00:11:38.419 Uttam Kumaran: And then, so this is basically, yeah, I mean, I come just relisting a little bit here. But we have these?

86 00:11:38.878 00:11:43.210 Uttam Kumaran: Can you tell me like how long? I don’t know. If we talked about this

87 00:11:43.540 00:11:50.520 Uttam Kumaran: so like how long it takes you end to end? And I would. I would honestly like you to consider.

88 00:11:51.470 00:11:54.989 Uttam Kumaran: Of course it’s like if you were to sit down. It’s the only thing you had to do in the day.

89 00:11:55.400 00:11:58.080 Uttam Kumaran: but I would rather kind of hear like.

90 00:11:58.550 00:12:01.849 Uttam Kumaran: how long does this take? Given all the other stuff you’re handling.

91 00:12:02.545 00:12:02.790 Matthew Good: Like.

92 00:12:02.790 00:12:08.210 Uttam Kumaran: How does this process typically take end to end? I don’t. We’re not looking for an exact number here.

93 00:12:08.560 00:12:09.829 Uttam Kumaran: but I’m looking for.

94 00:12:09.830 00:12:10.310 Matthew Good: Yeah.

95 00:12:10.310 00:12:10.880 Uttam Kumaran: Ground.

96 00:12:10.880 00:12:17.740 Matthew Good: I would, I would say, like, Yeah, I mean, let’s call it in between. Let’s call it an hour flat, because I’m oftentimes like

97 00:12:18.560 00:12:47.039 Matthew Good: I try to batch these things on days. So like Tuesday this week, I had no calls. And I was like, Okay, I gotta do a shit ton of these. I’d like 4 of them to do so. It takes time to like, gather all the inputs and be like copy and paste, transcript copy and paste questionnaire. Go look back in our dropbox and we onboard clients, and they give us kind of like their data room. We just throw it all into a dropbox. And I gotta go download those files and say, Okay, all this kind of stuff like, and not just like, throw it all at Claude. But like sequentially prompt it, and then and then kind of edit it and

98 00:12:47.306 00:12:50.850 Matthew Good: and stuff like that. And then inevitably, I’m like getting pinged about something or others.

99 00:12:50.850 00:12:51.330 Uttam Kumaran: Yeah, yeah.

100 00:12:51.330 00:12:56.560 Matthew Good: Getting back to that, I go. So I’d say, an hour. Ish, yeah.

101 00:12:57.400 00:13:03.910 Uttam Kumaran: Okay, okay. So I I mean, what do you think about the sort of slack

102 00:13:04.905 00:13:16.630 Uttam Kumaran: involvement here, like, right now, there’s you’re doing this manually, you know, in in my world I like to start things on slack just because there’s other people there. And ideally, this is something eventually that

103 00:13:17.000 00:13:20.809 Uttam Kumaran: we’ll be giving to someone else, most likely. So.

104 00:13:20.810 00:13:21.160 Matthew Good: Yeah.

105 00:13:21.160 00:13:23.290 Uttam Kumaran: Slack is like you would ping a bot.

106 00:13:23.750 00:13:35.070 Uttam Kumaran: It would, either. And this can we think about like exactly how the way it works? But let’s say, could give it all the files necessary. Give it the client to start with and have a draft.

107 00:13:35.280 00:13:37.530 Uttam Kumaran: Is that a good form factor.

108 00:13:38.580 00:13:42.830 Matthew Good: Yeah, I mean, I think slack is kind of like our home bait, like I’m in slack

109 00:13:43.771 00:13:51.259 Matthew Good: slack and notion, but primarily slack. So it’d be easy to just like ping a bot. And you know I don’t know if there’s some world in which the bot would.

110 00:13:51.720 00:14:00.690 Matthew Good: I guess we we have like shared client channels, but sometimes they’re like sending links in there. That might be. That might be too much. But even if I’m just like copy and pasting those links or dropping them in slack.

111 00:14:02.310 00:14:03.120 Matthew Good: Yeah.

112 00:14:05.760 00:14:11.750 Uttam Kumaran: Okay. So one thing I’m gonna put in is like, shared client panel X,

113 00:14:12.610 00:14:15.156 Uttam Kumaran: yeah. So sometimes. So I’ll think about

114 00:14:21.680 00:14:22.530 Uttam Kumaran: Okay,

115 00:14:24.920 00:14:36.000 Uttam Kumaran: okay, great. And so ideally, if you can picture a process and we’ll I can get you some like examples of this. But ideally you would you? It would be initiated by you. You ping.

116 00:14:36.240 00:14:39.980 Uttam Kumaran: it’s literally just be like, add a bot or whatever with information

117 00:14:40.410 00:14:46.000 Uttam Kumaran: it would it? There would be some input validation, basically like, if you missed sending something or

118 00:14:46.170 00:14:48.670 Uttam Kumaran: like for everything we try not to like.

119 00:14:48.900 00:14:56.209 Uttam Kumaran: I just want to enforce some context requirements. So it would basically make sure you have all the related information, and then it would do a 1st draft.

120 00:14:56.420 00:14:59.789 Uttam Kumaran: This 1st draft, as we saw, is kind of long.

121 00:14:59.940 00:15:03.330 Uttam Kumaran: but I guess I wanted to ask like is, would this best

122 00:15:03.660 00:15:08.489 Uttam Kumaran: to just send in the thread for you to be like? That’s good. Write the notion.

123 00:15:08.600 00:15:13.559 Uttam Kumaran: Would you? Rather it draft a notion, and then, like come

124 00:15:13.940 00:15:18.960 Uttam Kumaran: back like the easiest thing for us to do is for us to just like then.

125 00:15:18.960 00:15:19.710 Matthew Good: And thread.

126 00:15:19.880 00:15:25.369 Uttam Kumaran: Send it in thread as like either set like several messages, probably given the length.

127 00:15:25.520 00:15:28.679 Uttam Kumaran: And then you could basically provide feedback.

128 00:15:29.630 00:15:29.970 Matthew Good: You know.

129 00:15:29.970 00:15:34.040 Uttam Kumaran: It’s literally like you could just speech a text, a bunch of feedback or type it in.

130 00:15:34.150 00:15:40.170 Uttam Kumaran: and then it could trigger you could. I think we’ll we could probably decide whether we want to have

131 00:15:40.450 00:15:46.680 Uttam Kumaran: multiple loops there, or it automatically just goes to notion on the second one. But that would be the goal.

132 00:15:47.160 00:15:49.620 Matthew Good: Yeah. I think the threat, the threats fine.

133 00:15:49.620 00:15:53.930 Uttam Kumaran: It’s mostly text I saw. So I don’t like cause

134 00:15:54.250 00:15:57.969 Uttam Kumaran: like, for example, tables and stuff are hard to do.

135 00:15:58.280 00:15:59.990 Uttam Kumaran: Yeah in slack thread.

136 00:16:00.537 00:16:02.740 Uttam Kumaran: But if it’s mo mainly text.

137 00:16:03.430 00:16:05.709 Uttam Kumaran: it might be really convenient, because.

138 00:16:05.710 00:16:06.050 Matthew Good: Yeah.

139 00:16:06.050 00:16:06.990 Uttam Kumaran: This mobile.

140 00:16:08.240 00:16:22.970 Matthew Good: Yeah, true. That is, yeah, that is really true. Actually, I didn’t think about mobile. Yeah, because when I’m not like building tables, I’ll make it like a design note sometimes. Just be like, Hey, let’s do this, but oftentimes that’ll be a notion as well, because the end to look like or like not the end, but like that v. 1 deliverable

141 00:16:23.250 00:16:28.760 Matthew Good: that I send over is the public notion page, that oftentimes it has my comments

142 00:16:28.870 00:16:34.260 Matthew Good: you know there about like, Hey, we could do this for design, or like design idea here or whatnot, but I think in like the initial

143 00:16:34.750 00:16:37.619 Matthew Good: initial one in slack thread. I think it’s probably the easiest.

144 00:16:38.050 00:16:38.620 Uttam Kumaran: Okay.

145 00:16:39.230 00:16:47.199 Uttam Kumaran: Okay? So then, yeah, we can do that in slack. And then once it yeah, let me.

146 00:16:50.340 00:16:56.979 Uttam Kumaran: And then, yeah, it’ll come back to notion. I think we probably just need access to like a notion area.

147 00:16:57.080 00:17:00.150 Uttam Kumaran: or maybe even just a database.

148 00:17:00.430 00:17:02.280 Uttam Kumaran: If you can give us that

149 00:17:02.693 00:17:06.219 Uttam Kumaran: that’d be great. Like, I could. Also, yeah, we mean, we could.

150 00:17:06.220 00:17:11.359 Rafay Iqbal: Yeah, I’ll give you access to our database, and the way I’ve set it up I can give you a brief little walk through as well.

151 00:17:11.520 00:17:12.420 Uttam Kumaran: Okay, okay,

152 00:17:14.530 00:17:17.160 Uttam Kumaran: So that would be great. And then.

153 00:17:17.390 00:17:26.009 Uttam Kumaran: yeah, I feel like, otherwise, like, I feel pretty good as long as we’re good on the form factor, I think, Mustafa, what else do we need like? Api, wise

154 00:17:29.450 00:17:29.910 Uttam Kumaran: blackout.

155 00:17:31.200 00:17:37.830 Mustafa Raja: Yeah, yeah, slack we. We would need to install a slack bot in the workspace.

156 00:17:40.010 00:17:40.890 Mustafa Raja: Yeah.

157 00:17:41.760 00:17:46.149 Mustafa Raja: And I feel with the notion. Api, that’s all we would need.

158 00:17:46.870 00:17:51.626 Uttam Kumaran: Okay? So yeah, maybe I mean, we could honestly use, can we use our

159 00:17:52.300 00:17:55.109 Uttam Kumaran: Our joint channel, Mustafa, for testing.

160 00:17:56.040 00:17:58.060 Mustafa Raja: Yeah, I can look into it.

161 00:17:58.700 00:18:02.099 Uttam Kumaran: So we can use our joint channel for testing.

162 00:18:02.430 00:18:08.609 Uttam Kumaran: And then, yeah, I mean, we could transfer the bought to their workspace

163 00:18:08.730 00:18:12.530 Uttam Kumaran: whenever like. Yes, that’s that’s the that’s the easiest thing.

164 00:18:14.410 00:18:17.600 Uttam Kumaran: Okay, great. And then we can set up Nadn on our side.

165 00:18:20.470 00:18:32.209 Uttam Kumaran: yeah, we have some things around like, these are things like on our side that we’ll start to basically create like a little prompt library in here. And then we’ll start to main, create a little bit of like a

166 00:18:32.330 00:18:37.149 Uttam Kumaran: an Eval data sheet. What? What basically we will do for any of our agents is we?

167 00:18:37.380 00:18:48.369 Uttam Kumaran: As soon as like a response happens. We compare and score it. Based on what you’ve sent to us is like, these are good ones. So we’ll build what’s called a golden data sheet with like.

168 00:18:48.570 00:18:54.860 Uttam Kumaran: here are here are like good examples that way. When AI output something

169 00:18:55.430 00:18:59.740 Uttam Kumaran: on like a 3rd party, AI will basically judge and give a score.

170 00:19:00.154 00:19:05.980 Uttam Kumaran: And so that way we can. We sort of have some maintenance on. We have some understanding on performance versus, just like.

171 00:19:06.210 00:19:09.400 Matthew Good: Yeah, looks. Looks decent. Yeah.

172 00:19:09.400 00:19:16.609 Uttam Kumaran: And there’s a couple of other. There’s a couple of other things like, we’re we’re looking at like speed. I guess. My, yeah, my other question is like.

173 00:19:17.300 00:19:18.899 Uttam Kumaran: Of course, like

174 00:19:19.280 00:19:26.801 Uttam Kumaran: I, for this would like to use some of the reasoning models just because they’re very good. But I guess I want to get your sense of like

175 00:19:28.975 00:19:31.370 Uttam Kumaran: like requirements on

176 00:19:31.720 00:19:40.220 Uttam Kumaran: sla between the bot, I mean right now the full process and the end is taking an hour, but if, like a bot was took like a couple of minutes to respond.

177 00:19:41.395 00:19:43.410 Matthew Good: One or 2 min is okay.

178 00:19:43.850 00:19:52.830 Matthew Good: I would rather have like a a better, because I use whatever whatever the most powerful. I have. Like Claude Max, I use 4, which I guess is their most.

179 00:19:53.440 00:20:00.429 Matthew Good: yeah, it’s like the most, the most powerful model. So I’m primarily in there. And even if that would take a little bit longer. That’s I’m that’s fine. I’m fine with that.

180 00:20:00.670 00:20:01.230 Uttam Kumaran: Okay.

181 00:20:04.330 00:20:25.080 Uttam Kumaran: yeah, that’s what I tell a lot of clients is like, you’re actually lucky if you don’t have like super time, sensitive asks because you can take advantage of these reasoning models that are way. Better. Additionally, like, just here, you say, like, Yeah, I have it and make like an Lp role like we will, Mustafa. We should build that like we’ll build another. We’re sort of build some Llm. As a judge from a couple of different angles.

182 00:20:25.733 00:20:30.140 Matthew Good: Something that is like, Yeah, an lp, and if we can think about other roles.

183 00:20:31.149 00:20:34.889 Uttam Kumaran: We can have it. Judge the output and provide feedback, and

184 00:20:35.180 00:20:37.900 Uttam Kumaran: we can almost just or give you that back.

185 00:20:38.660 00:21:06.450 Matthew Good: Yeah, that would be that because that’s super helpful, you know, for both of the like. Say, there’s 2 types of decks right? Like Vcs that are selling to Lps, and then founders selling Vc. So for the founder decks, too. I’ll be like, okay, put your Vc hat on right and I’ll give it like a pro in that initial prompt of like you understand blah blah like venture investing and unity economics. And you know, blah blah, and I’ll just say, evaluate this this narrative and and poke any holes in it that you can like be critical. And then it’ll do that

186 00:21:06.560 00:21:09.860 Matthew Good: sometimes they’re like, okay, I know I need to worry about this like this was like too much.

187 00:21:09.860 00:21:10.390 Uttam Kumaran: Yeah, yeah.

188 00:21:10.390 00:21:13.270 Matthew Good: Has come up with good good feedback that I’ll pass along to the founder.

189 00:21:13.680 00:21:18.426 Uttam Kumaran: Okay, okay, great.

190 00:21:21.050 00:21:29.960 Uttam Kumaran: yeah. We’ll start to measure like tokens. And like, kind of like, what the cost ends up being. But it it doesn’t seem like too much text. So I don’t actually think it’s going to be that bad.

191 00:21:30.715 00:21:34.359 Uttam Kumaran: And yeah, I’m trying to think if there’s anything

192 00:21:38.210 00:21:39.320 Uttam Kumaran: else.

193 00:21:47.500 00:21:50.837 Uttam Kumaran: yeah, I guess my other question was, gonna tell me a little bit about like

194 00:21:51.930 00:21:56.200 Uttam Kumaran: I guess we’ll sort of find this out together. But I’m I’m sort of interested to

195 00:21:56.470 00:21:58.430 Uttam Kumaran: to see what your

196 00:21:59.220 00:22:08.430 Uttam Kumaran: what your feedback is when you see the 1st drafts like right now you’re sort of you. You come up with the feedback in your head, and then you come up with the instruction.

197 00:22:08.810 00:22:15.380 Uttam Kumaran: But instead, it’s like, I want to hear, like, Okay, if if we could get structured feedback from you, whether it’s on

198 00:22:15.570 00:22:24.409 Uttam Kumaran: the tone like, I want to start to break it down, so it makes it easier for you to give the feedback, for example, if you get a block thing, and you’re like.

199 00:22:24.710 00:22:33.570 Uttam Kumaran: well, like it’s like, I guess, like walk me through like if you were to, let’s say an intern produce one of these for you like, what are the things that you’re looking for like when you read one of these.

200 00:22:34.150 00:22:34.910 Matthew Good: Hmm.

201 00:22:35.323 00:22:46.089 Matthew Good: yeah, I would say I would break feedback down to like 2 main buckets. And that’s what we told clients like one like the narrative, like at more like a 10,000 foot level of like, are there?

202 00:22:46.840 00:23:03.279 Matthew Good: Are we hitting, like every piece of the story that we want to hit. We might not hit it in the right way, like we might not like the way this headline looks. We might not like the way this, you know these bullets are framed. We might want to use active voice, supposed to passive voice, or whatever. But I want to make sure on like the 1st pass. What I’m trying to solve for is.

203 00:23:03.520 00:23:09.419 Matthew Good: are all the pieces of the story there? And I have those in my head generally like what those should be

204 00:23:10.220 00:23:22.070 Matthew Good: and then I’ll say, Okay, great. And then I’ll when I go into I’ll move that into notion, and I might do a little bit like clean up in terms of copy. If I see anything that’s like flagrantly like, okay, we’re using way. Too much passive voice or like

205 00:23:22.690 00:23:40.259 Matthew Good: whatever but I level. Set that with clients. I’m like when you review this like, just look at the narrative and make sure, like holy shit, we’re missing this this massive piece. That’s what I wanna find out. And then I don’t really get into like nitty, gritty copy editing of each line until we’re in Figma. Because then the the words are playing.

206 00:23:40.260 00:23:41.330 Uttam Kumaran: With design.

207 00:23:41.330 00:23:44.530 Matthew Good: And I want to be able to like, use real estate effectively and stuff like that.

208 00:23:46.140 00:23:53.490 Uttam Kumaran: Okay. So so like, real copy review happens. And Sigma.

209 00:23:54.300 00:24:00.509 Uttam Kumaran: oh, interesting. Okay? So it’s almost like at the top, like.

210 00:24:02.730 00:24:09.420 Uttam Kumaran: it’s almost like there’s probably something between the questionnaire, the transcript and this final thing that you need at the top of the stock, which is like

211 00:24:09.790 00:24:11.709 Uttam Kumaran: the key points we’re hitting here

212 00:24:12.380 00:24:16.019 Uttam Kumaran: right? Because I’m sure that a lot of I mean just knowing

213 00:24:16.160 00:24:21.369 Uttam Kumaran: like I, I used to work in product and stuff, just knowing people get hooked on the language.

214 00:24:22.210 00:24:29.169 Uttam Kumaran: Probably like way too much, and so there almost could be a benefit of like at the top of each of these to have, like

215 00:24:29.370 00:24:32.009 Uttam Kumaran: literally that what you describe which is like

216 00:24:32.130 00:24:36.809 Uttam Kumaran: here’s the extraction of like the key themes. Is there anything missing here.

217 00:24:37.390 00:24:41.959 Matthew Good: Yeah, and I and I will. I’ve now kind of as I’ve evolved my prompting. I’ll

218 00:24:42.070 00:24:49.459 Matthew Good: and often if you saw in the in the notion, Doc, but I’ll have Claude put in the little rationale underneath each slide, so it’d be like the copy of like headline.

219 00:24:49.460 00:24:49.900 Uttam Kumaran: Yeah, yeah.

220 00:24:49.900 00:25:06.600 Matthew Good: Subheader body, and then, like a 1 sentence, rationale, and then I’ll have Claude also say, like, Okay, Meta level, tell me why you structured everything the way you did, because, like what we’re doing is like, there’s all these inputs. And it would take me forever to go through the Transcript. Go back to whatever Claude I’m like using as a way to like.

221 00:25:06.680 00:25:28.329 Matthew Good: If all these pieces are on the floor like scattered around, at least build something that has like a semblance of like directionally right. And then I can like take it and review and and fine tune it. But like that, that process of going from, like all the pieces are on the floor scattered around to like. Okay, at least, we have, like a directionally right. v, 1 is the most time intensive. And that’s what I’m using claw for.

222 00:25:28.950 00:25:30.879 Uttam Kumaran: And can you tell me, like what

223 00:25:31.574 00:25:35.019 Uttam Kumaran: what themes are sometimes missed like

224 00:25:35.230 00:25:40.800 Uttam Kumaran: are like what is what is sometimes not there, that they missed to say like they’ve that.

225 00:25:40.800 00:25:41.119 Matthew Good: You tell me.

226 00:25:41.120 00:25:41.690 Uttam Kumaran: Sharing.

227 00:25:41.990 00:25:43.350 Matthew Good: The founder, or.

228 00:25:43.350 00:25:49.650 Uttam Kumaran: No, no like you. You mentioned like, hey? Make sure that the top themes, or whatever. Here can you give an example of something that’s missed like.

229 00:25:49.810 00:25:54.964 Uttam Kumaran: I guess I’m trying to think like, what are they missing like a product announcement? Are they missing like, hey? We

230 00:25:55.470 00:25:58.509 Uttam Kumaran: like, I guess. Like. Give me some examples of like what gets missed.

231 00:25:59.030 00:26:02.119 Matthew Good: Like when I’m reviewing that v. 1 that Claude gives to me.

232 00:26:02.360 00:26:07.970 Uttam Kumaran: Yeah, but also what? No, but not actually not that. When you get send it to the client.

233 00:26:09.020 00:26:12.420 Uttam Kumaran: and you tell the client, hey? Make sure all the high level themes are here.

234 00:26:12.686 00:26:14.549 Matthew Good: I see! I see what you mean.

235 00:26:14.550 00:26:18.999 Uttam Kumaran: Are they having examples of some feedback that comes back before you? You get the check mark.

236 00:26:19.000 00:26:23.200 Matthew Good: Yeah, yeah, yeah. Oftentimes. Yeah.

237 00:26:23.470 00:26:26.070 Matthew Good: I mean, we’ve been doing this for a while. There’s not like.

238 00:26:26.740 00:26:29.610 Uttam Kumaran: Like, is it? Is it like we’re missing products like what like give me like, give me.

239 00:26:29.610 00:26:30.030 Matthew Good: Basically.

240 00:26:30.030 00:26:31.929 Uttam Kumaran: What, what or they’re like. I don’t.

241 00:26:32.620 00:26:33.130 Uttam Kumaran: We’re like.

242 00:26:33.130 00:26:33.770 Matthew Good: Normal Taste.

243 00:26:33.770 00:26:35.792 Uttam Kumaran: Slide like? What are they missing?

244 00:26:36.130 00:26:52.230 Matthew Good: Yeah, yeah, the slides are pretty straightforward, I think. So. We don’t. We never get feedback of like, we’re missing Tam, or we’re missing problem. It’s more so like, especially at the start of like setting up the framing, because, like everything, I mean, communication is just framing. So we’re trying to make sure that we’re not just like problem solution whatever, but like

245 00:26:52.800 00:26:59.440 Matthew Good: how the 1st couple of slides and set up like flow relative to the product and relative to

246 00:26:59.950 00:27:08.430 Matthew Good: the industry that they’re playing in relative to like macroeconomic and stuff that’s going on in the world. So oftentimes the feedback will not will be like.

247 00:27:08.890 00:27:12.549 Matthew Good: This isn’t like entirely wrong, but I think we should order it this way.

248 00:27:13.410 00:27:14.350 Uttam Kumaran: Okay.

249 00:27:14.570 00:27:17.120 Matthew Good: And then I’ll be like, okay, interesting like, let’s talk about that.

250 00:27:18.210 00:27:21.766 Uttam Kumaran: I see. Okay, that makes more sense. Yeah. So it’s kind of like,

251 00:27:23.250 00:27:27.219 Matthew Good: Not like, Oh, we don’t have a solution. But like, what if? Yeah.

252 00:27:27.630 00:27:36.509 Uttam Kumaran: Like, yeah, problem solutions that they’re like, Oh, maybe we should back up even one step further and explain like this whole industry, or explain who came before us.

253 00:27:36.870 00:27:37.739 Matthew Good: Exactly. That’s a good one.

254 00:27:37.740 00:27:41.290 Uttam Kumaran: It’s about the company, because maybe, like

255 00:27:41.520 00:27:46.080 Uttam Kumaran: their product hasn’t. There’s like no market for it, or like you’ve never heard of.

256 00:27:46.080 00:27:46.859 Matthew Good: Free product

257 00:27:47.200 00:27:54.500 Matthew Good: product, or it’s super obscure, like, people are like what the fuck we’re even talking about copper wiring. I don’t know anything about that. Like, let’s have a slide that talks about

258 00:27:54.880 00:27:57.799 Matthew Good: the history of copper. Sure. Okay, yeah, yeah.

259 00:27:57.800 00:28:03.259 Uttam Kumaran: Okay, okay, great, that’s actually, that’s actually helpful. Meaning the feedback that blocks you from moving to the figma step.

260 00:28:03.440 00:28:07.710 Uttam Kumaran: It’s typically overall ordering and narrative, and, in fact, I think.

261 00:28:08.240 00:28:16.789 Uttam Kumaran: What’s helpful is at the top. I want to have the AI basically like, give you like A,

262 00:28:17.110 00:28:28.669 Uttam Kumaran: you know, when you look at this. And you’re like, Hey, I always have like 5 questions. I wanted to answer those like, why, like, what is rereading this like? What is the kind of like narrative flow. I wanted to give that at the top. So that

263 00:28:29.250 00:28:32.240 Uttam Kumaran: one, I think, for when you go send this to a client

264 00:28:32.410 00:28:35.680 Uttam Kumaran: frankly like that is what they should start looking at

265 00:28:35.820 00:28:48.370 Uttam Kumaran: first, st because if they get jammed there, then there’s no reason you need to go slide by slide. Second, when you get the update from AI, you can read that 1st and be like, okay, it’s it’s gonna be on point for for the.

266 00:28:49.295 00:28:50.220 Matthew Good: Exactly.

267 00:28:50.950 00:28:56.800 Uttam Kumaran: Okay, cool. So then I think that’s kind of all the questions I had. We have all this stuff from you. I think we only need the

268 00:28:57.350 00:29:03.950 Uttam Kumaran: the notion. I don’t know. Ralph, if you want to share or tell me where to go, I can just double check that I have access to that.

269 00:29:05.522 00:29:10.419 Rafay Iqbal: Yeah, I’m gonna have to give you access to a few pages here, so you’ll probably see that come through after this call.

270 00:29:11.450 00:29:16.920 Uttam Kumaran: Okay, okay, okay, perfect. Yeah. Basically, we’ll just what we’re gonna do is I’ll tell you, if we need

271 00:29:17.040 00:29:22.620 Uttam Kumaran: Mustafa, what Api access do we need for a notion? I forgot how we get got that for ours.

272 00:29:24.590 00:29:27.019 Uttam Kumaran: Is there something in admin that we have to do.

273 00:29:29.530 00:29:34.150 Mustafa Raja: Yeah, I can look into it if you everyone.

274 00:29:34.810 00:29:38.080 Uttam Kumaran: Yeah, if you can look into it, and then we can send instruction if they need to get us a key.

275 00:29:38.080 00:29:39.090 Mustafa Raja: Oh, yeah.

276 00:29:39.430 00:29:46.773 Uttam Kumaran: I think we oh, yeah, it’s actually in connections. I feel like we have to create an integration or something. But yeah, if you could just get. Send me whatever we need to do.

277 00:29:47.310 00:29:57.010 Uttam Kumaran: And then, yeah, I think, I don’t know, Mustafa. What do you think? Like any questions I feel like. Probably by like mid next week, we can have something in slack that’s like dropping something into notion, and we could

278 00:29:57.550 00:29:58.380 Uttam Kumaran: test.

279 00:30:02.190 00:30:04.420 Mustafa Raja: Yeah. Mid. Next week. Looks good.

280 00:30:05.440 00:30:15.809 Matthew Good: Yeah, no, no rush, yeah. And and like, let me know, I’ll be traveling tonight. I’ll be in Michigan for the next week, like just doing some family stuff. But I’ll be around like in the mornings at night. So if there’s anything you need or like.

281 00:30:16.400 00:30:17.690 Matthew Good: that’s just something.

282 00:30:18.230 00:30:21.060 Uttam Kumaran: Yeah, I’ll keep something standing just in the middle of the week.

283 00:30:21.490 00:30:21.910 Matthew Good: Cool.

284 00:30:22.143 00:30:27.286 Uttam Kumaran: Just so we have the time. And then, yeah, I mean, we’ll send slacks and looms as we as we get through.

285 00:30:27.760 00:30:30.849 Uttam Kumaran: I think by Wednesday or or so, or Thursday we should have.

286 00:30:31.610 00:30:38.500 Uttam Kumaran: We should have, like all the systems hooked up. Then it’ll be just like starting to break down the agents a little bit more.

287 00:30:40.230 00:30:44.530 Uttam Kumaran: Yeah, but this is, this is cool. I feel like it’s it’s a good scope, like I think we should nail it.

288 00:30:45.070 00:30:49.029 Matthew Good: Awesome dude. We’re excited about this. Thank you guys for all the help here.

289 00:30:49.030 00:30:54.049 Uttam Kumaran: Of course. Yeah. And as we’re as you’re interacting and seeing some of the capabilities, think about some more stuff.

290 00:30:54.570 00:30:56.560 Matthew Good: Oh, yeah, I have long the list of stuff

291 00:30:57.720 00:31:01.129 Matthew Good: we’re we’re dying dude. We’re just talking about this morning.

292 00:31:01.130 00:31:10.530 Uttam Kumaran: Yeah. But also, like, I think, you know, even my business like, you want to get someone. It’s gonna be expensive to hire you guys, right? So you want to find someone who can use this

293 00:31:11.100 00:31:13.079 Uttam Kumaran: and then basically get there.

294 00:31:13.680 00:31:23.119 Matthew Good: Dude your post. When I texted you over the weekend about like a sales coordinator. I was like dude. Why did I not think of this, like Jesus. I was sitting at like my hotel, having breakfast. I was like, I gotta text him like right now.

295 00:31:23.320 00:31:34.109 Uttam Kumaran: Yeah, I don’t know. Man, I literally is like, I think that’s what our move is going to be, because people still want to talk to me, and I’m still the best at telling what we got to do. So crisp. And

296 00:31:34.380 00:31:38.419 Uttam Kumaran: I just it’s just everything around the sales process that’s so brutal

297 00:31:40.280 00:31:43.269 Uttam Kumaran: cause. I could go sit in 8 h like I could go dance meeting.

298 00:31:43.270 00:31:43.630 Matthew Good: Yeah.

299 00:31:43.630 00:31:44.640 Uttam Kumaran: Me and go do that.

300 00:31:44.810 00:31:55.430 Uttam Kumaran: I’m down to do that. But it’s everything else has to happen. Preparing stuff like that’s the stuff we really didn’t. And really, honestly, I’m hoping that one of our coordinators can graduate into like starting

301 00:31:55.430 00:31:59.260 Uttam Kumaran: didn’t hear that over time. I think that’s a natural flow of the folks want that, you know.

302 00:31:59.450 00:32:00.000 Matthew Good: Yeah.

303 00:32:00.000 00:32:06.660 Matthew Good: yeah, yeah, absolutely. But yeah, no, that hit me like a lightning bolt. I was like, yeah, it’s it’s like death by a thousand paper cuts of the follow up and notion

304 00:32:06.990 00:32:08.019 Matthew Good: updating my Cr.

305 00:32:08.020 00:32:11.641 Uttam Kumaran: No, I know, and it’s so hard because I want to be so organized.

306 00:32:11.900 00:32:12.509 Matthew Good: Yeah, same.

307 00:32:12.830 00:32:13.380 Uttam Kumaran: It’s

308 00:32:14.040 00:32:18.754 Uttam Kumaran: so it’s brutal. That’s why, when I asked you like, Hey, how long does this take in the context of everything you’re doing.

309 00:32:19.290 00:32:19.900 Matthew Good: Yeah.

310 00:32:20.160 00:32:22.259 Uttam Kumaran: It seems so simple from the outside.

311 00:32:22.932 00:32:27.999 Uttam Kumaran: But it’s just given, like all the other inputs, you know, like.

312 00:32:28.000 00:32:28.440 Matthew Good: Yeah.

313 00:32:28.440 00:32:28.950 Uttam Kumaran: Alright!

314 00:32:28.950 00:32:36.310 Matthew Good: Yeah, yeah, I hear you. Alright, man. Thank you for this. We’ll be in touch. We’ll stay in touch over slack and text me, if anything, but we appreciate it.

315 00:32:36.310 00:32:39.420 Uttam Kumaran: Okay. Alright, thanks. Guys. Bye, guys, yeah. Bye.