Meeting Title: Uttam <> Scott Date: 2024-09-30 Meeting participants: Joshuadeveyra, Uttam Kumaran


WEBVTT

1 00:01:07.310 00:01:08.010 Scott_Harmon: Alright!

2 00:07:23.910 00:07:25.009 joshuadeveyra: Hey! Scott! Good morning!

3 00:07:26.630 00:07:27.920 Scott_Harmon: Good morning, Joshua.

4 00:07:28.390 00:07:28.930 Scott_Harmon: Yep.

5 00:07:28.930 00:07:33.779 joshuadeveyra: Morning morning. Yeah, I’ll just contact utam. Give me a moment.

6 00:07:34.050 00:07:36.540 Scott_Harmon: I just did. He said he’d be on in 1 min.

7 00:07:37.290 00:07:41.070 joshuadeveyra: Oh, yeah, I actually thought the meeting was on 2 30. My time.

8 00:07:41.320 00:07:42.720 joshuadeveyra: I was a bit confused.

9 00:07:42.720 00:07:43.440 Scott_Harmon: First.st

10 00:07:43.710 00:07:44.920 Scott_Harmon: Where are you at?

11 00:07:45.360 00:07:47.740 joshuadeveyra: Manila. It’s currently 2 Am.

12 00:07:48.010 00:07:49.270 Scott_Harmon: Oh, my! Gosh!

13 00:07:49.400 00:07:56.669 joshuadeveyra: Yeah, it’s like 2. Then I checked my calendar. Oh, it’s actually 2 2 am. I was like, Oh, my God! Sorry for being late!

14 00:07:56.670 00:07:59.509 Scott_Harmon: Do you normally just adjust your time. So you’re awake

15 00:07:59.900 00:08:00.520 Scott_Harmon: this time.

16 00:08:00.520 00:08:04.910 joshuadeveyra: Yeah, yeah, I usually, yeah, usually sleep around 4 or 5 am.

17 00:08:04.980 00:08:07.460 joshuadeveyra: so I could, you know, be online alongside Tuta.

18 00:08:08.090 00:08:08.660 Scott_Harmon: Yeah.

19 00:08:10.900 00:08:12.439 Scott_Harmon: how are things in Manila?

20 00:08:14.014 00:08:18.419 joshuadeveyra: Honestly, I I’m not sure cause I I rarely go out of my house.

21 00:08:18.610 00:08:19.170 joshuadeveyra: Oh.

22 00:08:19.170 00:08:19.940 Scott_Harmon: Oh!

23 00:08:20.710 00:08:22.370 joshuadeveyra: I just stay indoors.

24 00:08:23.090 00:08:25.240 Scott_Harmon: Running out of my house.

25 00:08:25.270 00:08:29.629 Scott_Harmon: It’s a very big I’ve only been once. It’s a very big and bustling

26 00:08:30.050 00:08:31.770 Scott_Harmon: city. There’s so much.

27 00:08:32.179 00:08:33.099 joshuadeveyra: Yeah, very.

28 00:08:34.980 00:08:35.570 Scott_Harmon: Perfect.

29 00:08:35.730 00:08:36.280 Scott_Harmon: Yeah.

30 00:08:36.289 00:08:37.549 joshuadeveyra: Have you been to Manila?

31 00:08:37.830 00:08:39.104 Scott_Harmon: One time. Yeah.

32 00:08:39.659 00:08:42.519 Uttam Kumaran: Hey? Guys? Sorry about the delay just running around.

33 00:08:42.520 00:08:43.130 Scott_Harmon: Hey?

34 00:08:43.320 00:08:45.890 Scott_Harmon: No problem. We’re just talking about Manila.

35 00:08:45.890 00:08:46.420 Uttam Kumaran: Oh, okay.

36 00:08:46.420 00:08:49.440 Scott_Harmon: One time, very, very briefly.

37 00:08:49.440 00:08:49.760 Uttam Kumaran: Cool.

38 00:08:49.760 00:08:50.450 Scott_Harmon: That.

39 00:08:50.450 00:08:50.959 Uttam Kumaran: As you go.

40 00:08:50.960 00:08:51.599 Scott_Harmon: Like correct.

41 00:08:51.850 00:08:54.719 Scott_Harmon: Oh, years ago, you know, probably 15 years ago.

42 00:08:55.020 00:08:55.420 joshuadeveyra: Oh!

43 00:08:56.660 00:09:02.730 Scott_Harmon: and and I was there for just like part of a day, just, you know, really around the airport and in one meeting, and

44 00:09:03.140 00:09:03.780 Scott_Harmon: and then back

45 00:09:04.070 00:09:05.070 Scott_Harmon: and out again.

46 00:09:05.450 00:09:06.250 Scott_Harmon: Yeah.

47 00:09:06.820 00:09:08.939 Scott_Harmon: but it’s it’s

48 00:09:09.670 00:09:10.929 Scott_Harmon: kind of crazy.

49 00:09:12.470 00:09:14.476 Uttam Kumaran: It seems

50 00:09:16.680 00:09:20.852 Uttam Kumaran: I’ve been to I mean Thailand. I’ve been to Thailand, but Thailand was also.

51 00:09:21.730 00:09:22.620 Uttam Kumaran: It’s.

52 00:09:28.990 00:09:29.570 Scott_Harmon: Yeah, is it?

53 00:09:29.570 00:09:30.059 joshuadeveyra: I’m just gonna be.

54 00:09:30.060 00:09:30.440 Scott_Harmon: Set.

55 00:09:31.443 00:09:32.446 Uttam Kumaran: No.

56 00:09:34.090 00:09:37.440 Scott_Harmon: Utah. I think your cell reception is a little funny. Are you on your mobile.

57 00:09:38.412 00:09:41.200 Uttam Kumaran: Yeah, give me one second. Let me switch. Let me switch over.

58 00:09:41.200 00:09:43.289 joshuadeveyra: Oh, yeah, I thought it was on my end.

59 00:09:43.290 00:09:44.470 Scott_Harmon: No, no.

60 00:10:25.010 00:10:28.550 joshuadeveyra: So any plans to like come back to Manila anytime soon.

61 00:10:29.560 00:10:32.226 Scott_Harmon: No, I don’t have any travel plans, but

62 00:10:33.050 00:10:41.270 Scott_Harmon: with my the job I used to have. I used to travel internationally. I used to be a CEO, of A, of a Midsize company, and so.

63 00:10:41.660 00:10:42.909 Scott_Harmon: and we had.

64 00:10:42.960 00:10:45.000 Scott_Harmon: We had offices in

65 00:10:45.400 00:10:46.900 Scott_Harmon: the Pacific rim.

66 00:10:47.340 00:10:48.460 joshuadeveyra: And so.

67 00:10:48.460 00:10:52.730 Scott_Harmon: You know. I had to go there several times a year, you know, just on business.

68 00:10:53.682 00:10:58.149 Scott_Harmon: You know, we had an office in Australia, one in Japan and

69 00:10:58.650 00:11:00.880 Scott_Harmon: one in Hong Kong, briefly.

70 00:11:01.440 00:11:02.680 Scott_Harmon: And

71 00:11:02.810 00:11:05.440 Scott_Harmon: so, you know, I would go over there a couple of times a year.

72 00:11:06.130 00:11:12.689 Scott_Harmon: you know, and then, when I would go, I would like, oh, I’ll take a quick side jaunt to see this town or the city, or maybe have a meeting.

73 00:11:13.120 00:11:15.169 Scott_Harmon: you know in this city.

74 00:11:15.350 00:11:17.499 Scott_Harmon: but that was that was

75 00:11:17.770 00:11:20.509 Scott_Harmon: 15 years ago, so I don’t have the

76 00:11:21.660 00:11:25.989 Scott_Harmon: the occasion to go over there as often as I as I used to. But I.

77 00:11:26.390 00:11:27.300 Scott_Harmon: And

78 00:11:28.310 00:11:34.539 Scott_Harmon: obviously, it’s a lot. It’s a long travel. So when I would travel, I would probably go okay, because it’s so such a long trip.

79 00:11:34.680 00:11:35.470 Scott_Harmon: yeah.

80 00:11:35.470 00:11:36.589 joshuadeveyra: 4 HI think.

81 00:11:38.210 00:11:42.159 Scott_Harmon: Yeah, if once you make a connection, and you know you do

82 00:11:42.170 00:11:46.230 Scott_Harmon: add in the connection in San Francisco or la, then it’s usually.

83 00:11:46.800 00:11:51.240 Scott_Harmon: you know. Yeah, usually like 20 h, you know, by the time it’s all done.

84 00:11:52.370 00:11:53.610 Scott_Harmon: Yeah. So.

85 00:11:53.610 00:11:58.960 joshuadeveyra: I was planning to go States, too. But then I was like 20 h on a plane. Yeah, we could probably

86 00:11:59.584 00:12:08.709 joshuadeveyra: like the furthest place. I’ve I the longest time I’ve been on a plane was like 9 HI think that was a connecting flight to, I think, Barcelona.

87 00:12:10.470 00:12:15.689 joshuadeveyra: yeah, yeah, Manila to. I think we dropped. We stopped over at around Dubai or something.

88 00:12:16.324 00:12:19.780 joshuadeveyra: Yeah, my, my feet was like, yeah, it was it.

89 00:12:19.780 00:12:21.410 Scott_Harmon: Yeah, yeah, it’s true.

90 00:12:21.550 00:12:23.669 Scott_Harmon: It’s very true. It’s the

91 00:12:23.860 00:12:26.620 Scott_Harmon: just even even sitting.

92 00:12:27.100 00:12:29.249 Scott_Harmon: you know, sitting that long

93 00:12:29.800 00:12:32.460 Scott_Harmon: can. Can really, you know.

94 00:12:32.510 00:12:34.872 Scott_Harmon: mess up your legs and your back.

95 00:12:36.210 00:12:38.890 Scott_Harmon: You don’t feel don’t feel like a human being.

96 00:12:38.890 00:12:44.070 Uttam Kumaran: When I go. When I go to India it’s like 22 h, 24 h.

97 00:12:44.620 00:12:46.800 Uttam Kumaran: It’s like ridiculous.

98 00:12:47.110 00:12:51.350 Scott_Harmon: The Indian, and that’s what you go through. Frankfurt is that I’m usually go through Europe. Central.

99 00:12:51.350 00:12:54.700 Uttam Kumaran: Yeah, we go through yeah, or go through Singapore.

100 00:12:55.610 00:12:57.339 Scott_Harmon: Oh, really go the out the other way. Okay.

101 00:12:57.340 00:13:01.489 Uttam Kumaran: Yeah, when when I used to live in the Bay area, it goes through Singapore. Yeah.

102 00:13:01.490 00:13:03.300 Scott_Harmon: Gotcha gotcha.

103 00:13:03.940 00:13:08.382 Uttam Kumaran: Okay, I guess. Let’s like, let’s let’s take some stuff off. I know.

104 00:13:08.740 00:13:13.250 Uttam Kumaran: I’m glad that you and Miguel already have been talking. But Miguel is on my team.

105 00:13:13.744 00:13:17.385 Uttam Kumaran: Has basically been, you know, helping me kind of

106 00:13:17.850 00:13:43.099 Uttam Kumaran: do all the AI stuff that I’ve been wanting to do you know, and both do a lot of internal agent building, you know, for Brainforge but also do proof of concepts and prototypes. You know as we’re going after new clients. So we’ve both together learned a ton about how to actually productionize some of these systems. And I’ve been spending some time, you know, working on the Hpi use case.

107 00:13:43.472 00:13:50.210 Uttam Kumaran: So I guess we I guess we could just jump right in. I don’t know, Scott, if you had a chance to check out that loom video we sent.

108 00:13:50.250 00:13:54.570 Uttam Kumaran: If not, we can just walk through like a great basic demo of that, too.

109 00:13:54.950 00:14:03.499 Scott_Harmon: I didn’t. The thing I was. I I like, I said I went into town and kind of went offline for the last 5 days. But so I haven’t looked at the video. I did just go through the

110 00:14:03.640 00:14:05.050 Scott_Harmon: the analysis

111 00:14:05.507 00:14:09.439 Scott_Harmon: of the summary of change, you know, changes between the versions.

112 00:14:09.440 00:14:09.930 Uttam Kumaran: Yeah.

113 00:14:09.930 00:14:12.850 Scott_Harmon: Which I, which I thought was super interesting, and

114 00:14:13.030 00:14:16.870 Scott_Harmon: would would love to hear you, you know. Comment about.

115 00:14:17.130 00:14:23.630 Scott_Harmon: you know. Maybe dig into that one a little bit, because I think it was really cause. That’s kind of step one is just to get the

116 00:14:24.460 00:14:26.400 Scott_Harmon: the change history

117 00:14:26.660 00:14:27.780 Scott_Harmon: memorialize.

118 00:14:27.780 00:14:28.700 Uttam Kumaran: Yeah, exactly.

119 00:14:28.700 00:14:29.970 Scott_Harmon: Yeah, I think that.

120 00:14:29.970 00:14:38.449 Uttam Kumaran: Miguel. Maybe you want to just show briefly, like the Re, the ui of relevance, and then you can just pull up the mirror board, and we can talk a little bit about

121 00:14:38.540 00:14:39.200 Uttam Kumaran: like

122 00:14:39.890 00:14:48.580 Uttam Kumaran: the format. But you’re exactly right, like one of the things that we got a ton of documents from them we took a look at Bunch, and so we were just like scope down, scope down, until like, let’s take one

123 00:14:48.610 00:14:49.890 Uttam Kumaran: office lease

124 00:14:49.990 00:15:01.989 Uttam Kumaran: with like the 7 revisions, and basically see if we can 1st identify like the changes between each and then I think you’ll be interested in the kind of the way we wanted to architecture this like multi agent setup

125 00:15:02.473 00:15:08.676 Uttam Kumaran: but basically we arrived at like having one agent per like major clause or major

126 00:15:09.200 00:15:25.510 Uttam Kumaran: like section of the document, and then also having, like an exception agent that has a little bit of understanding of like. Is this a strategic client like? Is this a area we should grant exceptions. Again, like all this is just like throwing stuff at the wall and seeing what sticks. But we’ve been using this tool

127 00:15:25.910 00:15:30.670 Uttam Kumaran: called relevance, which is basically like an agent builder. That’s been like

128 00:15:30.940 00:15:36.620 Uttam Kumaran: super super great for us internally, that we wanted to see whether it would work for the system as well.

129 00:15:36.800 00:15:39.079 Uttam Kumaran: But yeah, Miguel, maybe if you want to share

130 00:15:39.580 00:15:44.580 Uttam Kumaran: your thoughts there and then, we could just like, I just want to have something on the screen that we can kind of talk to.

131 00:15:48.385 00:15:49.299 Uttam Kumaran: You’re on mute.

132 00:15:54.520 00:15:58.370 joshuadeveyra: Oh, yeah, sorry. I was just founding. I was just finding the link.

133 00:15:59.000 00:16:02.680 joshuadeveyra: Okay, yeah, I have it now. Yeah, let me just share screen.

134 00:16:03.120 00:16:07.239 Scott_Harmon: And rel is relevance its own model? Or is it a tool set built on top of somebody.

135 00:16:07.240 00:16:10.359 Uttam Kumaran: Tool. It’s a tool set built on top. Yeah, tool set.

136 00:16:10.360 00:16:10.960 Scott_Harmon: Gotcha.

137 00:16:13.656 00:16:16.949 joshuadeveyra: Wait, give me a moment. And how do I?

138 00:16:17.390 00:16:19.359 joshuadeveyra: Entire screen? Yeah.

139 00:16:19.760 00:16:21.350 joshuadeveyra: there you go. Can you guys see it?

140 00:16:21.700 00:16:23.200 joshuadeveyra: Yep, yeah.

141 00:16:23.780 00:16:24.919 joshuadeveyra: okay, yeah.

142 00:16:24.940 00:16:27.660 joshuadeveyra: so basically, here is like, relevance

143 00:16:28.500 00:16:31.199 joshuadeveyra: Utam, do you want me to go through this or.

144 00:16:31.200 00:16:34.354 Uttam Kumaran: Yeah, maybe. Just walk through one example, and then let’s

145 00:16:34.690 00:16:42.347 Uttam Kumaran: Let’s poke at the mirror board and at the notion. But yeah, just walk through like one example of what you did, and with a knowledge base like Scott’s, familiar with all the

146 00:16:43.298 00:16:46.690 Uttam Kumaran: like the technical capability stuff. So you can go as detailed.

147 00:16:47.310 00:16:52.210 joshuadeveyra: Okay, yeah. So wait. Sorry. Let me, just.

148 00:16:53.890 00:16:56.000 joshuadeveyra: okay. Okay. Yeah. So

149 00:16:56.410 00:16:58.809 joshuadeveyra: wait. Let me just find a good example.

150 00:17:07.650 00:17:08.700 joshuadeveyra: Okay, yeah.

151 00:17:08.859 00:17:16.070 joshuadeveyra: So one of the what do you call this lease agreements? That was there was between, you know.

152 00:17:16.690 00:17:19.000 joshuadeveyra: Barton Creek and this guy.

153 00:17:19.550 00:17:32.529 joshuadeveyra: And then the changes from basically v. 1. There was some skips. V. 2. Wasn’t there? V. 4. Wasn’t there? V. 6. Wasn’t there? But then v. 1, 3, 5, and 7 was there. So we were able to, you know.

154 00:17:32.540 00:17:33.690 joshuadeveyra: analyze it

155 00:17:34.317 00:17:40.979 joshuadeveyra: one like from v. 1. What changed to v, 3, and from v, 3 to v. 5, and v, 5 to v, 7.

156 00:17:41.820 00:17:44.900 joshuadeveyra: And then BA-, basically what I did.

157 00:17:45.070 00:17:45.710 joshuadeveyra: Oh, my God.

158 00:17:45.880 00:17:58.539 Scott_Harmon: And I’m sorry to get. Did you do it? Just technically, I’m curious. Did you do that by analyzing all all the text and just sort of doing a diff, or did you find the change in the Microsoft word

159 00:17:58.920 00:18:02.230 Scott_Harmon: metadata? Will highlight sort of change.

160 00:18:03.200 00:18:04.859 joshuadeveyra: I had AI do it.

161 00:18:05.640 00:18:10.469 Scott_Harmon: So you just took the 2 documents and just said, Show me the diffs. Basically, you did a diff.

162 00:18:11.566 00:18:12.400 joshuadeveyra: Yes, yes.

163 00:18:12.560 00:18:15.160 joshuadeveyra: in in layman’s terms. Yeah, that’s what I did.

164 00:18:15.160 00:18:18.099 Uttam Kumaran: We have we? I think we have those that metadata.

165 00:18:18.260 00:18:19.150 Uttam Kumaran: But

166 00:18:19.320 00:18:23.000 Uttam Kumaran: I think in this situation we just looked at the depths between the 3.

167 00:18:23.100 00:18:25.180 Scott_Harmon: Yeah, I mean, it’s probably a.

168 00:18:25.180 00:18:25.989 Uttam Kumaran: It’s the same thing.

169 00:18:25.990 00:18:30.300 Scott_Harmon: But I was just curious whether or not you’d get any better or worse results.

170 00:18:30.370 00:18:34.809 Scott_Harmon: you know, by using the metadata, or just basically just looking at the doc. You know the

171 00:18:35.200 00:18:37.570 Scott_Harmon: the the versions themselves are just.

172 00:18:37.750 00:18:40.389 Scott_Harmon: you know, having AI say, show me the differences.

173 00:18:42.180 00:18:48.119 Scott_Harmon: I don’t know that there’s any meaningful, you know, plus or minus between those 2 approaches.

174 00:18:52.040 00:18:53.199 joshuadeveyra: Okay, yeah.

175 00:18:54.430 00:19:00.850 joshuadeveyra: So yeah, basically, you know, this agent can take in like a lease, and then it’ll use. I can

176 00:19:01.070 00:19:06.659 joshuadeveyra: prob, technically, what do you call this like? Read that lease agreement. And then

177 00:19:07.160 00:19:11.800 joshuadeveyra: what we basically did was, for example, we wanna change the 10 name right?

178 00:19:12.170 00:19:15.989 joshuadeveyra: And then we’ll have, like a couple of sub agents to do that.

179 00:19:16.100 00:19:23.449 joshuadeveyra: So legal entity, if you wanna change, you know. Is it acceptable? Is it non acceptable? Or do we need like more clarifications.

180 00:19:23.450 00:19:27.850 Uttam Kumaran: Yeah. And then just show like, yeah, show a little bit of like, the prompt for this agent.

181 00:19:28.050 00:19:31.740 Uttam Kumaran: Oh, yeah. The reason the reason why we like using

182 00:19:32.590 00:19:37.570 Uttam Kumaran: using relevance is that you can set up these agents, but also tell it to like use another agent.

183 00:19:38.066 00:19:42.809 Uttam Kumaran: In X situation. Use a tool in Y situation? Can you actually open the sub agent?

184 00:19:43.010 00:19:43.730 Uttam Kumaran: Oh, yeah.

185 00:19:43.730 00:19:44.370 joshuadeveyra: For sure.

186 00:19:44.820 00:19:45.830 Uttam Kumaran: Prompt 2.

187 00:19:46.843 00:19:48.339 joshuadeveyra: Which one? Do you want this one.

188 00:19:48.340 00:19:49.410 Uttam Kumaran: Yeah, yeah.

189 00:19:49.950 00:19:59.019 Uttam Kumaran: so this is like, you would like this sub agent we had called. And if in case there’s a question about the security deposit. So if you go to the prompt for this

190 00:20:00.805 00:20:08.904 Uttam Kumaran: you can see that like, okay, here like this is exactly like what the fairway dock is for the security deposit.

191 00:20:09.430 00:20:10.899 Uttam Kumaran: you know, portion of the Doc.

192 00:20:11.310 00:20:15.079 Scott_Harmon: How’d you get this? How was this? This generated Uton.

193 00:20:17.040 00:20:20.040 Uttam Kumaran: Miguel, did you just generate like an example of this on your side?

194 00:20:20.868 00:20:24.539 joshuadeveyra: This was basically, yeah, yeah, it was basically from

195 00:20:24.860 00:20:28.830 joshuadeveyra: the what what changed on the prompt. And I just researched a bit on, you know

196 00:20:29.400 00:20:30.530 joshuadeveyra: this is all like.

197 00:20:30.530 00:20:31.689 Uttam Kumaran: We’ve produced it. Yeah.

198 00:20:31.690 00:20:32.320 joshuadeveyra: Yeah, yeah, we just.

199 00:20:32.320 00:20:37.260 Scott_Harmon: Gotcha. So you found somewhere. You either ask an AI or whatever give me.

200 00:20:37.570 00:20:40.170 Scott_Harmon: you know, guidelines on lease. Yeah.

201 00:20:40.260 00:20:41.670 Scott_Harmon: yes, yes. Okay.

202 00:20:41.920 00:20:42.620 Scott_Harmon: okay.

203 00:20:42.620 00:20:47.179 Uttam Kumaran: And these are the only 2 we needed, for, like some of the diffs on this one. So we just like

204 00:20:47.420 00:20:51.589 Uttam Kumaran: wanted to show this like sub agent process working. But yeah, again, we would

205 00:20:51.890 00:20:55.499 Uttam Kumaran: ideally, we for every single clause we have like these sorts of

206 00:20:55.920 00:20:58.189 Uttam Kumaran: this sort of a, this sort of like format.

207 00:20:58.655 00:21:00.530 Uttam Kumaran: I think probably this is

208 00:21:00.670 00:21:03.639 Uttam Kumaran: maybe like it’s it seems, a little bit of a lot.

209 00:21:03.951 00:21:07.609 Uttam Kumaran: But this is probably what I would try to get their feedback on.

210 00:21:07.610 00:21:14.240 Scott_Harmon: So just to tell me if if I’m seeing this rightly okay, if I’m understanding it. So

211 00:21:14.659 00:21:20.399 Scott_Harmon: and I and I’ll go. I’ll go look at the tool you’re using. It looks. It looks pretty interesting. What’s it called relevance.

212 00:21:20.400 00:21:23.299 Uttam Kumaran: Relevance. AI, yeah, it’s like an Australian startup.

213 00:21:23.780 00:21:27.070 Scott_Harmon: Yeah, okay? Good. So what you’re doing is you’re building

214 00:21:28.280 00:21:30.147 Scott_Harmon: relevancy. You’re building

215 00:21:31.060 00:21:34.139 Scott_Harmon: agents that can with some sort of a.

216 00:21:34.670 00:21:36.250 Scott_Harmon: with some sort of a

217 00:21:36.350 00:21:39.620 Scott_Harmon: methodology for calling agents and sub agents, which

218 00:21:39.800 00:21:42.539 Scott_Harmon: you know as we work through a workflow.

219 00:21:42.740 00:21:49.579 Scott_Harmon: But the way the fairway Doc concept is implemented is that they’re just another doc or a sub agent. And so a sub agent may say.

220 00:21:51.330 00:22:01.070 Scott_Harmon: tell me if this is accurate, compare this lease I have to my fairway dock for this example, security deposit guidelines. Yeah.

221 00:22:01.280 00:22:06.359 Scott_Harmon: And that sub agent would would sort of return an analysis, I guess.

222 00:22:07.630 00:22:13.489 Scott_Harmon: which would say, this is where this is inside or outside of the guidelines, is that the general idea.

223 00:22:13.660 00:22:17.219 Uttam Kumaran: That’s gen, that’s generally the idea. So this is actually the

224 00:22:17.870 00:22:19.219 Uttam Kumaran: like, where we kind of arrive.

225 00:22:19.220 00:22:19.840 Scott_Harmon: Right.

226 00:22:19.840 00:22:21.020 Uttam Kumaran: For this portion.

227 00:22:21.110 00:22:26.725 Uttam Kumaran: and then the the thing that Miguel had that I thought was like geniuses. We have an exceptions agent that sits above this.

228 00:22:27.160 00:22:29.170 Uttam Kumaran: This is where we capture, like

229 00:22:29.530 00:22:39.789 Uttam Kumaran: probably exceptions that, like strategic clients or like, I don’t know other documents right? But but like otherwise, we have these like guideline based sub agents. But yeah, they get called based on

230 00:22:39.990 00:22:41.570 Uttam Kumaran: what gets requested.

231 00:22:41.870 00:22:46.419 Scott_Harmon: On the sub on the subject area. They get called based on okay subject area. So

232 00:22:46.490 00:22:50.349 Scott_Harmon: do you. And so, how would the

233 00:22:50.590 00:22:56.390 Scott_Harmon: can you show me an example, and I know we’re super early of the result of one of those sub agents.

234 00:22:56.390 00:22:56.750 Uttam Kumaran: Let’s.

235 00:22:56.750 00:22:57.489 Scott_Harmon: Doing, a.

236 00:22:57.490 00:22:57.940 Uttam Kumaran: Yeah.

237 00:22:57.940 00:23:00.389 Scott_Harmon: Analysis, like, yeah, like.

238 00:23:00.790 00:23:02.560 Scott_Harmon: what? What might it look like?

239 00:23:02.560 00:23:03.170 Uttam Kumaran: Yeah, so.

240 00:23:03.170 00:23:03.900 Scott_Harmon: Take

241 00:23:04.060 00:23:05.020 Scott_Harmon: perfect.

242 00:23:05.020 00:23:06.840 Uttam Kumaran: Yeah, let’s go through one whole example.

243 00:23:06.890 00:23:08.689 Uttam Kumaran: So this is.

244 00:23:09.440 00:23:17.280 Uttam Kumaran: yeah. One of the potential tennis is off is offering to offer a year in advance security deposit. If we lower lower the terms of the lease from 60 to 48,

245 00:23:17.530 00:23:20.590 Uttam Kumaran: the tenant’s offer is not acceptable.

246 00:23:20.820 00:23:22.430 Uttam Kumaran: These are, here’s the reasoning.

247 00:23:23.000 00:23:25.139 Uttam Kumaran: The lease term is a critical component.

248 00:23:27.120 00:23:30.729 Uttam Kumaran: these guidelines emphasize setting the security deposit the highest reasonable amount.

249 00:23:31.219 00:23:37.300 Uttam Kumaran: So this is kind of like what you can expect again, all of this can be customized, but basically from the.

250 00:23:37.300 00:23:37.960 Scott_Harmon: Got it.

251 00:23:38.180 00:23:45.309 Uttam Kumaran: Security deposit agent. You get like the reasoning and like other things. And then, yeah, we’ve basically has tolerances set up. Yeah.

252 00:23:45.310 00:23:47.139 Scott_Harmon: Got. Okay. So what I

253 00:23:47.160 00:23:57.049 Scott_Harmon: this is probably this is probably, you know, I know we’re sort of at step. One of you know. Probably, you know, 10 like, we’re just trying to prototype basic functionality.

254 00:23:57.060 00:24:03.989 Scott_Harmon: This is this looks like a really interesting way to go through the documents kind of a section at a time.

255 00:24:06.000 00:24:10.580 Scott_Harmon: What I think we probably want to do in the Ui

256 00:24:11.570 00:24:14.489 Scott_Harmon: like this. This looks like the detailed.

257 00:24:14.670 00:24:17.950 Scott_Harmon: I think there needs to be some kind of summary scorecard new time, and we could sketch

258 00:24:18.510 00:24:20.180 Scott_Harmon: when we’re on a whiteboard together.

259 00:24:20.290 00:24:26.249 Scott_Harmon: you know, my at least in my mind, and I know we’re going to get together on Wednesday, and Greg Greg’s going to be there, and he can help us with this.

260 00:24:26.630 00:24:28.710 Scott_Harmon: What might the 1st

261 00:24:29.000 00:24:34.000 Scott_Harmon: ui look like, and maybe it’s just a high level scorecard, and then you drill down to this level

262 00:24:34.600 00:24:35.480 Scott_Harmon: right? So.

263 00:24:35.480 00:24:35.940 Uttam Kumaran: Yeah, I see.

264 00:24:35.940 00:24:45.249 Scott_Harmon: You know you. You may have a scorecard showing all the sections with some score like, Imagine a Slider or a, you know, red, yellow, green, or some other kind of visual.

265 00:24:45.790 00:24:54.939 Scott_Harmon: you know. Indicator of Hey, this is good, or this is troubling, you know whatever, so that at a glance you can see, boy, this one looks. There’s like 2 areas that are

266 00:24:55.160 00:25:00.160 Scott_Harmon: problematic. Oh, let me drill in, and then maybe you click on it. And now we’re seeing the detail.

267 00:25:00.997 00:25:04.489 Scott_Harmon: You know exactly why this one was unacceptable. So

268 00:25:04.760 00:25:09.070 Scott_Harmon: the I think that’s just almost a it’s not an afterthought. But it’s kind of a ux.

269 00:25:09.620 00:25:14.159 Scott_Harmon: Thought that I think would make it easier for a user to sort of

270 00:25:17.070 00:25:25.010 Scott_Harmon: employ the tool. But but I think so. That’s just it off the top of my head. Thought we could drill more into that on Wednesday, when we get together.

271 00:25:25.010 00:25:25.400 Uttam Kumaran: Yeah, yeah.

272 00:25:25.400 00:25:25.890 Scott_Harmon: Up to me.

273 00:25:26.204 00:25:27.460 Uttam Kumaran: That makes sense. Yeah.

274 00:25:30.040 00:25:32.100 Scott_Harmon: But this looks

275 00:25:33.250 00:25:34.330 Scott_Harmon: really

276 00:25:34.450 00:25:38.250 Scott_Harmon: powerful. If I’m getting this, then we would. We could just go change.

277 00:25:39.300 00:25:44.300 Scott_Harmon: You know, these reference agents, the Fairway dock or the the reference, Doc.

278 00:25:45.810 00:25:49.990 Scott_Harmon: and over time Hpi can kind of refine those right like.

279 00:25:49.990 00:25:50.780 Uttam Kumaran: Yes.

280 00:25:50.780 00:25:53.570 Scott_Harmon: Like it would be their own corpus of

281 00:25:55.130 00:25:59.159 Scott_Harmon: I call it best practice, or you know how they feel about leases.

282 00:25:59.210 00:26:04.759 Scott_Harmon: Exactly those those could be refined over, you know, through time they may be quite sophisticated.

283 00:26:05.608 00:26:06.800 Scott_Harmon: So that was exactly.

284 00:26:06.800 00:26:07.850 Uttam Kumaran: And we will have.

285 00:26:08.830 00:26:11.459 Uttam Kumaran: Yeah, we’re gonna have these prompts

286 00:26:11.530 00:26:21.589 Uttam Kumaran: and like these sorts of contacts saved. And then again, ideally, we come up with, like the minimum viable prompts for each of these. But we can think about like what the output

287 00:26:21.920 00:26:24.370 Uttam Kumaran: like. Now that we have the insides working.

288 00:26:24.480 00:26:45.509 Uttam Kumaran: it’s easy now that we cause for us. It was like, Okay, can we make sure? We, we have the all the least examples. And then what we do is you? Basically, you have one chat, instance per like negotiation. So you could keep asking things like that like, they just asked for this, what should I do? They just asked for this. What should I do? And it retains the memory of that like chat back and forth.

289 00:26:45.510 00:26:47.809 Scott_Harmon: Well, that’s that’s fabulous, I think.

290 00:26:47.810 00:26:48.130 Uttam Kumaran: Yeah.

291 00:26:48.130 00:26:53.300 Scott_Harmon: Exactly right. So you could have, you know, multiple, you know, one extended, you know, thread or chat.

292 00:26:53.300 00:26:56.720 Uttam Kumaran: Exactly. So. That’s the benefit of this tool is that

293 00:26:56.820 00:27:05.910 Uttam Kumaran: you can use the sub agents. But they’re the context are confined to like one, you know, chat where it’s just like, Hey, this is the. This is the, for example. This is like what

294 00:27:05.950 00:27:12.210 Uttam Kumaran: Miguel triggered. A tenant with a rent of 15 K. Is asking to lower deposit to 15 for 45 K.

295 00:27:12.440 00:27:16.310 Uttam Kumaran: And then here are the reasons why. And again, all this is actually

296 00:27:16.340 00:27:25.430 Uttam Kumaran: just based on those sub agent prompts you created. But those again, we just went for something we can now like make this really really adhere to

297 00:27:25.480 00:27:28.070 Uttam Kumaran: anything we want, but I do think that

298 00:27:28.380 00:27:29.774 Uttam Kumaran: I’m seeing how

299 00:27:30.360 00:27:44.159 Uttam Kumaran: how many components of at least there are. I think it’ll be great to have specific sub agents for each, and then I think the the lovely thing is you can drill into what those sub agents send back like. Can you show an example of like

300 00:27:44.290 00:27:47.669 Uttam Kumaran: what one of the return answers are from a sub agent.

301 00:27:48.330 00:27:51.339 joshuadeveyra: Yeah, sure. I think this one is a better one.

302 00:27:51.570 00:27:58.500 Uttam Kumaran: Yeah, like, when it calls. Yeah. So it calls the security deposit agent and the security deposit agent sent over these things.

303 00:27:59.560 00:28:02.649 Scott_Harmon: This is what it’s got sent to the agent, or what the agent said. Map.

304 00:28:02.650 00:28:04.209 Uttam Kumaran: What the agent sent back.

305 00:28:08.060 00:28:12.400 Scott_Harmon: Oh, wow! Look at that! Let me re scroll down. I want to read the summary paragraph

306 00:28:13.180 00:28:21.550 Scott_Harmon: tenants offer might seem beneficial in the short run it does not align. The strict and non-negotiable guidelines was created by the landlord, should maintain the original lease.

307 00:28:21.660 00:28:28.870 Scott_Harmon: determine screen deposit conditions. So that that’s really cool. And so what I’m suggesting is, you could almost use that

308 00:28:28.950 00:28:35.490 Scott_Harmon: the sentiment of that summary to inform a scorecard right? Just imagine a scorecard, where.

309 00:28:35.730 00:28:38.729 Scott_Harmon: you know, for each agent there was just some little visual.

310 00:28:39.730 00:28:43.640 Scott_Harmon: you know. Again we have to think about the ux, but it’s color coded something or other.

311 00:28:43.970 00:28:48.789 Scott_Harmon: and maybe with just a couple of words, and then you just drill in. And this is what you see.

312 00:28:49.840 00:28:50.410 Uttam Kumaran: Correct.

313 00:28:50.720 00:28:53.809 Uttam Kumaran: And so this this gets returned to like the

314 00:28:53.980 00:29:00.860 Uttam Kumaran: the top level Hpi lease negotiation agent, and then that one takes this and then crafts up.

315 00:29:01.280 00:29:02.630 Uttam Kumaran: That’s the response.

316 00:29:03.820 00:29:05.670 Scott_Harmon: Wow, that’s really

317 00:29:05.980 00:29:07.410 Scott_Harmon: interesting.

318 00:29:07.710 00:29:09.501 joshuadeveyra: It’s basically a back office.

319 00:29:11.840 00:29:15.130 Scott_Harmon: Well, that’s and that’s that’s right. That’s exactly

320 00:29:15.320 00:29:18.791 Scott_Harmon: what we’re hoping to to automate

321 00:29:19.660 00:29:21.670 Scott_Harmon: to give them back office

322 00:29:22.070 00:29:23.590 Scott_Harmon: intelligence.

323 00:29:23.810 00:29:32.119 Uttam Kumaran: The nice thing is you can you don’t have to like answer. Ask one and get one back. You can ask it, and then you can see, like the components of the

324 00:29:32.490 00:29:38.130 Uttam Kumaran: I again, the scorecard response, or whatever you can then say, Okay, I wanna drill into like, what was the reason

325 00:29:38.190 00:29:39.290 Uttam Kumaran: we didn’t.

326 00:29:39.320 00:29:47.720 Uttam Kumaran: It’s like, if you were to call on somebody like your security positive person like, why, why, why did you? What was it? Can you expand on your answer, and you can kind of click into see like what the

327 00:29:48.450 00:29:49.790 Uttam Kumaran: what the inputs were.

328 00:29:51.490 00:29:54.480 Scott_Harmon: I’m not sure I followed that time. Can you say that another way?

329 00:29:54.480 00:30:01.789 Uttam Kumaran: Yeah, meaning like, if you were to like, if you’re in a round table, and you’re like, okay, give me a summary of like what our decision is cool. Our decision is like red.

330 00:30:01.800 00:30:09.820 Uttam Kumaran: And you’re like, okay, I want to look at all the contributing inputs to those from a security deposit angle from this angle. From that angle you can see

331 00:30:09.930 00:30:16.149 Uttam Kumaran: all of like, actually very like detailed responses from each of the agents. Or you can also just do the high level decision

332 00:30:16.260 00:30:18.110 Uttam Kumaran: with some summary. Yeah.

333 00:30:18.350 00:30:20.840 Scott_Harmon: Right? Right? Right? Right and

334 00:30:22.530 00:30:28.319 Scott_Harmon: excellent. So this, you know, step one would be detailed analysis.

335 00:30:28.890 00:30:30.420 Scott_Harmon: And it

336 00:30:30.640 00:30:37.959 Scott_Harmon: that that I think we’re agreeing, we’ll have 2 levels. It’ll have some kind of a summary level that the agents roll up.

337 00:30:38.810 00:30:42.049 Scott_Harmon: you know, where you can add a sort of an at a glance.

338 00:30:42.250 00:30:48.620 Scott_Harmon: big picture, and then you’ll be able to drive, drill into and look at the detailed feedback from the agents.

339 00:30:50.530 00:30:55.239 Scott_Harmon: you know the the detail of of why it gave the opinion

340 00:30:55.330 00:30:56.580 Scott_Harmon: that it did.

341 00:30:57.440 00:31:07.309 joshuadeveyra: Maybe something like this is a ui but this one this time will. It’ll change to something, you know. It’s 7 out of 10, and then view details. Click it. It shows something like this.

342 00:31:08.870 00:31:18.899 Scott_Harmon: Yeah. And I think allowing people to. So I do think it’ll be multi party. So you could imagine it’ll be like the leasing. It’ll be like the leasing guy, and maybe 1, 1 or 2 other people will.

343 00:31:19.973 00:31:22.869 Scott_Harmon: Sometimes people need to debate this stuff

344 00:31:23.354 00:31:27.249 Scott_Harmon: not on the other side. Not not not not the counterparty, but

345 00:31:27.480 00:31:32.740 Scott_Harmon: the the landlord. The way this works is, they have a leasing agent

346 00:31:32.860 00:31:35.190 Scott_Harmon: so utam. That would be Richard Paddock

347 00:31:36.044 00:31:50.230 Scott_Harmon: who you met. And and then there’s a landlord who actually owns the building and has the final say. And so sometimes both people, both of those 2 people need to discuss a debate so.

348 00:31:50.380 00:31:55.860 Scott_Harmon: and but the and so. Some things are quite easy, and and the leasing agent

349 00:31:56.120 00:31:59.420 Scott_Harmon: is is able to make the decision by himself.

350 00:32:00.091 00:32:03.459 Scott_Harmon: But for for matters that would be more.

351 00:32:03.910 00:32:07.769 Scott_Harmon: I guess, on the line or or whatnot. Then he would kind of say

352 00:32:08.140 00:32:11.570 Scott_Harmon: he wanna consult or make sure he ran

353 00:32:11.660 00:32:13.140 Scott_Harmon: or involved

354 00:32:13.340 00:32:16.460 Scott_Harmon: the landlord in certain decisions.

355 00:32:16.490 00:32:24.629 Scott_Harmon: And so they and you can imagine that they may actually want to debate a little. Hey? Why do we do this again? Why are we suggesting 38 months? Well, as you can see.

356 00:32:24.980 00:32:28.459 Scott_Harmon: the agent recommended this, and I really think that. Okay, great.

357 00:32:28.610 00:32:32.189 Scott_Harmon: So I’m just saying, there, there’d be like a little conversational.

358 00:32:32.500 00:32:33.200 Uttam Kumaran: Yeah.

359 00:32:33.500 00:32:34.810 Scott_Harmon: You know, Banter.

360 00:32:34.960 00:32:37.729 Scott_Harmon: this wouldn’t be a single user.

361 00:32:38.490 00:32:39.379 Uttam Kumaran: Yeah, makes sense.

362 00:32:39.520 00:32:44.160 Scott_Harmon: Is is what I’m saying so, and and maybe we can spec that out.

363 00:32:44.820 00:32:48.830 Scott_Harmon: you know, as we get deeper into it, like how the different roles.

364 00:32:48.830 00:32:50.569 Uttam Kumaran: Yeah, use the.

365 00:32:51.880 00:32:56.750 Scott_Harmon: Does it? Does everybody see the same thing? Or just what’s the multi-threading thing look like? We can sort of sketch that out.

366 00:32:56.750 00:32:57.390 Uttam Kumaran: Yeah.

367 00:32:59.250 00:33:01.370 Scott_Harmon: Wow! This looks really exciting.

368 00:33:01.910 00:33:06.240 Uttam Kumaran: I know. That’s why I wanted to get you on the phone last week, as we’ve been sitting on this, we’ve been sitting on this goal.

369 00:33:06.240 00:33:07.060 Scott_Harmon: Wow!

370 00:33:07.610 00:33:08.890 Uttam Kumaran: I think it’s really.

371 00:33:08.890 00:33:09.420 Scott_Harmon: Cool.

372 00:33:09.420 00:33:11.295 Uttam Kumaran: I think it. I think it looks really

373 00:33:11.610 00:33:17.860 Uttam Kumaran: amazing, too. And this tool we’ve been using for maybe a month or 2. And Miguel has used this for almost 6, 6, over 7 months, but

374 00:33:17.890 00:33:26.349 Uttam Kumaran: it’s this great startup out of Australia. We met with the team there as well, and like, we’re starting to use this for a bunch of stuff for just automating

375 00:33:26.510 00:33:33.599 Uttam Kumaran: the company itself. But it’s so easy to use this tool. I mean I again, the nice thing is this is like

376 00:33:33.760 00:33:39.310 Uttam Kumaran: probably the least technical agent building tool. I’ve seen that like works really well, and you can call tools and things to like.

377 00:33:39.320 00:33:47.180 Uttam Kumaran: do run a Google search extract something from a document. So then, once we get, I would say, like 80 or 90% there from here we can consider if we wanna

378 00:33:47.540 00:33:55.089 Uttam Kumaran: use something else more technical. But again, like we can connect the vector database into here, you can do a whole host of stuff, and it comes with this ui.

379 00:33:55.220 00:33:57.569 Uttam Kumaran: so like I was gonna ask.

380 00:33:57.570 00:34:00.670 Scott_Harmon: About. I was gonna ask about the Ui, does the tool include.

381 00:34:00.670 00:34:02.090 Uttam Kumaran: Tool. Itself. Yeah.

382 00:34:02.440 00:34:07.449 Scott_Harmon: Right. So the Ui that hill partners would use would actually be hosted or served, or.

383 00:34:07.760 00:34:11.480 Scott_Harmon: you know, provided by relevance, we wouldn’t have to

384 00:34:11.880 00:34:13.689 Scott_Harmon: effectively program the Ui.

385 00:34:13.690 00:34:14.150 Uttam Kumaran: Yeah.

386 00:34:14.150 00:34:14.610 Scott_Harmon: Some other.

387 00:34:14.610 00:34:22.440 Uttam Kumaran: Which would be great. We, you know, we could put this behind an Api and like, if they wanted to do this somewhere else, it’ll it’ll basically still funnel back and forth through here. But

388 00:34:22.469 00:34:23.739 Uttam Kumaran: I think, like

389 00:34:23.929 00:34:30.860 Uttam Kumaran: in in the in this like area where we are like, how early like, this is a great thing we could just hand off to say, like, play around with it.

390 00:34:30.909 00:34:32.569 Uttam Kumaran: It’s something that, like.

391 00:34:32.790 00:34:36.919 Uttam Kumaran: you know, it’s pretty cost effective for us to to maintain. Yeah.

392 00:34:38.002 00:34:40.259 Scott_Harmon: I’m gonna I’ll dig into it so before.

393 00:34:41.320 00:34:43.796 Uttam Kumaran: Miguel, let’s make sure that. Yes, Scott,

394 00:34:44.520 00:34:48.929 Uttam Kumaran: is in here and cannot, and like, go ask questions through each of these threads.

395 00:34:49.800 00:34:53.649 Scott_Harmon: Yeah, okay, I got it. Build your own AI work for? Okay, I’ll dig into it.

396 00:34:54.790 00:34:58.430 Scott_Harmon: Oh, yeah, interesting. Okay, so

397 00:34:58.620 00:35:00.120 Scott_Harmon: this is perfect.

398 00:35:00.360 00:35:05.939 Scott_Harmon: I think it’s a great start gives us something to get to. I think if we could whiteboard a little more on Wednesday with Greg.

399 00:35:06.517 00:35:11.069 Scott_Harmon: and Greg’s doing a bunch. You know. He’s a friend of mine. I think I told you

400 00:35:11.400 00:35:18.299 Scott_Harmon: actually worked together for a long, long time. He’s run usually really big dev organizations, you know, 3 or 400 people.

401 00:35:18.610 00:35:23.059 Scott_Harmon: but he’s doing AI dev. Now for small projects, kind of like you are.

402 00:35:23.702 00:35:27.629 Scott_Harmon: And he and I just work together. We’re pals, and he’s looking for

403 00:35:28.020 00:35:32.160 Scott_Harmon: fun stuff to work on, so there, there may be a way he can help or get involved.

404 00:35:32.160 00:35:32.420 Uttam Kumaran: Cool.

405 00:35:33.040 00:35:33.660 Scott_Harmon: Yeah.

406 00:35:33.660 00:35:34.059 Uttam Kumaran: I mean, let’s.

407 00:35:34.060 00:35:35.639 Scott_Harmon: Start of the stuck up

408 00:35:36.740 00:35:37.310 Scott_Harmon: and.

409 00:35:37.310 00:35:39.591 Uttam Kumaran: I think I saw on Linkedin that he started like

410 00:35:39.980 00:35:45.920 Uttam Kumaran: his own, like like consultancy, or it’s like, I didn’t know whether they were starting a product or something.

411 00:35:46.270 00:35:53.230 Scott_Harmon: No, he’s just hanging out he’s just a consultant right now, but he was a founder of spice works, which got to be a pretty big company for a while, Alan.

412 00:35:53.400 00:35:55.619 Scott_Harmon: But yeah.

413 00:35:55.630 00:35:59.230 Scott_Harmon: he’s he’s just, you know, looking to do fun.

414 00:35:59.800 00:36:07.410 Scott_Harmon: you know, kind of dev project work. So he’s he’s really just a good friend. I think you’ll like I’m just a good guy to, you know. No one’s how to bounce ideas off of.

415 00:36:07.590 00:36:14.730 Scott_Harmon: you know there may be a way for him to help us on this project, but he’s doing done some pretty cool things already.

416 00:36:14.870 00:36:18.809 Scott_Harmon: so he may have some suggestions for us.

417 00:36:19.530 00:36:23.279 Scott_Harmon: You know, to to get this into a firm prototype. But

418 00:36:23.800 00:36:29.929 Scott_Harmon: yeah, super exciting. I think. I think, Hbi, if we can get a ui paste it onto it.

419 00:36:32.450 00:36:38.240 Scott_Harmon: you know. I think we’d be at the point, Tom, where we could get their feedback like, have a meeting, or even let them

420 00:36:38.800 00:36:42.310 Scott_Harmon: play around with a prototype and then get and then talk more.

421 00:36:42.570 00:36:46.140 Scott_Harmon: Yeah, you know concretely about a prod, you know, financing a project.

422 00:36:46.640 00:37:02.209 Uttam Kumaran: So I think one thing, maybe by Wednesday, because for on my side, like Miguel, has enough, has, like all the technical stuff that we need. I think the biggest thing that we can help him is what the scorecard output we wanna look like. So maybe we can wire. We can wireframe that on Wednesday.

423 00:37:02.550 00:37:03.130 Uttam Kumaran: like we.

424 00:37:03.130 00:37:05.110 Scott_Harmon: Yeah, I’d like to. And then any.

425 00:37:05.110 00:37:07.820 Uttam Kumaran: Anything else on the on the real estate side

426 00:37:07.970 00:37:15.919 Uttam Kumaran: cause. That’s also the area where, even for me, like I’ve I’ve been reading the leases, but I think we can infer anything about anything cool that would for them

427 00:37:15.980 00:37:18.510 Uttam Kumaran: like they care more about or less about like we can.

428 00:37:18.580 00:37:25.230 Uttam Kumaran: And then, basically, what we’ll do is we’ll create all that necessary agents, the sub agents, and then also frame the scorecard.

429 00:37:25.550 00:37:30.399 Scott_Harmon: So can I let me share? Let me share my screen real quick. Could you unshare? Let me share so.

430 00:37:30.400 00:37:31.030 Uttam Kumaran: Yeah. Go ahead.

431 00:37:31.500 00:37:33.539 Scott_Harmon: Yeah. So I’ll sketch out.

432 00:37:35.060 00:37:39.109 Scott_Harmon: maybe just a real rough scorecard. Just so we’ll have something to look at on

433 00:37:39.420 00:37:42.280 Scott_Harmon: on Wednesday. So you probably saw this.

434 00:37:42.810 00:37:48.390 Scott_Harmon: Miguel. I don’t know if you saw this this. These were the what I’m envisioning is 4 different sections, or

435 00:37:48.620 00:37:52.439 Scott_Harmon: you know, you can imagine these things. Maybe they’re, you know, the collapsing span.

436 00:37:52.950 00:37:59.799 Scott_Harmon: But the 4 major areas that they want to analyze leases are are the financial terms.

437 00:38:00.080 00:38:05.279 Scott_Harmon: the timing, the options, and then the legal, and then the legal items. Those are the 4 buckets

438 00:38:05.680 00:38:08.370 Scott_Harmon: that we’d like to roll up into a scorecard.

439 00:38:08.690 00:38:12.959 Scott_Harmon: And and he even provided like. So you can imagine a scorecard with 4 sections.

440 00:38:13.040 00:38:15.769 Scott_Harmon: maybe even like, you know, 4 icons.

441 00:38:16.200 00:38:22.329 Scott_Harmon: Yeah, you could imagine an icon having a coloration or a score like a slider from one to 10 who taught, you know, something that

442 00:38:22.420 00:38:23.740 Scott_Harmon: indicated.

443 00:38:23.970 00:38:25.000 Scott_Harmon: you know.

444 00:38:25.380 00:38:27.240 Scott_Harmon: some initial score.

445 00:38:27.340 00:38:38.099 Scott_Harmon: So maybe the financial terms imagine they were great. You know that that the sub agents all came back and go, man, these are great. And then maybe that’s that’s all. Green, or maybe 10 out of 10.

446 00:38:38.960 00:38:44.220 Scott_Harmon: And then you could expand it, and then you could see rate Ti free rent, and then parking.

447 00:38:45.030 00:38:48.929 Scott_Harmon: and then for each one you could drill into the sub agent feedback and see the details.

448 00:38:49.502 00:38:53.660 Scott_Harmon: Analysis right? And then, okay, maybe the second section

449 00:38:53.690 00:39:00.139 Scott_Harmon: is is is a, and that’s a calendar. So timing means the timing just means at

450 00:39:00.490 00:39:06.689 Scott_Harmon: when over time do all these? Does the tenant move in? And does their lease come in so that they start paying us. And

451 00:39:07.030 00:39:10.020 Scott_Harmon: the timeline is really important.

452 00:39:10.730 00:39:13.970 Scott_Harmon: And so you could almost imagine that being visualized.

453 00:39:14.310 00:39:22.150 Scott_Harmon: maybe even as a timeline, you know, with with indicators of whether where there are problems, or

454 00:39:22.570 00:39:24.649 Scott_Harmon: you know, good or bad things right?

455 00:39:24.870 00:39:26.090 Scott_Harmon: And then

456 00:39:26.140 00:39:31.049 Scott_Harmon: the 3rd thing, or you know, just options, the clients or the the tenants always ask for

457 00:39:33.160 00:39:36.760 Scott_Harmon: And then the 4th items are the hardcore legal

458 00:39:37.150 00:39:39.020 Scott_Harmon: things that lawyers

459 00:39:39.240 00:39:42.410 Scott_Harmon: get really worried about when the tenant.

460 00:39:42.500 00:39:45.490 Scott_Harmon: So they’re mostly things like

461 00:39:45.750 00:39:51.199 Scott_Harmon: audit rights and indemnification. Right? So that’s where slight changes in the language are really problematic.

462 00:39:52.580 00:39:58.869 Scott_Harmon: So anyway, I’ll sketch something out or maybe slap something together, and then we can. We can write for a little more.

463 00:39:58.890 00:40:04.419 Scott_Harmon: You probably got as much experience with Scorecard as I do from your bi days. You’re pretty familiar.

464 00:40:05.880 00:40:11.879 Scott_Harmon: Greg’s Greg’s done a ton like, you know. I can’t. He’s probably done 30 different scorecard kind of tools before. So

465 00:40:12.020 00:40:12.980 Scott_Harmon: so

466 00:40:14.790 00:40:17.250 Uttam Kumaran: But I do think we’re we’re probably like another.

467 00:40:17.290 00:40:20.299 Uttam Kumaran: you know, 2 weeks from something, I think, like enough

468 00:40:20.530 00:40:24.490 Uttam Kumaran: for them to be like, okay, this is like, what, how do we feel?

469 00:40:24.650 00:40:28.959 Uttam Kumaran: Right? So that’s kind of where I think, I, yeah, go ahead.

470 00:40:28.960 00:40:31.499 Scott_Harmon: So the other meeting I can get. So so

471 00:40:31.510 00:40:35.760 Scott_Harmon: Hbi is the primary focus. Obviously, that’s who I envision.

472 00:40:36.351 00:40:40.850 Scott_Harmon: you know, would would fund the project right? So we want to keep them in focus.

473 00:40:41.160 00:40:46.169 Scott_Harmon: But I may also set up a meeting. I think I’ve mentioned in the past

474 00:40:46.430 00:40:49.799 Scott_Harmon: with a guy named Bill Campbell.

475 00:40:50.400 00:40:54.870 Scott_Harmon: So Bill Campbell is the senior partner of a law firm downtown called

476 00:40:55.340 00:40:57.470 Scott_Harmon: Bryant Dubois Campbell.

477 00:40:58.300 00:41:04.250 Scott_Harmon: and so he’s a name partner of the firm, and they’re probably one of the 2 or 3 biggest

478 00:41:04.630 00:41:05.970 Scott_Harmon: real estate

479 00:41:06.000 00:41:07.869 Scott_Harmon: law firms in Texas.

480 00:41:09.180 00:41:10.050 Scott_Harmon: So

481 00:41:10.140 00:41:13.839 Scott_Harmon: you know, he’s a just a real, highly trained lawyer on real estate law.

482 00:41:13.970 00:41:17.610 Scott_Harmon: and it’s and he’s he’s just a really good friend of mine. And so

483 00:41:17.850 00:41:21.820 Scott_Harmon: he’d love to see this, too, and give us feedback. Yeah, he’s

484 00:41:22.050 00:41:26.699 Scott_Harmon: and he’s also unusually for a lawyer. I think he’s really open minded.

485 00:41:27.170 00:41:28.150 Scott_Harmon: and

486 00:41:28.860 00:41:37.629 Scott_Harmon: you know, I’d be surprised if he didn’t have a bunch of really intriguing feedback. So I’ll probably want to set up a meeting for him to look at it, too.

487 00:41:37.900 00:41:39.860 Scott_Harmon: Once we get something put together.

488 00:41:40.340 00:41:46.560 Uttam Kumaran: Yeah, let’s do that. I mean, I think once we get feedback from them on a demo, we can have something that like again, even if they don’t buy it. I think we can

489 00:41:46.750 00:41:53.689 Uttam Kumaran: something to shop around, but also something succinct like we can even make a little loom video or something that’s

490 00:41:53.700 00:41:54.789 Uttam Kumaran: that we can

491 00:41:55.080 00:42:00.460 Uttam Kumaran: haven’t share cause again. We’ve just been working on this on our own. For the most part, I think we could.

492 00:42:00.640 00:42:05.180 Uttam Kumaran: If our whole idea was like you sub in your what you care about, and then

493 00:42:05.390 00:42:09.385 Uttam Kumaran: that the pipes are all connected, you know. And

494 00:42:10.830 00:42:14.210 Uttam Kumaran: yeah, I mean, I think it looks really compelling now, especially compared to

495 00:42:14.320 00:42:15.740 Uttam Kumaran: yeah. Even a month ago.

496 00:42:16.290 00:42:17.810 Scott_Harmon: Yeah, that’s super exciting.

497 00:42:17.830 00:42:21.230 Scott_Harmon: Really. That’s I. That’s I’m gonna fiddle around with it and

498 00:42:21.420 00:42:22.840 Scott_Harmon: get a look at the tool and

499 00:42:23.431 00:42:27.300 Scott_Harmon: wednesday should be good that should that should give us a really good meeting on Wednesday.

500 00:42:27.300 00:42:35.769 Uttam Kumaran: Yeah. And if you have any questions about relevance, please. Message Miguel, because he’s the expert but we’ve we’ve we’ve been working with their team, and they’re great, and

501 00:42:35.870 00:42:39.270 Uttam Kumaran: I don’t know out of all the agent tools builders that I’ve tried like

502 00:42:39.650 00:42:41.629 Uttam Kumaran: they’ve been awesome. So.

503 00:42:42.220 00:42:49.289 Scott_Harmon: Okay, and I may. I’ll let you know there may be one other person. Join us on Wednesday. He’s another friend of mine who’s a lawyer.

504 00:42:49.320 00:42:54.770 Scott_Harmon: Okay, that is, actually, he’s actually the Cfo of a software company. Well, he’s actually the chief legal counsel at

505 00:42:55.250 00:42:57.020 Scott_Harmon: Wp engine.

506 00:42:57.210 00:42:57.880 Uttam Kumaran: Oh, yeah.

507 00:42:57.880 00:42:59.600 Scott_Harmon: Impressed. Yeah, here in Austin.

508 00:43:00.000 00:43:00.860 Uttam Kumaran: Nice.

509 00:43:01.250 00:43:03.819 Scott_Harmon: And his old buddy of mine used to work for me, and

510 00:43:04.936 00:43:06.450 Scott_Harmon: he’s just.

511 00:43:06.680 00:43:13.188 Scott_Harmon: He only likes to look at startups and give feedback, so I may see if he can drop over and spend a few minutes with us. Cool

512 00:43:13.510 00:43:16.369 Scott_Harmon: But I’ll let you know via email, if that’s the case.

513 00:43:16.370 00:43:17.120 Uttam Kumaran: Okay.

514 00:43:18.040 00:43:18.830 Scott_Harmon: Okay?

515 00:43:19.000 00:43:20.631 Scott_Harmon: Well, that looks great.

516 00:43:22.010 00:43:24.350 Scott_Harmon: terribly exciting. Thanks for showing it to me.

517 00:43:24.350 00:43:27.970 Uttam Kumaran: Yeah, thanks. Guys, appreciate it. And yeah, let’s check on. Let’s chat on slack.

518 00:43:29.110 00:43:29.470 joshuadeveyra: Forever.

519 00:43:29.470 00:43:30.609 Scott_Harmon: Chat soon. Thank you.

520 00:43:30.610 00:43:32.090 Uttam Kumaran: Thanks, guys, bye.