Meeting Title: Robert Tseng and Connor Nevelle Date: 2025-06-17 Meeting participants: Connor Nevelle, Ben Redfield, Robert Tseng, Uttam Kumaran
WEBVTT
1 00:00:37.090 ⇒ 00:00:38.330 Uttam Kumaran: Hello!
2 00:00:38.690 ⇒ 00:00:39.480 Robert Tseng: Hey!
3 00:00:39.730 ⇒ 00:00:40.780 Uttam Kumaran: Hey! How are you?
4 00:00:41.380 ⇒ 00:00:42.470 Robert Tseng: Good! How are you?
5 00:00:43.250 ⇒ 00:00:45.179 Uttam Kumaran: Good dude. I’m kind of tired.
6 00:00:45.560 ⇒ 00:00:46.160 Robert Tseng: Yeah.
7 00:00:46.920 ⇒ 00:00:54.624 Uttam Kumaran: And a waste just deleted our entire notion database. Well, I’m getting it back. I’m getting it back.
8 00:00:56.050 ⇒ 00:00:57.860 Robert Tseng: What is going on with our team?
9 00:00:58.760 ⇒ 00:00:59.970 Robert Tseng: Oh, my goodness.
10 00:01:05.290 ⇒ 00:01:09.963 Uttam Kumaran: Yeah, I had to sleep in a little bit. I woke up like I’m not feeling it right now.
11 00:01:10.400 ⇒ 00:01:16.410 Robert Tseng: Yeah, better better today than yesterday. That’s still kind of crazy.
12 00:01:20.540 ⇒ 00:01:30.620 Robert Tseng: Okay, so I mean, just to give you the primary like. So Connor is the guy talking to you brought? He bring Ben on the call. I believe Pet Ben is a partner at blue blueprint.
13 00:01:30.880 ⇒ 00:01:31.225 Uttam Kumaran: Okay.
14 00:01:31.570 ⇒ 00:01:36.129 Robert Tseng: So I’m expecting him to do more of the talking. And then.
15 00:01:36.390 ⇒ 00:01:41.939 Robert Tseng: yeah, he’ll probably just ask, like, how this like, how we want to do the AI workshop and stuff. So I think
16 00:01:42.070 ⇒ 00:01:45.029 Robert Tseng: we’ll probably leave it to you to kind of handle that a bit more.
17 00:01:45.160 ⇒ 00:01:50.630 Robert Tseng: They have port coast coming on Site end of summer, either August or September. They’re based in San Diego.
18 00:01:52.410 ⇒ 00:01:59.130 Robert Tseng: I don’t know if you’ve looked at their portfolio. But yeah, I mean. It seems like as mostly
19 00:02:00.480 ⇒ 00:02:03.820 Robert Tseng: I forgot now in the Sas or something.
20 00:02:07.850 ⇒ 00:02:13.610 Robert Tseng: Yeah, they’re mostly they’re mostly sas. So keep that in mind.
21 00:02:14.950 ⇒ 00:02:15.550 Uttam Kumaran: Okay.
22 00:02:15.800 ⇒ 00:02:16.440 Robert Tseng: Yeah.
23 00:02:26.060 ⇒ 00:02:30.899 Ben Redfield: Hey, guys? Sorry I was I was on another call, and it ran a minute long.
24 00:02:31.180 ⇒ 00:02:32.690 Robert Tseng: In order, of course. Hey, Ben?
25 00:02:32.860 ⇒ 00:02:33.890 Uttam Kumaran: Nice to meet you.
26 00:02:34.770 ⇒ 00:02:35.949 Ben Redfield: Nice to meet you, too.
27 00:02:38.380 ⇒ 00:02:39.870 Robert Tseng: Are you based in San Diego?
28 00:02:40.807 ⇒ 00:02:54.899 Ben Redfield: No, I live in Los Angeles. But it’s close enough that I come down to the blueprint office every couple of weeks. I’m there like today. I’m about ready to head home. But I’ll usually come in for a day or 2 couple of times a month.
29 00:02:56.110 ⇒ 00:03:00.299 Robert Tseng: Very cool. Nice, hey, Connor? Good to see you.
30 00:03:00.300 ⇒ 00:03:00.880 Uttam Kumaran: Go ahead!
31 00:03:06.160 ⇒ 00:03:07.750 Robert Tseng: Everybody’s audio loading a bit.
32 00:03:08.460 ⇒ 00:03:10.070 Connor Nevelle: Sorry. Everybody. New- new computer.
33 00:03:10.540 ⇒ 00:03:11.260 Connor Nevelle: It’s relative.
34 00:03:11.260 ⇒ 00:03:12.350 Robert Tseng: Yeah, I can hear you now.
35 00:03:12.820 ⇒ 00:03:13.590 Connor Nevelle: Perfect.
36 00:03:15.390 ⇒ 00:03:43.620 Robert Tseng: Yeah, thanks for setting up this call, Connor. I just wanted to. I brought Utah, my business partner onto the call. Yeah, I’m assuming we’re probably gonna talk a bit more about kind of how we would structure the day workshops, and then maybe logistically kind of when you bring your port close outside. And I think, yeah, would be able to speak to all that. So not sure if you had the chance to look over everything we’d sent over. But yeah, I mean, I’ll turn over to you and let you kind of drive. However, you want. After this conversation.
37 00:03:43.620 ⇒ 00:03:51.369 Connor Nevelle: Yeah, I did. Thanks. Thanks. Joined a minute or 2, and Ben leads kind of our value. Creation
38 00:03:51.370 ⇒ 00:04:14.419 Connor Nevelle: at at blueprint does many different things. One of it is helping plan this on site, so I want to bring him as well. Let you all meet each other. As a quick recap. I’m not sure how much Robert shared. With you on just the type of event we’re looking to have. And the timeline. It’s really just a 1 day overview maybe one and a half days with our.
39 00:04:14.480 ⇒ 00:04:21.449 Connor Nevelle: We probably should have, you know, mid twenties, portfolio companies at the time, you know, 22, maybe 23.
40 00:04:21.910 ⇒ 00:04:27.779 Connor Nevelle: hopefully, all of them attend. We want to have an AI on site where we walk through. Just you know.
41 00:04:28.290 ⇒ 00:04:51.340 Connor Nevelle: high level more of a consultative approach. The thing, I think, is, gonna be difficult with our Portfolio company is, not only do we have a pretty wide range of size. We have, you know, companies that are close to a hundred 1 million in revenue. We have companies that are 1 million in revenue. So it’s not just a size thing which probably translates into like scope and budget of the type of consulting arrangement they’d agree with agree to after
42 00:04:51.340 ⇒ 00:05:13.649 Connor Nevelle: there’s a pretty wide range of just sophistication. When it comes to AI. We have some of our smaller companies that are AI forward. From the beginning we have thought leaders within our portfolio that likely could probably lead one of these seminars if they were asked to, and on the other side of it we have even some of our larger companies are doing little to none when it comes to
43 00:05:13.750 ⇒ 00:05:19.529 Connor Nevelle: like truly being a top to bottom. AI, native product and platform. So
44 00:05:19.720 ⇒ 00:05:36.460 Connor Nevelle: a kind of a couple of different moving pieces that we need to work around with in terms of this on site. When Ben and I have been talking like we think it makes most sense to break these companies up into different cohorts and groups based on different things. It could be business model could be size, the AI maturity. It could be
45 00:05:36.990 ⇒ 00:05:39.739 Connor Nevelle: a lot of different reasons and mixing and matching of
46 00:05:39.810 ⇒ 00:05:56.520 Connor Nevelle: who gets in with it? Who gets put within each cohort. But yeah, that’s you know our high level, what we’re looking to accomplish. We’ve started to iron out more of what we would like to do between now and September, and we think it would probably make sense for
47 00:05:56.520 ⇒ 00:06:10.750 Connor Nevelle: you all, or whoever is working with us, to meet with each of our individual portfolio companies prior and do just a kind of like a mini assessment with each of them, and identify common themes and maybe opportunities that
48 00:06:10.820 ⇒ 00:06:30.640 Connor Nevelle: we could, you know, bring those common themes to the onsite. If you know, every Portfolio company of ours says these are the 2 or 3 biggest issues, and there’s an overlap where it’s 1 thing that everyone seems to care about probably makes sense to focus on those sort of things at the Onsite, and at the same time meeting them. Prior would give you all a good sense of
49 00:06:30.910 ⇒ 00:06:35.829 Connor Nevelle: who makes sense to be in which cohort, or which maturity level, and
50 00:06:36.080 ⇒ 00:06:39.339 Connor Nevelle: those those sort of things. So that’s, you know
51 00:06:39.400 ⇒ 00:07:00.809 Connor Nevelle: high level. I think if we left that on site with every Portfolio company had one or 2, or even 3 action items of what they want to accomplish over the next 3, 6, 9 months, and even you know, type of engagement of, we can use Brainforge to get these things done over the next 9 months we feel really good. We don’t need to.
52 00:07:00.840 ⇒ 00:07:09.930 Connor Nevelle: you know, build anything prior or create anything. Prior, it’s just we want to get them thinking about things the right way, and, you know, moving forward on their own respective paths.
53 00:07:11.450 ⇒ 00:07:24.240 Uttam Kumaran: Yeah, I think to talk about that. I mean, we would actually, for all of our workshops, we would actually require a lot of that prep work like, I’ve been in a lot of like corporate workshops and a lot of them suck.
54 00:07:24.360 ⇒ 00:07:28.870 Uttam Kumaran: And I actually was pretty averse to like doing a sort of workshop.
55 00:07:28.960 ⇒ 00:07:29.490 Robert Tseng: Shit.
56 00:07:29.490 ⇒ 00:07:32.679 Uttam Kumaran: Webinar type product, just because
57 00:07:32.960 ⇒ 00:07:57.149 Uttam Kumaran: it’s typically like someone just talks at a screen. And everybody like checks a box. So for this, we actually do have like a quite a fair bit of discovery, and like a questionnaire and stuff we do beforehand, which we think would fit really nicely into that and then we sort of prep the like activities and the conversation that way, based on like, who’s in the room
58 00:07:57.260 ⇒ 00:08:27.180 Uttam Kumaran: what the topics are and like exactly like what the outcomes we’re trying to get to. Of course, that will happen over hours. But it is like a facilitation towards a couple of questions that are pretty scripted beforehand but are unique towards the cohort that you’re talking to like. Just like you mentioned the folks at 50 to 100 million very different set of problems, and the folks, and very different set of constraints than the folks in like the one or 2 million range
59 00:08:27.546 ⇒ 00:08:48.020 Uttam Kumaran: and so those are the exact kind of things we want to delineate on who’s gonna have, who in their cohort is gonna have the best conversations, and even within their portfolio companies. I don’t know if it’s just gonna be the the founders that are there. But who’s there that are that are gonna talk through this. And I again. I don’t. I don’t know necessarily whether like
60 00:08:48.660 ⇒ 00:09:06.190 Uttam Kumaran: type of company or like. It’s sort of the business model is the best way, I do think that what we found is the toughest part in a lot of the AI work is one identifying problems really identifying how to prioritize the problems that are that are feasible to solve in the short term and worth solving.
61 00:09:06.963 ⇒ 00:09:19.560 Uttam Kumaran: So that’s so, we kind of bring the little bit of the feasibility, but worth solving and worth getting. Motivation around is something that you have to discuss. Second is the adoption. So we really find that
62 00:09:19.870 ⇒ 00:09:39.170 Uttam Kumaran: it has to be adopted like there, and not only by the founder, who could be just like super driven, and even in our company we have challenges with adoption of AI tools, not, it may not necessarily be because people are pushing back, but it’s just because people aren’t familiar. So those are the things that would like. I feel like we have a lot of conversations with.
63 00:09:39.440 ⇒ 00:09:46.640 Connor Nevelle: Yeah, it’s interesting. You say internally, your company. You’re you’re seeing the same thing because blueprint has that same problem. We.
64 00:09:46.640 ⇒ 00:09:52.130 Uttam Kumaran: Yeah, you’d be surprised on a word company. Yeah.
65 00:09:52.130 ⇒ 00:09:55.860 Connor Nevelle: We’re we’re. We’re kind of preaching that you need to use AI tooling. But, interesting
66 00:09:55.860 ⇒ 00:10:23.639 Connor Nevelle: admittedly, like our internal adoption of AI tool has quite a long ways to to go, and we were kind of kicking the can around. It might be a good 1st step is as you do. These initial assessments. Blueprint could almost be the 1st initial assessment that you’ll do is like a test run. So we can get a sense of what it’s gonna look like when you’re working with our portfolio companies. You could almost treat us like a company. I’ve said, like, you know, our team, we have
67 00:10:23.660 ⇒ 00:10:36.010 Connor Nevelle: investment associates that more or less mirror the work responsibilities of what Bdr. Or an account executive would do at one of our portfolio companies. We have, you know, our Vps sort of sit. And
68 00:10:36.100 ⇒ 00:10:50.960 Connor Nevelle: somewhere between, like a sales manager and finance manager role, they do do a bit of both. We have our portfolio operations team, that’s everything from recruiting to like kind of an Fp and a type person to. So our our
69 00:10:51.060 ⇒ 00:11:06.119 Connor Nevelle: business, even though we’re a investment firm, pretty closely mirrors. What a lot of our portfolio companies are doing anyways, and like our go to market motion, it’s similar. We have a pipeline and funnel. We have deal stages. We have all these different things that we could definitely
70 00:11:06.240 ⇒ 00:11:11.789 Connor Nevelle: be better at practicing what we preach in terms of, you know, bringing AI into our own processes.
71 00:11:12.100 ⇒ 00:11:34.879 Uttam Kumaran: Yeah, I feel like, I don’t know, Robert. We’ve been that like we’ve been sort of pushing towards a lot of different types of adoption. Internally, we have tools that we built. We, of course, have tools like Chat Gpt that we buy, and we just sort of give out to everybody. But it has been slow, and it’s been slow, based on different roles, like we have engineers. We have project managers. We have folks like me and Rob are on the sales side. We have internal operations, folks
72 00:11:35.030 ⇒ 00:12:00.839 Uttam Kumaran: and the tools that they use and the types of things that they try to automate. And the way that people learn we just found are are very different. But, like I, I don’t know, I would say in the last, probably 3 or 4 months. We’re seeing a lot of success. But it took a but it took a lot of work, and I was also like, if I was really, if our company can’t adopt it, and like, what are we selling? And so for us? I think we found different ways of
73 00:12:00.840 ⇒ 00:12:23.920 Uttam Kumaran: making it possible whether it’s through like quick, like vibe coded applications that like need to exist, to just do simple things that we can’t buy a sas off the shelf, for whether it is something that’s like for us all of our stuff live tries to live in slack, because I realize that people aren’t gonna go outside of there, and I want to see other folks using it. So some of those like practices and like ways of
74 00:12:24.000 ⇒ 00:12:39.639 Uttam Kumaran: pushing adoption. And then the other thing is like we try to measure. So for our clients that we deploy AI agents or agentic workflows, we run evals. We measure like how often it’s getting used. We sort of give to the stakeholders like, who’s using it? Who’s not using it, like.
75 00:12:39.870 ⇒ 00:13:00.619 Uttam Kumaran: you know, all that data around the adoption, because that’s what we’re going for beyond, just like, Hey, it worked like I said something and it got it. I’m like, cool. We’re we’re not done until people actually use it. And then you you sort of can not only like, see that. Okay. Given this amount of usage, I should be getting back this Roi, but you can feel that that like people are like, Wow, I actually just like
76 00:13:00.870 ⇒ 00:13:07.700 Uttam Kumaran: not doing this really boring thing. Or I’m not doing this thing that you should take me 10 HA week. It’s now taking me one or 2 HA week, you know.
77 00:13:07.700 ⇒ 00:13:13.610 Connor Nevelle: Yeah. So Robert mentioned that you did something similar, for I believe it was live oak ventures. Is that is that right, Robert?
78 00:13:13.880 ⇒ 00:13:24.390 Connor Nevelle: Yeah, just curious. What? Like what that looked like, how similar that was to what we’re proposing like even the type of work. If you work with their portfolio companies. Afterward.
79 00:13:24.820 ⇒ 00:13:45.690 Connor Nevelle: taking a step back, we invest only in b 2 b early stage vertical Saas companies. So all of our portfolio very different industries. But in general have similar business models and internal processes to to work in. So yeah, just curious of what your experience live up is like, what? What that type of work with their portfolio companies look like after.
80 00:13:46.500 ⇒ 00:13:57.819 Robert Tseng: Yeah. So I mean, that was really just workshop for their for their team. And yeah, kind of like I mentioned like getting the getting some of the getting a partner. Some of these like operator type folks that are working directly with the port coast
81 00:13:58.470 ⇒ 00:14:22.300 Robert Tseng: into a room. And then we’re kind of just like basically running through this, this workshop checklist that we kind of like sent you using a similar framework and us doing like 80 AI readiness. And so, yeah, we’re still kind of reaping the the fruit of like afterwards. Or we’re getting to continue to follow up with with those companies didn’t really have an opportunity to bring everyone on site. In the way that kind of you guys are planning for September. So I think this
82 00:14:22.430 ⇒ 00:14:50.590 Robert Tseng: event would definitely be be different, and we get to do a lot more like direct. Seems like we would have the opportunity to meet with companies before we do it, which I think is, is good. I think that’s a big learning for us that, rather than like waiting for the firm to kind of disseminate it. If we can establish kind of that 1st point of contact as well, and kind of be the ones to kick off the conversations. That kind of helps bring more excitement around it. So I think that was pretty aligned with what you had pitched last time. Connor.
83 00:14:50.590 ⇒ 00:14:59.250 Connor Nevelle: Was there any like things that at the live up meeting you could tell, the operating leaders were gravitating more, more towards.
84 00:15:00.860 ⇒ 00:15:12.419 Robert Tseng: Yeah, well, so I think we were focused a lot more on the business development side. Right? So for them, you know, there’s a lot of like this lead research and kind of like, kind of what you’re describing as investment associate workflows like.
85 00:15:12.420 ⇒ 00:15:35.220 Robert Tseng: yeah, go to market is kind of the place that where people go go to 1st and within our own, within our own team. That’s that’s the those are the workflows that have gotten the most adoption most quickly. Right? And so, yeah, we we have demos like, we could even show you some stuff today of like what that stuff that we’ve built internally and be able to walk people through those and just field questions. I think that’s
86 00:15:35.500 ⇒ 00:15:41.840 Robert Tseng: you know the show. The show that over. Tell is kind of the is the way to kind of get get the engagement going.
87 00:15:41.840 ⇒ 00:15:53.600 Uttam Kumaran: Yeah, and on on that piece, like, you know, part of the reason. So our company started primarily as a data, analytics and data engineering related consultancy. So both of our backgrounds are in data. But we worked at
88 00:15:53.620 ⇒ 00:16:14.070 Uttam Kumaran: bunch of different B, 2 B like software tech focus, Vc backed startups. So a lot of our work on the data side is spans the whole company. But of course, a lot of it goes to sales operations, product analytics, and so naturally, even within our company, the stuff that’s hardest to automate is the engine is like the engineering work.
89 00:16:14.070 ⇒ 00:16:35.789 Uttam Kumaran: but really, like top of our mind is like, how do we spend more time with clients? So how do our project managers not like sit and do summaries, and have to look through all the tickets. We use AI to help them summarize that, and then quickly get an email out to a client more often follow up more often schedule, get a get a chance for just like one more 30 min meeting to get some face time.
90 00:16:35.790 ⇒ 00:16:55.839 Uttam Kumaran: The second thing is on sales, so the time in between calls all the momentum that typically can get lost. That’s like what we decided to really attack but again, for all of the sort of investment firms or venture firms that we’re talking to. It’s it’s always on the either the diligence side or like sourcing side.
91 00:16:55.840 ⇒ 00:17:13.340 Uttam Kumaran: and like trying to trying to speed up sort of either research or qualifying leads, or just giving like a a summary or perspective. So that you can prime a meeting. You know. And so we’re seeing a lot of that. I think in in your companies you’ll probably see similarly, where I think sales is the most ripe.
92 00:17:13.770 ⇒ 00:17:28.469 Uttam Kumaran: I think it goes to probably like project management and like internal operations. Next, engineering is the probably the hardest. One easy thing is for everyone, just to make sure their folks are using cursor, and like, probably just requiring that. But
93 00:17:28.880 ⇒ 00:17:39.399 Uttam Kumaran: building more in the engineering side is a bit harder in sales project management. There’s a lot of an internal like ops like contract handling. There’s a lot of opportunity.
94 00:17:39.400 ⇒ 00:17:53.669 Connor Nevelle: Got it. I agree with that. I feel like our portfolio companies. There’s just there’s less risk, and there’s less buy in and trust you need from them to automate lead prospecting internally versus releasing a product facing plat like a.
95 00:17:53.670 ⇒ 00:17:54.080 Uttam Kumaran: Yes.
96 00:17:54.080 ⇒ 00:17:57.368 Connor Nevelle: Facing product and so completely agree with that.
97 00:17:58.130 ⇒ 00:18:07.889 Connor Nevelle: I you mentioned Demos. Well, like, let’s just save 5 min at the end of this call definitely want to see some of the things you’ve built that’d be very helpful to to walk through a couple of those.
98 00:18:08.360 ⇒ 00:18:20.440 Connor Nevelle: looking at your website. What drew me to it initially, was, I like your focus on using existing 3rd party tools and just tying them together from an automation perspective like
99 00:18:20.520 ⇒ 00:18:49.700 Connor Nevelle: of our 20 or so portfolio companies. There’s maybe one or 2 that would benefit from a 6 month Mckinsey level consulting engagement where they get 50 slides as an output. Like most of them, you’re going to get buy in when you pop in and you create them an workflow on the 1st day that links a Bdr process that they’re doing to an ae like if if they can see a tangible output. And
100 00:18:49.900 ⇒ 00:19:00.139 Connor Nevelle: I actually like the aspect of having these things be on kind of low to mid code platforms like an where non technical Ceos can like look at it and play around with it.
101 00:19:00.140 ⇒ 00:19:04.030 Uttam Kumaran: Spotify prompts like, add things. Yeah, I I mean.
102 00:19:04.130 ⇒ 00:19:09.570 Uttam Kumaran: that was our sort of my internal battle, too, because I’m like a typical software engineer. I was like.
103 00:19:10.400 ⇒ 00:19:32.390 Uttam Kumaran: I really don’t want to build these on low code, and we tried a lot of different ones. And a couple are really good because one they do have a code back end, and they’re open source. And second, you’d be surprised that the AI people we hire have no formal like software background, like the folks we hire started their career in the last like 5 years, purely just doing AI stuff. So for them
104 00:19:32.530 ⇒ 00:19:37.730 Uttam Kumaran: to even like work in an environment like this. It needed to be something like that, but is technical enough
105 00:19:38.020 ⇒ 00:19:58.249 Uttam Kumaran: for it to actually, basically you can, you can do what they what comes out of the box, and then you can also hit python. Apis hit different rest. Apis. Do web hooks. You can do everything you need, and then, of course, like, if it’s something that they’re gonna take to production for their clients, go the distance, but otherwise don’t go the distance. There’s like no point.
106 00:19:58.680 ⇒ 00:20:19.799 Connor Nevelle: Yeah, no, I I like that approach. I feel like, our our portfolio companies, especially in smaller ones. If they can engage in like a 2, 3 week engagement, whatever that means, and get a single M. 8 and output like that that leads to a second one, the 3rd one before, you know, hopefully, a lot of their internal processes are optimized.
107 00:20:20.620 ⇒ 00:20:22.720 Connor Nevelle: do you have? Ex, so
108 00:20:23.390 ⇒ 00:20:42.720 Connor Nevelle: 90% of our portfolio is using quickbooks. 80% of our portfolio, using either probably 90% is using either Hubspot or salesforce. And like, there’s 1 or 2 that uses a different accounting platform or a different Crm like, do you have experts within your team that
109 00:20:43.210 ⇒ 00:20:54.510 Connor Nevelle: know the ins and outs of working with these platforms? Because, admittedly like, there are portfolio as companies where their crms are in different states, cleanliness and like
110 00:20:55.000 ⇒ 00:21:00.599 Connor Nevelle: standard data, connector is definitely not gonna not gonna work. In the majority of our cases.
111 00:21:00.820 ⇒ 00:21:01.139 Uttam Kumaran: Yeah.
112 00:21:02.230 ⇒ 00:21:27.370 Uttam Kumaran: so we do a lot of work in name, your favorite Crm, like. So we like, we do a lot of stuff in Hubspot done a lot of work in in salesforce. We, you know, people use klaviyo like anything around sort of like bringing disparate data sets into one area to either make available to an A to an AI Llm as contacts, or for reporting
113 00:21:27.460 ⇒ 00:21:45.370 Uttam Kumaran: like that’s what we do. In fact, like we found that typically the AI stuff we’re doing ends up being mostly like a data challenge of like getting the right documents unstructured or structured data into a place where you can run rag over it. That’s like a lot of data stuff. The last piece is sort of forming the right prompt
114 00:21:45.370 ⇒ 00:22:01.060 Uttam Kumaran: and then piecing in the AI and then on the back end of that, you need to then send that somewhere. Right? So if your AI, for example, is, gonna take in 5 data points about a client, and then come up with a low medium. High risk of success like for opportunity success
115 00:22:01.060 ⇒ 00:22:24.840 Uttam Kumaran: that needs to go back into Hubspot somewhere. Right. That is, that’s the sort of activation piece where, like, yeah, you could sort of give it all you could do a 1 time thing where you give it all the information says this lead is like at a low odds of success, for this Xyz reasons. But if it lives in someone’s chat, Gbt, like. Nothing happens to that, and that person’s the only one empowered just for that one moment where they had all the perfect, prompt right.
116 00:22:24.840 ⇒ 00:22:25.430 Connor Nevelle: Yeah, yeah.
117 00:22:25.430 ⇒ 00:22:37.919 Uttam Kumaran: You kind of need that to happen all the time, and you need that to like sort of automatically filter back into wherever your you know your sales folks are taking actions. If that’s their Crm, if that’s back in slack.
118 00:22:38.070 ⇒ 00:22:52.349 Uttam Kumaran: we’ve even done stuff for clients where we’ve drafted emails into their inbox like where to going one step further from like, just draft me this, the AI actually like basically writes it and puts it into their email inbox.
119 00:22:53.410 ⇒ 00:22:55.989 Uttam Kumaran: Leave it there for them to just tweak and hit send. So
120 00:22:56.570 ⇒ 00:23:06.189 Uttam Kumaran: it’s there is sort of a world of things. But this is where again, you’re not gonna right now, we don’t reck there’s not a lot of tools that can sort of give you that flexibility, or there are a lot of like
121 00:23:06.560 ⇒ 00:23:28.296 Uttam Kumaran: tools that are like, Oh, build whatever thing you want. But your your folks aren’t gonna build that. It’s really kind of hard. Instead, you want them to really buy in and adopt and deal with like the hiccups that are in the proof of concept. Mvp. Phase, things aren’t gonna work perfectly. But you want them to partner with you, to work through that until finally it’s starting to take care of things.
122 00:23:29.740 ⇒ 00:23:37.550 Connor Nevelle: How about on the like, Zendesk and intercom? And you know Hubspot’s customer success modules like, Do you?
123 00:23:37.700 ⇒ 00:23:41.249 Connor Nevelle: Do? You have ex- experience working in those.
124 00:23:42.490 ⇒ 00:24:10.750 Uttam Kumaran: Yeah, we’ve done work with with Zendesk and with intercom before Hubspot customer. Success is very similar. I would say, we try to do most of our work on. Like the revenue generating like sales side of the businesses. Teams tend to not get a lot of attention or love, but they are very important. It’s just commonly not like the number one problem. But we’ve done a lot analytics around like, yeah, customer support analytics, agent performance Nps, scores like
125 00:24:11.340 ⇒ 00:24:16.480 Uttam Kumaran: everything around sort of customer service related reporting. Yeah, we’ve done a lot of that.
126 00:24:16.480 ⇒ 00:24:16.993 Connor Nevelle: Got it
127 00:24:17.250 ⇒ 00:24:19.109 Robert Tseng: Yeah, we’ll say, Go ahead again.
128 00:24:19.110 ⇒ 00:24:36.590 Connor Nevelle: Well, I know we have 2 min. We want to do 5 min, Demo. I do have to hop at the 30. But maybe, Ben, if you could. I don’t know if you have another call. If I have to hop you could stay on in terms of next steps, like what you’re building definitely sounds interesting. It’d be great to grab, you know, an hour of your time, probably with me and Ben.
129 00:24:36.650 ⇒ 00:24:51.969 Connor Nevelle: maybe someone else. And we can walk you through like blueprints, internal processes, and would love for you to treat us like a company that you’re running an assessment for, and then maybe, like, suggest one or 2 action items. And if you even created like
130 00:24:52.050 ⇒ 00:25:03.736 Connor Nevelle: one of these, I don’t know what the lift is of the time involvement like, don’t we? Spending a lot of time on like a trial consulting, but just so that we can get a sense of what it would look like to work with our portfolio companies.
131 00:25:04.190 ⇒ 00:25:20.928 Connor Nevelle: would love to do that as a next step. So maybe I know I have your calendar, Robert. I can just grab, grab some time for this week or next week. Would love you as a next step, but would also love to see some some demos of things you’ve built in the past sounds like you understand what we’re trying to accomplish.
132 00:25:21.530 ⇒ 00:25:22.200 Uttam Kumaran: Yeah, yeah.
133 00:25:22.200 ⇒ 00:25:24.939 Robert Tseng: I’m just scheduling with you, and we, Tom, can do a demo real quick. Yeah.
134 00:25:24.940 ⇒ 00:25:28.150 Uttam Kumaran: Yeah, let me share a couple. Let me share a couple of things. So
135 00:25:32.268 ⇒ 00:25:37.520 Uttam Kumaran: so probably I’ll start with like something that I think we’re using a lot internally, which is
136 00:25:38.013 ⇒ 00:25:43.679 Uttam Kumaran: just like any company. We have a ton of meetings. We’ve tried Granola. We’ve tried zoom stuff.
137 00:25:44.135 ⇒ 00:25:51.520 Uttam Kumaran: And ultimately what I wanted is a quick place for people in my company to go see the meetings that are happening.
138 00:25:51.920 ⇒ 00:26:15.309 Uttam Kumaran: Who’s in the meeting? What was talked about quickly, be able to chat with the meeting and then take some resulting actions from there. So this start off is just something someone on my team. We just we just by coded together, and we built what you probably don’t see here is all these meetings or things were happening either internally or clients. For example, this is a conversation that I led this morning about something on the data observability side.
139 00:26:15.558 ⇒ 00:26:23.500 Uttam Kumaran: These are all folks on our team immediately. What you get is like, what if I had to pay for granola for everybody? Which is like chat with your meeting right? So
140 00:26:23.540 ⇒ 00:26:27.680 Uttam Kumaran: right at that point I’ve eliminated. I don’t know. 20 bucks a month per person per month.
141 00:26:27.760 ⇒ 00:26:45.782 Uttam Kumaran: Second piece is we now can see the transcript, so people can quickly. Go here, copy this into their chat, Gbt. Or whatever, and take this with them. It’s like pretty clean. This is all hitting the Zoom Api, which is completely free, and we’re storing this in s. 3, which is near free.
142 00:26:46.240 ⇒ 00:26:59.409 Uttam Kumaran: and so the other thing you could do is common workflow is we wanted. Our Pm’s have to go, sit through, look through their notes, or look through a meeting and generate tickets. So we built a little thing that you can click on here. This meeting is a little bit long, so may take a sec.
143 00:26:59.510 ⇒ 00:27:02.649 Uttam Kumaran: But it’ll just basically go through the transcript.
144 00:27:03.018 ⇒ 00:27:30.859 Uttam Kumaran: And we run all of our projects on linear. So it’ll basically generate a bunch of sample tickets that our project managers can edit tweak and then create. This is a process that I was a product manager for a while. This is a process that takes quite a long time, as you can see here. These are a couple of things we talked about this morning, for example, establish a rotating on call system. Your Pm. Can go in here and edit this, and this isn’t a product. This is something that we built in like 2 weeks.
145 00:27:30.990 ⇒ 00:27:32.219 Uttam Kumaran: But what what does.
146 00:27:32.220 ⇒ 00:27:35.180 Connor Nevelle: What platform is this like? What is this rival.
147 00:27:35.180 ⇒ 00:27:40.010 Uttam Kumaran: This is just running on our own heroku like this is just. This is not a platform. This is like.
148 00:27:40.429 ⇒ 00:27:40.850 Connor Nevelle: Okay.
149 00:27:40.850 ⇒ 00:27:48.479 Uttam Kumaran: We just. We just built this for us, right? But there’s not a product like this that exists, and I don’t know what I would have to pay for it. And it took us
150 00:27:48.690 ⇒ 00:27:50.180 Uttam Kumaran: 2 weeks to build this.
151 00:27:50.410 ⇒ 00:27:51.410 Uttam Kumaran: Yeah, what did it?
152 00:27:51.410 ⇒ 00:28:14.349 Uttam Kumaran: 2 weeks, though, is the fact that I’m leading this. I know what the pain points for our project managers are. It’s so painful to go to meetings, sit through them, have all these notes creating these tickets in the right project in linear is such a pain with the right assignee. And now you can just go and click immediately, create it, and immediately open it and go see it in linear.
153 00:28:15.470 ⇒ 00:28:19.979 Connor Nevelle: Yeah, we we have this like, you see this, the fireflies that’s in this.
154 00:28:19.980 ⇒ 00:28:20.690 Uttam Kumaran: Yes.
155 00:28:20.690 ⇒ 00:28:22.299 Connor Nevelle: Way less clean, but
156 00:28:22.400 ⇒ 00:28:33.690 Connor Nevelle: goes from here goes to Zapier. I need to put it on an end instead of Zapier finds like the account that we’re speaking with, and updates a couple of fields, but it’s clunky. If it breaks, it’s in a black box. We don’t know what.
157 00:28:33.690 ⇒ 00:29:00.890 Uttam Kumaran: You have to go to fireflies to do that. Yeah. And it’s like, it’s really painful. So this is an opportunity where I’m like, I was okay with the build versus buy another example of something. You know, we have several demos on like a demo site. Here’s another example that we built for like a a potential medical clinic. This is all like synthetic data. But we had a medical clinic that was like, Hey, I want you to take these Pdfs that I’m getting, which have, like medical records.
158 00:29:01.080 ⇒ 00:29:08.440 Uttam Kumaran: put them into a very standardized data format. And then, like you could just click on standardize. And it’ll actually clean a lot of this up.
159 00:29:08.480 ⇒ 00:29:32.769 Uttam Kumaran: And so this is actually again, synthetic data. But like basically like live standardization, what I’m showing here is skipping sort of probably a lot of the sort of like mn, like, oh, generate 100,000 lead that you can cold email like moving past that to actually building like bespoke tooling for your clients. And here’s like a version of like the standardized data that
160 00:29:32.880 ⇒ 00:29:53.359 Uttam Kumaran: they would want going from a big, like basic, unstructured data set into something that they can copy and paste or extract and move into a system. Also, again, we would then immediately hook this up to like whatever this is going to, and you can click and and move it there. The other thing that we’re doing here is like we do a lot of work in slack. Here’s something we’re actually working on today.
161 00:29:53.726 ⇒ 00:30:18.890 Uttam Kumaran: Sprint updates for our clients and we do all the time. And Mustafa is on our team. And he’s working on basically pulling all the linear tickets for a given sprint and not only just shoving that in the Chat Gpt, and saying, Generate me a summary, but we have we have agents for every single one of our clients that has access to all of their Zoom Meetings, all of the slack channels that we’ve been discussing with them.
162 00:30:19.186 ⇒ 00:30:42.040 Uttam Kumaran: And all the tickets we’ve worked on all the code related to their project. So we have an agent that has all this rich context, that then when you give it like, here’s 10 tickets we’re working on, it’s gonna give you a really bomb summary of like what the sprint and like, what? Why, it really matters that our Pm’s can then copy. Make sure it’s right, and send doing this is again something that takes
163 00:30:42.580 ⇒ 00:30:45.519 Uttam Kumaran: at least an hour. If you are not like.
164 00:30:45.890 ⇒ 00:30:46.360 Connor Nevelle: Yeah.
165 00:30:46.360 ⇒ 00:30:50.899 Uttam Kumaran: Multitasking like crazy, you know. So I don’t know.
166 00:30:50.900 ⇒ 00:30:51.359 Connor Nevelle: Oh, my! Gosh!
167 00:30:51.360 ⇒ 00:31:17.350 Uttam Kumaran: We do. We do a bunch of work in clay as well like we use clay a lot for our go to market stuff. What you’re seeing here is we’re we’re trying to go after system integrators for some common data tools like mixed panel amplitude to kind of start the partnerships with them. We built some stuff around that to go Google, search access, search for like system integrators with amplitude, experience rank them by a bunch of parameters, and then it helps us sort of target them.
168 00:31:18.130 ⇒ 00:31:20.280 Uttam Kumaran: And then the last. Yeah.
169 00:31:20.922 ⇒ 00:31:26.060 Connor Nevelle: Should get a commission from them by now, for how many people I recognize them to? I?
170 00:31:26.060 ⇒ 00:31:27.509 Connor Nevelle: They’re good. They’re pretty good.
171 00:31:28.990 ⇒ 00:31:39.519 Uttam Kumaran: Yeah. And then the last thing is like we actually save, we have, like a big, prompt library in our company. So like, when when our company sort of like finds like, Oh, I where I’m I’m always using this like
172 00:31:39.800 ⇒ 00:32:04.779 Uttam Kumaran: like 1 1 thing, for example, I get asked to do a lot of intro. So I build a great prompt to do intros. I know Robert does. I know a couple of other people do. So we share that internally, we also have prompts that help you write better prompts. So if you have like a prompt that you wrote like sort of randomly, and you’re like, well, I want to use open AI’s best practices for using Xml for having rules and examples. We have prompts that you can literally paste your thing in some commentary, it’ll help you write through that.
173 00:32:04.800 ⇒ 00:32:12.340 Uttam Kumaran: So like, this is the stuff where, like internally, when you think about your company adopting these things have to be there, because
174 00:32:12.490 ⇒ 00:32:28.259 Uttam Kumaran: the random person who discovers chat should be today is gonna like, hit these walls that you hit probably a year ago. And then we’re gonna realize that. Like, Oh, oh, yeah, we have a prompt library with a great prompt on how to do like something for project scoping or something for case studies things like that, you know. So.
175 00:32:28.260 ⇒ 00:32:49.500 Connor Nevelle: Perfect. I I do have. I do have to hop. This is really interesting. I’d love. I’ll grab some time you next this week or next week. We have quite a bit of internal tools and optimizations. It’s turning into a bit of like spaghetti code just because it’s living in so many different places, but even when you just showed me like, if we could standardize it in that way, would be
176 00:32:49.840 ⇒ 00:32:53.630 Connor Nevelle: would be great. But I do have to hop but talk to you, talk to you soon.
177 00:32:53.880 ⇒ 00:32:54.540 Uttam Kumaran: Okay. Okay.
178 00:32:54.540 ⇒ 00:32:55.540 Robert Tseng: Alright! Thanks, Connor, thank you.
179 00:32:55.540 ⇒ 00:32:56.850 Uttam Kumaran: Thanks everyone, bye.