Meeting Title: Brainforge <> HelloFresh: AI-Automation Initiatives Date: 2025-08-11 Meeting participants: Sam Roberts, Uttam Kumaran, Rockaway Peach (United States-NYCHQ, 7)
WEBVTT
1 00:01:55.450 ⇒ 00:01:56.810 Uttam Kumaran: Hello!
2 00:01:59.140 ⇒ 00:02:00.270 Sam Roberts: Ehhh….
3 00:02:00.920 ⇒ 00:02:02.079 Uttam Kumaran: Hey, how are ya?
4 00:02:03.560 ⇒ 00:02:04.419 Sam Roberts: How you doing?
5 00:02:04.800 ⇒ 00:02:05.810 Uttam Kumaran: Good.
6 00:02:06.140 ⇒ 00:02:21.429 Uttam Kumaran: So yeah, for this meeting, it’s kind of gonna be a little bit of discovery. We got put into touch, with HelloFresh, actually by Robert’s wife, she works there, and they’re doing a lot of stuff
7 00:02:21.550 ⇒ 00:02:31.609 Uttam Kumaran: Around AI and AI initiatives, and really, it’s just gonna be a little bit of discovery. I’m gonna sort of start to ask some questions.
8 00:02:31.870 ⇒ 00:02:35.929 Uttam Kumaran: About, like, what’s… what they’re currently building.
9 00:02:36.130 ⇒ 00:02:47.279 Uttam Kumaran: I already have a little bit of a sense, but I’ll be asking, like, what they’re currently building, what challenges they’re having, like, you know, what are they… what are they interested in from a firm like ours, or where we could be helpful?
10 00:02:47.350 ⇒ 00:02:48.510 Uttam Kumaran: …
11 00:02:48.540 ⇒ 00:03:01.289 Uttam Kumaran: You know, and yeah, I think we’ll just probably start with a round of intros, but kind of two goals in, like, these sort of initial meetings is, one, just establishing ourselves as subject matter experts.
12 00:03:01.290 ⇒ 00:03:11.430 Uttam Kumaran: And then second, it’s going to really be, like, almost like a VIN session for them, where we try to dissect… we try to get here as much as possible, and then, …
13 00:03:11.470 ⇒ 00:03:21.510 Uttam Kumaran: you know, try to, come back and say, cool, like, let’s go prepare something, or prepare some things you can look at, and then, you know, get another follow-up book. So…
14 00:03:22.110 ⇒ 00:03:38.539 Uttam Kumaran: That’s probably it. I would say, you know, if you have any questions while we’re talking, feel free to ask. Okay. I think it’s helpful, you know, just so we’re both on the same page, and then we’ll probably take this back and see if we can prepare anything, you know, to share with them, so….
15 00:03:38.900 ⇒ 00:03:39.870 Sam Roberts: Cool, alright.
16 00:03:40.210 ⇒ 00:03:40.860 Uttam Kumaran: Damn.
17 00:03:42.580 ⇒ 00:03:43.340 Sam Roberts: It’s good.
18 00:04:15.340 ⇒ 00:04:17.640 Sam Roberts: But who, who are we meeting with today?
19 00:04:17.640 ⇒ 00:04:21.440 Uttam Kumaran: We’re meeting with Allie Day. She is…
20 00:04:29.970 ⇒ 00:04:40.090 Uttam Kumaran: She’s the Director of Pathways at HelloFresh, so… partnerships, operational improvements… Fair enough.
21 00:05:58.640 ⇒ 00:06:01.260 Sam Roberts: this properly, I’m not confident.
22 00:06:01.260 ⇒ 00:06:01.890 Uttam Kumaran: what?
23 00:06:01.890 ⇒ 00:06:04.840 Sam Roberts: I thought I didn’t time this properly, I should have made coffee right before.
24 00:06:05.340 ⇒ 00:06:06.229 Sam Roberts: how low I was getting.
25 00:06:06.230 ⇒ 00:06:12.549 Uttam Kumaran: I usually grab 45 minutes, because sometimes it can go a bit longer, but…
26 00:06:12.730 ⇒ 00:06:15.130 Uttam Kumaran: It’s probably gonna be closer to 30 minutes.
27 00:06:15.130 ⇒ 00:06:22.599 Sam Roberts: Okay, no, we’re good. I just… I usually make it and bring it upstairs to my office in this little thermos, and I didn’t realize it was empty. I thought I had another cup in there.
28 00:06:29.650 ⇒ 00:06:30.049 Sam Roberts: E.
29 00:06:30.760 ⇒ 00:06:36.620 Uttam Kumaran: Yeah, today’s a big meeting day for me, but it’s been good, like, we have a… we have several clients in, so just like…
30 00:06:37.350 ⇒ 00:06:52.849 Uttam Kumaran: working my tail off so, like, we can get you in full-time, and I just want to lock that down and give you some clarity, so… Yeah. We’re close. I don’t know if you’re… I can add… I don’t… I sometimes don’t want to add people to every channel, but there’s a sales channel if you want to join, you can kind of see, but….
31 00:06:52.990 ⇒ 00:06:54.350 Sam Roberts: Okay, a lot of….
32 00:06:54.990 ⇒ 00:07:00.740 Uttam Kumaran: Lot of stuff going on, and maybe it’ll give you some understanding of, like, the types of people we’re pitching.
33 00:07:01.040 ⇒ 00:07:06.630 Rockaway Peach (United States-NYCHQ, 7): Hi, … okay, good, the video’s working.
34 00:07:06.950 ⇒ 00:07:07.750 Uttam Kumaran: Hey!
35 00:07:09.070 ⇒ 00:07:10.269 Rockaway Peach (United States-NYCHQ, 7): What’s going on?
36 00:07:10.270 ⇒ 00:07:11.300 Uttam Kumaran: How’s it going?
37 00:07:11.790 ⇒ 00:07:30.470 Rockaway Peach (United States-NYCHQ, 7): Good, good. I’m Ali Day. I am, I don’t think we’ve met before, but I guess the intro would be a great place to start. I am in the Pathways program, here at HelloFresh. I’m trying to remember, someone on the team had reached out to
38 00:07:30.470 ⇒ 00:07:34.709 Rockaway Peach (United States-NYCHQ, 7): provide an intro, since, one of the projects I’m working on is the
39 00:07:34.710 ⇒ 00:07:54.459 Rockaway Peach (United States-NYCHQ, 7): focus on AI operations and identifying opportunities, to layer in tools, but also upskill, folks in the growth marketing team to, really enhance, our operations from, you know, performance management perspective, research and analysis.
40 00:07:54.460 ⇒ 00:07:59.209 Rockaway Peach (United States-NYCHQ, 7): QA, and, you know, improving our,
41 00:07:59.210 ⇒ 00:08:03.939 Rockaway Peach (United States-NYCHQ, 7): You know, speed to market, and all things that, you know, line up to our marketing objectives.
42 00:08:03.940 ⇒ 00:08:10.570 Uttam Kumaran: Cool. Yeah, and maybe intro from our side. Yeah, I think it’s Rachel, put us in touch, Rachel Sao, …
43 00:08:10.570 ⇒ 00:08:26.009 Uttam Kumaran: she’s actually, married to my business partner, Robert. Okay. This is, like, you know, it’s funny because even my partner, like, they probably don’t know what we’re even working on, and she’s like, aren’t you guys doing AI stuff? And we’re like, yes, that’s the whole company.
44 00:08:26.010 ⇒ 00:08:44.779 Uttam Kumaran: She’s like, well, we’re at HelloFresh, so we’re looking at some stuff, so I’m like, I’m happy to be helpful. So, yeah, Brainforge is a company I started about 2 years ago. We’re primarily started as a data analytics, consultancy. That’s my background. I led several data teams in New York before this, led product and a data startup, and then
45 00:08:44.780 ⇒ 00:09:02.409 Uttam Kumaran: sort of in the last year, we actually were using AI very heavily to build a business, in a lot of the same ways that you described, in kind of every nook and cranny, in operations, for finance, a lot on the sales and marketing side, and additionally on, like, the engineering, given we’re, you know, we’re an engineering firm.
46 00:09:02.410 ⇒ 00:09:05.019 Uttam Kumaran: And through that process, I found that
47 00:09:05.020 ⇒ 00:09:08.399 Uttam Kumaran: Okay, we should go offer this as a service, because I’m a business owner.
48 00:09:08.400 ⇒ 00:09:31.500 Uttam Kumaran: who is pretty technical, and we do have this great team, but it’s hard to get adoption, it’s hard to take the right bets, and so that’s what we’ve been doing for over a year now, is we have several clients that we’re working on AI initiatives, and AI initiatives and automation initiatives can have several flavors. So, on one hand, that can just be sort of just strategy, where we’ve tried and messed up a lot of things, and
49 00:09:31.500 ⇒ 00:09:37.089 Uttam Kumaran: We have a lot of experience, so it’s one, just being a resource there. Second, we also come in and develop
50 00:09:37.090 ⇒ 00:09:53.880 Uttam Kumaran: proof of concepts, you know, applications, but all geared towards internal, sort of, optimization and internal efficiency. So, tons of stuff I can talk about, and you actually have Sam here on the call. Sam, if you want to give a brief intro, Sam is a lead on our AI team.
51 00:09:55.090 ⇒ 00:10:10.889 Sam Roberts: Yeah, hi, Sam Roberts. I’ve been, an engineer, software engineer, web dev engineer, however you want to call it, for over a decade, and obviously more recently, jumping into AI stuff in the past 5 years. So I have a decent amount of experience in startups,
52 00:10:10.960 ⇒ 00:10:17.299 Sam Roberts: from various roles, from development all the way up to CTO, and I’ve joined Brainforge fairly recently to help.
53 00:10:17.460 ⇒ 00:10:19.520 Sam Roberts: Build out this, AI…
54 00:10:20.070 ⇒ 00:10:21.989 Sam Roberts: Team, and manage it moving forward.
55 00:10:23.180 ⇒ 00:10:24.600 Rockaway Peach (United States-NYCHQ, 7): Nice.
56 00:10:24.600 ⇒ 00:10:35.999 Uttam Kumaran: Yeah, so I guess, Ali, like, tell me the best place to start. I have your notes, I sort of heard kind of what you said, but what would be helpful, like, examples of stuff we’ve done, or even give me a sense of, like.
57 00:10:36.150 ⇒ 00:10:40.260 Uttam Kumaran: What’s been painful through the process for your side to figure this all out?
58 00:10:40.570 ⇒ 00:10:58.919 Rockaway Peach (United States-NYCHQ, 7): Yeah, I’d love to, like, start with, like, getting a… I could definitely, like, share a little bit of what I’ve seen so far. I’ve been on this project for about 2 months, but another piece is, like, I’d love to, like, understand the way you work and, like, maybe some sample use cases of how you’ve worked with similar,
59 00:10:58.920 ⇒ 00:11:12.029 Rockaway Peach (United States-NYCHQ, 7): teams, like, particularly, I work on a growth marketing team focused in new customer acquisition. So we’re focused on, you know, most of our paid media channels, influencer marketing.
60 00:11:12.030 ⇒ 00:11:18.979 Rockaway Peach (United States-NYCHQ, 7): paid search, SEO, all those types of great things. …
61 00:11:19.070 ⇒ 00:11:42.369 Rockaway Peach (United States-NYCHQ, 7): I’d love to, like, hear some examples of, like, how you’ve integrated with teams, like, what solutions that you’ve brought forward, and what technology, that you use, like, laying the groundwork from, like, things that I’ve seen, like, we do… I’d say, like, our maturity level is kind of, like, in the middle, in the sense that, like, A, we have, like, a really clear AI policy here at Total Fresh.
62 00:11:42.570 ⇒ 00:11:55.470 Rockaway Peach (United States-NYCHQ, 7): We have some enterprise-wide solutions that we leverage, that create, like, data-secure environments. We also have, like, a few ad hoc, you know,
63 00:11:55.470 ⇒ 00:12:06.170 Rockaway Peach (United States-NYCHQ, 7): different tools that we, like, allow, you know, teams for exploration, and we encourage people to, you know, put in requests for security approvals for new tools, and manage those
64 00:12:06.170 ⇒ 00:12:14.339 Rockaway Peach (United States-NYCHQ, 7): SaaS budget, so I feel like we’re embracing the change, but we’re at, like, an inflection point of…
65 00:12:14.340 ⇒ 00:12:28.359 Rockaway Peach (United States-NYCHQ, 7): like, moving from, like, prototypes and ideas to scalable solutions, and that’s where I’m really focused on, like, how do I bring those forward? And then how do I make it easy to, integrate? I think there’s a lot of,
66 00:12:28.530 ⇒ 00:12:48.329 Rockaway Peach (United States-NYCHQ, 7): folks that, you know, have been burned, this last 6 months have been transformational, in the sense of, okay, in theory, I can do X, but then when I go to try to do X, like, I get a really crappy outcome, right? So, there’s a little bit of, like… but now, the models are, like.
67 00:12:48.390 ⇒ 00:13:12.220 Rockaway Peach (United States-NYCHQ, 7): 25 times better, and I can… I have real-world use cases where, like, I’m trying to do simple analysis for tests, where 6 months ago it was awful, and, you know, now I’m getting what I want. I also have a better understanding of, like, how to structure my data. So there’s an element that goes hand-in-hand with having the right tool, but also teaching people how to use those tools.
68 00:13:12.560 ⇒ 00:13:28.340 Rockaway Peach (United States-NYCHQ, 7): And at the same time, these tools are getting even smarter, so the people that want to use them need to know less, right? So, that’s, like, the kind of weird loop that we’re in. But I know that, like, the faster we move, the more we experiment, the better we’ll be in the long run.
69 00:13:28.340 ⇒ 00:13:34.449 Rockaway Peach (United States-NYCHQ, 7): If more people are learning at the same rate I’m learning, like, the business will be in a better place. …
70 00:13:34.550 ⇒ 00:13:45.989 Rockaway Peach (United States-NYCHQ, 7): So that’s, you know, the mindset that we’re embracing, and I wanted to hear from you, like, you know, I guess I’ll throw it back over to you now, and hear your side and how you approach the situation.
71 00:13:46.220 ⇒ 00:14:05.629 Uttam Kumaran: Yeah, so maybe I’ll just start off with, like, an example of, you know, kind of when we walk into, I would say, a similar org like Coors, we tend to do a lot… a lot of the AI work is actually pulling a lot of the work we do on the data side forward, so maybe I can just start with that, and when we come in, and you probably have a very similar system to this, is you have tons of data sources where you’re running.
72 00:14:05.630 ⇒ 00:14:10.210 Uttam Kumaran: spend, you have organic, you have probably Google Sheets or some manual tracking for things.
73 00:14:10.210 ⇒ 00:14:12.359 Uttam Kumaran: All of that gets pulled into a warehouse.
74 00:14:12.360 ⇒ 00:14:30.020 Uttam Kumaran: and you have, sort of, data marts for different sections of the business. All of that gets visualized in a BI tool, gets sent through some reverse ETL, maybe back into a CRM. But another point here is that, actually, what we’re finding in a lot of clients is that AI becomes another output of the data, and that
75 00:14:30.040 ⇒ 00:14:35.210 Uttam Kumaran: Just like you mentioned, you want to run analysis over specific data, you want to
76 00:14:35.400 ⇒ 00:14:50.740 Uttam Kumaran: ask questions over a set of documents. You also want to use it for brainstorming. So, what we found, even in our company and in clients, is that it’s still very necessary for you to have a really great understanding of all your data sources.
77 00:14:50.740 ⇒ 00:15:10.589 Uttam Kumaran: In fact, I think AI actually forces you to understand even what data is not being reported on. So, if you want to ask over documents, company policies, procedures, you know, you need to have a really good understanding of your data. So, for us, it starts with this. So, getting a really good understanding of what the current
78 00:15:10.590 ⇒ 00:15:14.800 Uttam Kumaran: data mart and data lifecycle is, you know, at HelloFresh.
79 00:15:14.830 ⇒ 00:15:28.499 Uttam Kumaran: And then the next piece is really understanding, like, what are the metrics we’re trying to affect? You know, we have several clients that, you know, they call us and they say, like, hey, we want to implement AI, or I have a board directive to do this.
80 00:15:28.500 ⇒ 00:15:39.040 Uttam Kumaran: But where they… that’s sort of where they end. And so, for us, it’s helpful to understand what is the KPI you’re trying to affect, and ideally, that’s one measurable thing that we can attack.
81 00:15:39.040 ⇒ 00:15:47.929 Uttam Kumaran: Whether it is time towards a process, whether it is maybe, like, hey, we want to make sure that a… you know, in my company, right, the KPI for us is.
82 00:15:47.930 ⇒ 00:16:07.470 Uttam Kumaran: I want to see that everybody’s using an AI tool every day. One of several that we have available, right? So that’s a measurement problem. But for several companies, additionally, we have a company that is a home services, large home services company, here in Texas, and for them, they want to look at the amount of times their customer service reps
83 00:16:07.470 ⇒ 00:16:10.869 Uttam Kumaran: use the AI to answer a question, and there’s a positive resolution.
84 00:16:10.870 ⇒ 00:16:13.629 Uttam Kumaran: And so, the first thing is, like, what is the…
85 00:16:13.880 ⇒ 00:16:22.579 Uttam Kumaran: what is the core KPI? Because broadly, everyone using AI is a nice, like, anecdotal thing to get a feeling about, but
86 00:16:22.830 ⇒ 00:16:40.119 Uttam Kumaran: there’s no way you’ll know whether you’ve… you’ve hit the nail on the head or not. And so that’s probably my question back, is like, is there… do you have a clear KPI in mind on what you’re tackling? Are you measuring that today? Or is that sort of still being decided?
87 00:16:40.290 ⇒ 00:16:48.569 Rockaway Peach (United States-NYCHQ, 7): Yeah, we definitely have very clear KPIs around, time savings driven by AI.
88 00:16:48.920 ⇒ 00:16:56.550 Rockaway Peach (United States-NYCHQ, 7): That those are quantifiable, and we have measurement tools in place, at least, like, at a high level, to understand progress.
89 00:16:56.680 ⇒ 00:16:59.989 Rockaway Peach (United States-NYCHQ, 7): We also have…
90 00:16:59.990 ⇒ 00:17:20.320 Rockaway Peach (United States-NYCHQ, 7): aspiration on, like, additional projects. Like, maybe it’s not about just saving time that we’re doing today, but it’s also now about, like, increasing productivity. So, say we didn’t have time to do XYZ because we were so busy doing ABC, now that ABC takes, you know, half an amount of time, we can go do XYZ, right? Now we’re a smarter
91 00:17:20.319 ⇒ 00:17:21.700 Rockaway Peach (United States-NYCHQ, 7): more strategic team.
92 00:17:22.200 ⇒ 00:17:22.770 Uttam Kumaran: Yeah.
93 00:17:23.190 ⇒ 00:17:42.070 Uttam Kumaran: So, I mean, given that you have KPIs, I think for me, it’s a good understanding about what you mentioned, which is moving from, sort of, you kind of have two things. One, you want to enable anybody in your company to use AI tools, whether it’s basically just ChatGPT, or at maximum, it’s, like, some internal tool that’s been built, right? So that’s more of, like.
94 00:17:42.070 ⇒ 00:17:50.420 Uttam Kumaran: this platform enablement. The other piece is, some of these proof of concepts, you may now want to develop into a product. So what is the stack?
95 00:17:50.520 ⇒ 00:17:52.620 Uttam Kumaran: Like, who needs to be staffed on there?
96 00:17:52.670 ⇒ 00:18:04.999 Uttam Kumaran: what is the strategy around that? So we kind of support in two areas. One is a lot of our work starts at the strategy level. So we come into clients where maybe they have no understanding of
97 00:18:05.010 ⇒ 00:18:20.039 Uttam Kumaran: of, like, what tools are on the market, what the opportunities are. It’s clear that you guys are past that, and you guys have developed, you know, things internally. I think the second piece is certainly running some sort of cross-functional council is what we typically, you know, like to encourage, is that
98 00:18:20.040 ⇒ 00:18:38.840 Uttam Kumaran: there are teams… there are people from separate teams that are able to contribute ideas, and they have some sort of monthly or some cadence at which you can get that, and also you can share out wins. But if alignment is not a problem, and it’s really, hey, we have these proof of concepts, we need to take them to the next level, then it becomes more of…
99 00:18:38.840 ⇒ 00:18:52.900 Uttam Kumaran: you know, engineering, you know, work, which is, okay, we have maybe someone vibe-coded something on their laptop, it worked in a demo, now we want to roll this out to the entire company in a secure environment. You know, what does that look like?
100 00:18:52.930 ⇒ 00:18:58.150 Uttam Kumaran: So do any of those kind of, you know, more ring a bell than the other?
101 00:18:58.240 ⇒ 00:19:10.139 Rockaway Peach (United States-NYCHQ, 7): I would say we’re at, like, that last stage of, like, we have people live coding that we want to build out in secure ways, and, you know, part of the things that I’m thinking about is, like.
102 00:19:10.180 ⇒ 00:19:19.739 Rockaway Peach (United States-NYCHQ, 7): you know, if I make a… like, how do we get engineering help, right? And how do we get people who understand these systems? Or, like, we have access to a tool called N8N.
103 00:19:19.740 ⇒ 00:19:20.170 Uttam Kumaran: Yes.
104 00:19:20.170 ⇒ 00:19:38.909 Rockaway Peach (United States-NYCHQ, 7): And, like, how do we, you know, and it’s… you know, there’s a lot of issues, not issues, there’s a lot of challenges that you run into, like, on the technical side, when you’re trying to create a strategic idea, and then you go and try to set it up, and you’re like, oh my god, I don’t have the credentials, like, why am I in Google Cloud Services?
105 00:19:38.910 ⇒ 00:19:39.690 Uttam Kumaran: What is this?
106 00:19:39.840 ⇒ 00:19:48.449 Rockaway Peach (United States-NYCHQ, 7): security code, why is it not working? You know, like, that kind of stuff is, like, where I could see there being a real business need in the future.
107 00:19:48.450 ⇒ 00:20:03.950 Rockaway Peach (United States-NYCHQ, 7): like, the near-term future of, okay, we have this idea, we’ve shown the tools can do it, let’s plug the idea into NAN, let’s train the team how to use it, and, like, create all these, like, end-to-end workflows, or at least, like, partially end-to-end workflows with another…
108 00:20:03.960 ⇒ 00:20:08.199 Rockaway Peach (United States-NYCHQ, 7): thing, you know, we were at Gemini as our enterprise tool, so…
109 00:20:08.200 ⇒ 00:20:29.879 Rockaway Peach (United States-NYCHQ, 7): when we create, a really powerful agent within Gemini that we want the team to use, we can’t share it right now, right? So, I… in my mind, I hope that, like, one day you can share it, but for now, okay, what’s the solution to that? Like, build an N8N and create the prompt in N8N so that it lives in a shareable cloud, right?
110 00:20:29.880 ⇒ 00:20:30.480 Uttam Kumaran: Yeah.
111 00:20:30.480 ⇒ 00:20:33.009 Rockaway Peach (United States-NYCHQ, 7): So, like, but that’s…
112 00:20:33.200 ⇒ 00:20:40.000 Rockaway Peach (United States-NYCHQ, 7): I want to do that, like, 10 times a week, right? So, and I don’t know how to do N8N, so, like, how do we get from this.
113 00:20:40.000 ⇒ 00:21:04.650 Uttam Kumaran: Yeah, so it’s actually a great point. So, one, it’s a great decision going with N8N. You know, we use N8N very frequently. In fact, it sort of starts off in a very similar way. The reason we chose N8N among all the sort of vendors, and even chose sort of a low-code, no-code tool, and I’m not really a big fan, I would rather just have everything as code, but what we found is that we want to enable people to build
114 00:21:04.650 ⇒ 00:21:05.250 Uttam Kumaran: gold.
115 00:21:05.250 ⇒ 00:21:18.840 Uttam Kumaran: And code is actually the limiting factor. But the great thing about N8N as an engineering decision is it has the flexibility to do basically everything you need. But second, you will need some support from the engineering team
116 00:21:18.840 ⇒ 00:21:35.170 Uttam Kumaran: To make sure that it’s provisioned in the right place, that all the users can get provisioned, that they have right access controls, and that there are certain, like, platform implementations, right? Okay, if you need to talk to the database, here is, like, our HelloFresh database node, right, that’s, like, certified by the team.
117 00:21:35.180 ⇒ 00:21:50.240 Uttam Kumaran: Similarly, you know, in our business, we use NNN for tons of stuff, and I would say, as NNN workflows stabilize, we then move it to code, right? But until it gets stabilized in there, where there edits to prompts, and you want that to be happening
118 00:21:50.240 ⇒ 00:22:06.929 Uttam Kumaran: so frequently, and then once it sort of works, it could then probably get moved to a more secure, stable environment, and that’s… that’s usually our process. But in terms of, you know, NN1 is, like, really having a great, you know, support system between
119 00:22:06.980 ⇒ 00:22:17.169 Uttam Kumaran: everyone building, and sort of this, like, AI sort of department, and the actual traditional engineering department. Because they’re going to be the gatekeepers behind access to databases.
120 00:22:17.230 ⇒ 00:22:21.810 Uttam Kumaran: Access to different platform resources, and certainly it’s gonna be…
121 00:22:21.810 ⇒ 00:22:39.060 Uttam Kumaran: important to get them involved in how N8N is going to be hosted, and what resources they’re going to need to actually run those workflows. But this is… that’s something that, you know, we spend a lot of time in N8N doing. But I think you’re right in that it’s… it’s like, how do you set N8N up as a platform?
122 00:22:39.080 ⇒ 00:22:53.259 Uttam Kumaran: for any team member to come in and build their first workflows. Moving out of, like, Gemini agents, right? As soon as they get a sense of something’s working, how do we actually move it into an environment where maybe it can chat with your Slack, it can chat with your document store, things like that?
123 00:22:53.260 ⇒ 00:23:11.700 Rockaway Peach (United States-NYCHQ, 7): Yeah. Like, and, like, we are… we have an engineering prop team that runs at N, and they are quick to adopt it, and they’re doing things to enhance the experience, so I feel like they would probably be a partner. I guess my question for you, and I think one of our blockers is, like.
124 00:23:11.820 ⇒ 00:23:23.790 Rockaway Peach (United States-NYCHQ, 7): the team is busy, and they don’t have time to, debug things. Yes. Right? And that’s frustrating, right? It’s gonna take them 10 hours to create an NAN clone and…
125 00:23:23.790 ⇒ 00:23:36.289 Rockaway Peach (United States-NYCHQ, 7): and work it out, but the report’s due now. So, how, like, do you guys ever come in, like, a resource of, okay, if I give you… and I just ran this whole project with my interns, where…
126 00:23:36.290 ⇒ 00:23:52.199 Rockaway Peach (United States-NYCHQ, 7): we did, like, a discovery call, then we did… we documented the current SOP, then I had them research tools, and then created, like, a new SOP with using AI, and the goal was for the outcome to be, like, automated and N8N.
127 00:23:52.200 ⇒ 00:24:00.309 Rockaway Peach (United States-NYCHQ, 7): Yes. Granted, I’m doing this with interns. I feel like if I was doing it with people who actually have the business knowledge and real engineers.
128 00:24:00.310 ⇒ 00:24:05.079 Rockaway Peach (United States-NYCHQ, 7): That, you know, 8-week process would take 2 days, or 3 days, right?
129 00:24:05.080 ⇒ 00:24:22.950 Uttam Kumaran: Yeah, so let me be two parts there. One is, that’s… it’s a very similar… I would say the process you got right is, like, one, many times these SOPs may not even exist for existing workflows, so there is a pretty heavy, like, step-by-step discovery process, and then, you know, kind of what we do in our team is we basically
130 00:24:23.000 ⇒ 00:24:24.719 Uttam Kumaran: Kind of sketch it out, and then…
131 00:24:24.730 ⇒ 00:24:31.730 Uttam Kumaran: the goal is, like, how to cut this by 50%, and… but what also, what is the 80-20 of automation? Because what you’ll find…
132 00:24:31.730 ⇒ 00:24:34.150 Uttam Kumaran: Is that the last piece of automation
133 00:24:34.150 ⇒ 00:24:57.100 Uttam Kumaran: you know, maybe the, like, the… because you, like, have to input something in a CSV somewhere. Okay, maybe that’s, like, worth leaving for a human, but everything up to it, let’s try to automate. And so, we build these trade-offs, and typically, for our clients or the stakeholders to understand what are the trade-offs on how much level of automation, but also things like human in the loop, alerting, and then, of course, like, governance.
134 00:24:57.200 ⇒ 00:25:11.490 Uttam Kumaran: And then reporting, right? So one of the things that we try to do is, how do you report out that your agents are being used on errors? Because as soon as you throw something like this into the wild and people use it, there will 100% be problems, and it won’t be perfect.
135 00:25:11.490 ⇒ 00:25:28.569 Uttam Kumaran: So what is the cadence and the SLA in which your team is tackling issues? So that’s exactly, like, a perfect area to utilize us. I think, one, you know, we’re a traditional engineering firm in that I’ve worked with a lot of those folks before, and so whatever time that they’re taking to sort of ramp up and figure out N8N,
136 00:25:28.600 ⇒ 00:25:42.430 Uttam Kumaran: I can give them, you know, the answers there and say, hey, this is where we need help. The other thing is, how do we enable anyone to come in and actually build on NNN and go through that SOP to, like, AI SOP to then building
137 00:25:42.430 ⇒ 00:25:50.179 Uttam Kumaran: and have, like, how do you plug into, like, a data stream where we can actually measure the results? That’s a great place to use us.
138 00:25:50.180 ⇒ 00:26:01.950 Uttam Kumaran: In two ways. One is, you know, myself or Sam both are, you know, at, like, the solution architect level, where we could come in and either run those meetings or basically be the integrator.
139 00:26:01.950 ⇒ 00:26:11.679 Uttam Kumaran: and keep pushing things along, and we sort of speak both languages. I think also we have, you know, a lot of technical talent that all we do every day is a lot of in-and-done work.
140 00:26:11.720 ⇒ 00:26:29.170 Uttam Kumaran: And so whether it’s like, hey, we have to run a custom webhook, or we have to call an external function, or we’re stuck, like, how do you do this in NAN? It’s a great place to use this as well. But what I’m hearing is that, you know, for me, what would be most important is, like, how do we build with your team? Like, how do we…
141 00:26:29.170 ⇒ 00:26:41.660 Uttam Kumaran: like, you know, you could… we could come on and build the things, but it’s like, how do we do it with your team, and how do we coach them to start building the next most advanced integration? And so, you know, of course, like, if I have a couple of
142 00:26:41.700 ⇒ 00:26:51.699 Uttam Kumaran: example workflows I can share with you some things that we built, but those are the two ways that I would say you could best utilize us, is sort of in that integrator, like.
143 00:26:51.850 ⇒ 00:26:56.709 Uttam Kumaran: someone that speaks both sides of the house and can drive that forward, and then additionally, just, like.
144 00:26:56.980 ⇒ 00:27:05.110 Uttam Kumaran: what is the tech problem that’s happening with NIDN that we need to, like… can someone just tell me, like, what’s… how do we get this done, you know?
145 00:27:05.920 ⇒ 00:27:20.529 Rockaway Peach (United States-NYCHQ, 7): Cool. Yeah, I think this was a great conversation, like, you know, your technical expertise, I think, could be of value. I’d love to see, like, some samples of, like, exit workflows, that you’ve,
146 00:27:20.530 ⇒ 00:27:45.359 Rockaway Peach (United States-NYCHQ, 7): aligned, you know, aligned into, like, particularly, like, I don’t know if you have to… if you can share any, like, larger companies that you’ve worked for. Obviously, like, we’re a billion dollar, you know, $7.5 billion business, so we need to ensure, like, we’re vetting, like, enterprise-wide solutions. Also, any information you have on, like, your approach to, like, data privacy and security. Like, when you guys come in.
147 00:27:45.510 ⇒ 00:27:58.389 Rockaway Peach (United States-NYCHQ, 7): like, what do you need access to, right? Or is it, like, can you just work on our servers and our environment without exposing our company to risk, because there is going to be, you know, a lot of data sharing?
148 00:27:58.770 ⇒ 00:28:12.280 Rockaway Peach (United States-NYCHQ, 7): And I think, and you know, one more, like, big solution that we are looking to solve right now is marrying the data and the context, and creating, like, a better structure for, okay, I have all the metrics for the data, but, like.
149 00:28:12.280 ⇒ 00:28:20.479 Rockaway Peach (United States-NYCHQ, 7): how do I get a really good description of the creative so that I can actually have insights and recommendations? Because right now, they’re just very, like, separately of, like.
150 00:28:20.480 ⇒ 00:28:33.259 Rockaway Peach (United States-NYCHQ, 7): poor, you know, that’s the manual piece of, like, poor descriptions of the creative. Brying with really good data leads to really bad, you know, or mediocre insights or recommendations. So, that’s, like, if you have a use case around that.
151 00:28:33.330 ⇒ 00:28:36.720 Rockaway Peach (United States-NYCHQ, 7): That would be awesome, but anything that you guys
152 00:28:36.810 ⇒ 00:28:58.820 Rockaway Peach (United States-NYCHQ, 7): have, related to, like, performance management, test management on, like, the paid media channels, anything end-to-end automation of some of those processes. We also work a lot on, like, monthly business reviews and weekly business reviews, so, like, a lot of the inputs there are very focused on, like, high-level metrics.
153 00:28:58.820 ⇒ 00:29:09.359 Rockaway Peach (United States-NYCHQ, 7): But obviously, like, the reason why, like, lives somewhere else. So trying to, like, plug in, like, the contextual knowledge of, like, the human brain in a structured way to mirror the data is…
154 00:29:09.360 ⇒ 00:29:11.270 Rockaway Peach (United States-NYCHQ, 7): I think it’s, like, that’s where…
155 00:29:11.290 ⇒ 00:29:24.080 Rockaway Peach (United States-NYCHQ, 7): That’s where, personally, that’s my, like, passion project of, like, how we’re really gonna move the business forward and improve our feedback loop, and our performance for managing our, growth marketing campaigns.
156 00:29:24.410 ⇒ 00:29:32.629 Uttam Kumaran: And then for the, you know, the content descriptions, are you guys using any content management for… for that right now, or where does… where would that, like, live?
157 00:29:32.630 ⇒ 00:29:47.170 Rockaway Peach (United States-NYCHQ, 7): So, content management. So, like, if I’m thinking about… like, we have links, like, in most of our data, like, the links to that creative asset is, like, available, right? Like, if I have a social ad, I link to that creative asset.
158 00:29:47.170 ⇒ 00:29:49.580 Uttam Kumaran: It’s in the Facebook library or something like that. Okay.
159 00:29:49.580 ⇒ 00:30:06.390 Rockaway Peach (United States-NYCHQ, 7): Yeah, and it’s already merged into a data, it’s kind of like… and I came into this in the intern project, which is why I’m raising it. It was like, that became a blocker for a lot of people, was, okay, I can tell you if there’s an offer or not an offer, but I care about more things like what was the hook, or
160 00:30:06.800 ⇒ 00:30:14.899 Rockaway Peach (United States-NYCHQ, 7): how is the claim positioned? So something that, like, is a little bit more sophisticated, because we’re not, like, the product that’s
161 00:30:14.900 ⇒ 00:30:33.280 Rockaway Peach (United States-NYCHQ, 7): new, and we’re trying to find out the basic information about, like, what our consumers want or not. We’re at a more sophisticated stage where the nuance, the, you know, very much, like, trying to fine-tune our brand voice and what resonates, for a product that’s been on the market for, you know, 10 years, or whatever, 7 years.
162 00:30:33.280 ⇒ 00:30:44.959 Uttam Kumaran: Okay, so let me… maybe we’ll prepare a few things. I mean, we have several of those we can share over. I think it’s a… it’s a great project. I mean, this is some of the stuff that we’re pitching to a lot of our clients, as well as… now that we’ve sort of
163 00:30:45.030 ⇒ 00:31:02.790 Uttam Kumaran: built out great data marts. The decision-making is actually… the dashboard is, like, the limiting factor. We want to go one step further, where AI can take the results and take in input, whether… you know, for some of our clients, that’s, like, meeting notes, that’s Slack messages, that’s, like, everything around the context.
164 00:31:02.790 ⇒ 00:31:16.260 Uttam Kumaran: And then AI is like, hey, okay, I can now have enough information to help you either guide towards a decision or research possibilities. So I think it’s a great, you know, move in that direction. Certainly, we’ll send over a bunch of stuff
165 00:31:16.260 ⇒ 00:31:34.650 Uttam Kumaran: that we’ve done on the N8N side. You know, I’ll look through our library of past work and see if there’s anything on the, sort of paid media, but also just, like, complicated workflows that we’ve enabled. And then certainly I can tell you everything about, sort of, the way we work, in terms of security and governance.
166 00:31:35.500 ⇒ 00:31:38.199 Uttam Kumaran: So yeah, we can try to get that over to you this week.
167 00:31:38.690 ⇒ 00:31:54.159 Rockaway Peach (United States-NYCHQ, 7): Okay, nice. Alright, nice, I’m excited to learn more. Yeah. And we’ll keep the conversation going, and if you need any additional guidance from me on info that would be useful, please feel free to reach out.
168 00:31:54.160 ⇒ 00:31:58.380 Uttam Kumaran: Cool, yeah, and if there’s any N8in questions that we could answer in the meantime, please, you can email us.
169 00:31:58.740 ⇒ 00:32:01.880 Uttam Kumaran: Happy to answer those. That’s, like, all we do, so….
170 00:32:01.880 ⇒ 00:32:02.210 Rockaway Peach (United States-NYCHQ, 7): Thanks.
171 00:32:02.210 ⇒ 00:32:02.529 Uttam Kumaran: share with you.
172 00:32:02.530 ⇒ 00:32:17.149 Rockaway Peach (United States-NYCHQ, 7): Okay, NAN experts, what’s your favorite… what’s your, like, most used… I know this is kind of a dumb question, but what’s your most used, like, NATN workflow, or most impactful NAN workflow within your business as a small business owner?
173 00:32:18.210 ⇒ 00:32:19.870 Uttam Kumaran: Yeah, …
174 00:32:20.890 ⇒ 00:32:37.810 Uttam Kumaran: So, one of the things that we do is we actually create, what’s called client hubs for every single client, which actually… we basically use RAG to bring in all of the emails we’ve sent to them, all of the codebase that we’re working with them on, all of the Slack
175 00:32:38.070 ⇒ 00:32:48.799 Uttam Kumaran: like, back and forth, all of the Zoom meetings we’ve had with them, and then at any moment, you can actually go and just chat with the entire history of, like, our engagement with the client.
176 00:32:48.800 ⇒ 00:32:55.800 Uttam Kumaran: So, at the small level, this works to solve things like, can I send an update for what we did this week, right? So our project managers
177 00:32:55.800 ⇒ 00:33:13.560 Uttam Kumaran: aren’t sitting and, like, thinking through or looking through their notes. And… or they’re not even, like, having to copy-paste transcripts into ChatGPT. We’ve sort of solved that. But additionally, like, if you go one step further, my ask for the team is, like, now I want to chat over every meeting we’ve had with any client.
178 00:33:14.460 ⇒ 00:33:28.340 Uttam Kumaran: saying, like, what are common objections that we find in the sales process? Like, what are the most impactful, like, analyses or outcomes that, when we’ve presented, the client had a great reaction? Like, those are some really complex things that
179 00:33:28.740 ⇒ 00:33:39.950 Uttam Kumaran: I don’t know if we would ever have had a chance to do, especially not at our size, where if we have everything segmented by client… and so we’re… again, a lot of our work is communication, so client updates.
180 00:33:40.010 ⇒ 00:33:56.610 Uttam Kumaran: following up on tickets, and so we’ve sort of went step-by-step and are sort of dissecting that. Like, another common use case is, again, the project management workflow, you go to a client, and you have a great meeting, there’s a bunch of action items, and then it’s, like, on that PM to go, like, write that into JIRA,
181 00:33:56.960 ⇒ 00:34:03.379 Uttam Kumaran: And, like, or maybe they just didn’t take notes, or, like, that gets lost, and then… so we built a very simple workflow that…
182 00:34:03.380 ⇒ 00:34:19.310 Uttam Kumaran: takes the transcript, proposes, like, the set of linear tickets to create or modify, and then the PM can literally… we’ve just biccoded a pretty simple UI that they… that the PM can check which tickets they want created. It automatically does the assignments, the estimations.
183 00:34:19.600 ⇒ 00:34:24.200 Uttam Kumaran: So we kind of tackled it as, like, let’s go after sales and marketing first, because…
184 00:34:24.340 ⇒ 00:34:35.530 Uttam Kumaran: I think that was, like, where a lot of the AI went to augment in the beginning, and then we sort of now are moving towards project management, which is a lot of, like, you need the nuance of the client, what the scope of work is.
185 00:34:35.530 ⇒ 00:34:52.479 Uttam Kumaran: And then the last piece I think we’ll go towards is engineering. Everyone on our team uses Cursor and has access to that, but I think engineering, you know, we want to start having AI do all of our PR reviews, we want to start using, like, the Cursor agent to go straight from ticket to, like, the first
186 00:34:52.679 ⇒ 00:34:56.490 Uttam Kumaran: version, and then have the… have an engineer more, like, review that.
187 00:34:56.600 ⇒ 00:34:59.730 Uttam Kumaran: So, like, that’s where we’ll be pushing next. …
188 00:34:59.870 ⇒ 00:35:09.609 Uttam Kumaran: But this, again, that’s all work for us, which is not… we’re not the number one clients here, so we… those are things we work… we work on, like, with our side time, ….
189 00:35:09.660 ⇒ 00:35:12.290 Rockaway Peach (United States-NYCHQ, 7): But we use NNN for all of that, so….
190 00:35:12.290 ⇒ 00:35:22.859 Uttam Kumaran: we use NNN and Supabase for all of our indexing, for all of our vectorization, and then we also use some other helpful tools for RAG. You know, one of the things for us is, I don’t want to go
191 00:35:22.860 ⇒ 00:35:39.469 Uttam Kumaran: like, we may be forced to, but I don’t want to have to become an expert in any of these new technologies, like, read every paper. And so we found some great vendors that offer, like, RAG as a service, where they can handle, like, pre-filtering, tons of document indexing, and so we use them for all of our RAG systems.
192 00:35:39.470 ⇒ 00:35:44.359 Uttam Kumaran: And so that’s been another thing, is, like, as we’ve explored every vendor.
193 00:35:44.470 ⇒ 00:35:58.299 Uttam Kumaran: basically across the entire stack for AI, we found some of the best ones that save us some time, and then we focus on making sure we have the right data, we have the best prompts, we have a great human-loop feedback system, we’re running evals.
194 00:35:58.480 ⇒ 00:36:06.689 Uttam Kumaran: Like, those are the things that we try to do, but yeah, it’s a lot on the project management, client management side that saves some time.
195 00:36:06.690 ⇒ 00:36:16.869 Rockaway Peach (United States-NYCHQ, 7): I’m interested to learn more about that. I mean, I’m in this Pathways program, and, like, one of my projects is, like, I work cross-functionally, and I end up taking on a lot of, like.
196 00:36:16.890 ⇒ 00:36:28.610 Rockaway Peach (United States-NYCHQ, 7): project management-type functions within the team, which, you know, like, I’m the only one who updates the status tracker, and I have to go follow up on every… you know, and so I’d love to, I mean, not now, but, like.
197 00:36:28.980 ⇒ 00:36:35.169 Rockaway Peach (United States-NYCHQ, 7): pick your brain, like, I’ve started using some techniques where, like, I’ll create gems for myself for each project, which helps, like.
198 00:36:35.170 ⇒ 00:36:35.620 Uttam Kumaran: Yeah.
199 00:36:35.620 ⇒ 00:36:50.120 Rockaway Peach (United States-NYCHQ, 7): but, like, I feel like I could… and I’ll, you know, I’m good at, like, Slack AI workflows, but I feel like I could bring it to, like, the next level of, okay, populate my stat, like, here are the notes from this week, like, you know, follow up on things that weren’t tackled, like, something like that.
200 00:36:50.140 ⇒ 00:36:50.810 Uttam Kumaran: Yes.
201 00:36:50.810 ⇒ 00:36:52.290 Rockaway Peach (United States-NYCHQ, 7): That’s what I’d love to get.
202 00:36:52.290 ⇒ 00:37:11.290 Uttam Kumaran: Yeah, and for us, like, all of these started as, like, I had a ChatGPT… we use a chat GPT further, we had a ChatGPT project, where I had a system prompt, and then I was just pasting transcripts in, and as soon as I do that a couple times, I’m like, perfect, this should now move to an agent, because for me, this is something that I want everybody in the company to be able to use, and so we have a little platform website.
203 00:37:11.290 ⇒ 00:37:23.830 Uttam Kumaran: And then we generate these, like, helpful UIs. For example, going from transcript to, like, tickets, it’s hard to see that all in text, and I was like, okay, can we just, like, build something simply where maybe it’s boxes for every ticket.
204 00:37:23.890 ⇒ 00:37:33.999 Uttam Kumaran: it auto-generates the title, who it’s assigned to, and then if it’s wrong, you can go in and make edits, right? And that was a very simple way. Now any PM who joins our team
205 00:37:34.000 ⇒ 00:37:51.380 Uttam Kumaran: you have this thing where you don’t have… now you can sit in a meeting and not worry about all the tickets you have to create, or you could ideally sit in more meetings, or, like, sit more time with the client, you know, and get them more information, and now our PMs can scale maybe not only across one or two clients, but three or four.
206 00:37:51.380 ⇒ 00:37:52.370 Rockaway Peach (United States-NYCHQ, 7): Right? Because….
207 00:37:52.390 ⇒ 00:37:55.869 Uttam Kumaran: What was… they may previously had to spend 10 hours a week
208 00:37:55.980 ⇒ 00:37:58.829 Uttam Kumaran: Creating tickets, and it’s a shame that the…
209 00:37:58.950 ⇒ 00:38:16.759 Uttam Kumaran: the ticket companies are not building this, but I can’t wait for them, right? So, I need this now. And so, those are the kinds of things that, in your workflow, would be like, cool, I’m doing this with gems, now how do we scale something any PM can use? And very simply, paste in your transcript, get tickets out, and then it kind of
210 00:38:16.760 ⇒ 00:38:22.810 Uttam Kumaran: Of course, they immediately start flooding in with feature requests, feature requests, and that gets built up into something larger, you know?
211 00:38:22.960 ⇒ 00:38:24.050 Rockaway Peach (United States-NYCHQ, 7): Yeah.
212 00:38:24.260 ⇒ 00:38:38.409 Uttam Kumaran: But that’s… that’s for our business, like, our constraint is that we… we’re completely bootstrapped, and I want to make sure that we can generate the best outcome for our clients, and sitting in tickets is not something I want my PMs to do at all.
213 00:38:38.440 ⇒ 00:38:46.359 Uttam Kumaran: Right? I want them to worry about what tickets are dropping, and I want them to go maybe have one more meeting with a client to make sure we sharpen something
214 00:38:46.440 ⇒ 00:38:54.680 Uttam Kumaran: Versus, like, every person spending so many… so much time there. It’s still important, because that’s the glue between everyone, that’s everyone’s to-do list, but…
215 00:38:54.930 ⇒ 00:39:02.670 Uttam Kumaran: there has to be smarter ways. So even though we’re doing stuff around ticket grooming, automatic flagging of, like, which tickets are, like.
216 00:39:02.840 ⇒ 00:39:12.179 Uttam Kumaran: about to, you know, fall, or not get… not get used. So we’re, like, dissecting… as far as we can go, we’re trying, you know, so….
217 00:39:12.500 ⇒ 00:39:19.490 Rockaway Peach (United States-NYCHQ, 7): Yeah, awesome. Cool. Do you have any other final questions for me?
218 00:39:20.070 ⇒ 00:39:29.109 Uttam Kumaran: I think that’s it. I guess my question would be, if there’s anyone else you think we should add to that email to send this over to, that would be helpful.
219 00:39:29.250 ⇒ 00:39:29.900 Uttam Kumaran: ….
220 00:39:29.900 ⇒ 00:39:46.509 Rockaway Peach (United States-NYCHQ, 7): Yeah, send it to me, and I’ll circulate. We have a few stakeholders internally. We do have some B2B, folks here, so if you have, like, a use case there, like, I can look them in, too.
221 00:39:46.510 ⇒ 00:39:47.220 Uttam Kumaran: Okay.
222 00:39:48.200 ⇒ 00:39:57.290 Uttam Kumaran: Yeah, we’ll add several of our case studies, several examples of us working with kind of similar companies in sort of digital marketing, growth marketing.
223 00:39:57.290 ⇒ 00:39:57.750 Rockaway Peach (United States-NYCHQ, 7): Awesome.
224 00:39:57.750 ⇒ 00:39:59.509 Uttam Kumaran: And then, yeah, we’ll get that over this week.
225 00:39:59.870 ⇒ 00:40:02.980 Rockaway Peach (United States-NYCHQ, 7): Cool. Thank you so much. It was nice to meet you both.
226 00:40:02.980 ⇒ 00:40:04.750 Uttam Kumaran: Yeah, thank you so much, have a great day.
227 00:40:04.990 ⇒ 00:40:05.670 Rockaway Peach (United States-NYCHQ, 7): Bye.
228 00:40:05.800 ⇒ 00:40:06.460 Uttam Kumaran: Bye.