Meeting Title: Uttam <> Josh Date: 2025-05-13 Meeting participants: Josh Crittenden, Uttam Kumaran
WEBVTT
1 00:00:55.980 ⇒ 00:00:57.110 Uttam Kumaran: Hey, Josh.
2 00:00:57.350 ⇒ 00:00:58.350 Josh Crittenden: Hey udam!
3 00:00:58.850 ⇒ 00:00:59.910 Uttam Kumaran: Hey! How are you?
4 00:00:59.910 ⇒ 00:01:01.319 Josh Crittenden: Good sorry I’m late.
5 00:01:01.720 ⇒ 00:01:09.409 Uttam Kumaran: No, all good. Sorry this background noise. I just. I’ve been working on a lot of sales stuff in the morning, so I try to go for a walk.
6 00:01:09.560 ⇒ 00:01:13.959 Uttam Kumaran: and like take some calls instead of sitting behind the desk. So.
7 00:01:14.540 ⇒ 00:01:21.739 Josh Crittenden: Yeah, no worries at all. I I do the same thing. Whenever I’m on a non customer facing call, I try to try to walk outside when the weather’s nice.
8 00:01:22.490 ⇒ 00:01:27.950 Uttam Kumaran: Yeah, I’m here in Austin, and it’s just like turning over to summer. So
9 00:01:28.100 ⇒ 00:01:36.590 Uttam Kumaran: it will be nice. And then it’s around 9 am here, so it’ll start. It’ll be 100 here at like 9 am. In a in like a month or 2. So.
10 00:01:36.860 ⇒ 00:01:39.319 Josh Crittenden: Oh, yeah, I know where whereabouts in Austin, are you.
11 00:01:39.920 ⇒ 00:01:44.089 Uttam Kumaran: I’m in East Austin. If you’re familiar, like near the Mueller area.
12 00:01:44.260 ⇒ 00:01:55.560 Josh Crittenden: Okay, yeah, I’m I’m somewhat familiar. My my entire in laws, mother in law, father in law, and both my brother in laws are in up in Round Rock, so they’re out in the suburbs.
13 00:01:55.560 ⇒ 00:01:56.760 Uttam Kumaran: Oh, nice. Okay.
14 00:01:57.200 ⇒ 00:02:07.679 Josh Crittenden: Yeah, but we’ll we’ll be. We’ll be in Austin in in August. So I know I know very well if I want to get my run, and I have to. I have to go by 8 am. Otherwise it’s gonna be.
15 00:02:07.840 ⇒ 00:02:09.779 Josh Crittenden: It’s gonna be like hell outside.
16 00:02:11.000 ⇒ 00:02:33.890 Uttam Kumaran: Yeah. And even up there, too, like the roads there with your running. They’re so wide and like people drive so fast here. So I feel like I I like being down here because there’s like still sidewalks, and there’s still like you can. You’re not like walking across like a 10 Lane highway. But I, my girlfriend, is from Cedar Park like right near there, and I go to Round Rock. I have some friends up there, so been there a bunch.
17 00:02:34.890 ⇒ 00:02:35.800 Josh Crittenden: Cool.
18 00:02:36.590 ⇒ 00:02:42.139 Uttam Kumaran: Yeah. So I I mean, it’s it’s great to be in touch like, how do you? How did you originally like meet Steve.
19 00:02:43.080 ⇒ 00:02:52.629 Josh Crittenden: So Steve and I worked for a boutique Microsoft partner called Blue Granite, and Steve was
20 00:02:53.660 ⇒ 00:03:08.329 Josh Crittenden: Steve was there before before I joined, so I joined back in 2,017, and I left in early 2022, and that’s when I came to Snowflake. But Steve had been there, I think, for a few years before I even joined as a
21 00:03:08.530 ⇒ 00:03:13.080 Josh Crittenden: I think he was an a consultant at 1 point, and then he switched into the sales side.
22 00:03:14.200 ⇒ 00:03:14.680 Uttam Kumaran: Nice.
23 00:03:14.680 ⇒ 00:03:18.774 Josh Crittenden: But yeah, we work together for for 5 years.
24 00:03:19.877 ⇒ 00:03:33.700 Josh Crittenden: and it. It was a lot of fun back then, because it it’s kind of like what you said on Linkedin, that you’re looking to build where, when I joined, I think I was probably employee around like we had around 40 to 50 employees at the time.
25 00:03:33.700 ⇒ 00:03:34.490 Uttam Kumaran: Nice.
26 00:03:34.490 ⇒ 00:03:35.559 Josh Crittenden: The Us. Yeah.
27 00:03:35.560 ⇒ 00:03:36.200 Uttam Kumaran: Yeah.
28 00:03:36.530 ⇒ 00:03:41.790 Josh Crittenden: And then by the time I left it was, it was definitely over a hundred. And
29 00:03:42.710 ⇒ 00:03:49.649 Josh Crittenden: yeah, I think, yeah, I think like, 9 months after I left. That’s when blue granite sold to to 3 Cloud.
30 00:03:50.070 ⇒ 00:03:50.430 Uttam Kumaran: Yes.
31 00:03:50.830 ⇒ 00:03:51.380 Josh Crittenden: Yeah.
32 00:03:52.100 ⇒ 00:03:55.380 Josh Crittenden: So I think Steve left
33 00:03:56.010 ⇒ 00:04:04.390 Josh Crittenden: shortly after that acquisition. If I’m not mistaken, maybe a year or so after, and then he’s been bouncing around since. But yeah, we worked together for 5 years at blue granite.
34 00:04:04.690 ⇒ 00:04:06.773 Uttam Kumaran: Nice. Yeah, we got connected.
35 00:04:07.380 ⇒ 00:04:21.919 Uttam Kumaran: where I we’ve been running some local Linkedin campaigns in Austin, mainly just for me to be able to meet people. I mean, I feel lucky. I my career. I started my career in New York. So I know in New York it’s really hard to just like
36 00:04:21.990 ⇒ 00:04:47.810 Uttam Kumaran: meet people in sort of like a non transactional way. So in Austin, I just try to meet people that are in our space. Whether it’s like former employees at at these places or execs or anyone, because I’m my background is isn’t. I’m an engineer. I’m a data engineer. So I’m learning about like sort of all these different, like sales, motions and sales partnerships. And so Steve was great, like we just connected, and you know he’s been nothing but helpful, like just
37 00:04:47.880 ⇒ 00:05:13.910 Uttam Kumaran: being open to to hearing about our journey, and then telling me a lot about how you know partnerships work at like Microsoft and these hyperscalers. And I’ve told him a lot about our business. He actually came to a conference that that we presented at. And and yeah, I mean, he’s he’s like maniacally focused on partnerships to a degree that’s like very inspirational like I’ve never seen. I mean, I’ve only talked to him. I’ve talked to a bunch of people in partnerships. But I’ve never talked to someone that’s like
38 00:05:14.010 ⇒ 00:05:22.999 Uttam Kumaran: keeping track of people and like and doing it in a very systematic way, which for me, as a data person like it’s like, Okay, I’m like, that’s how I’m thinking about it. So.
39 00:05:23.000 ⇒ 00:05:23.680 Josh Crittenden: Yeah.
40 00:05:24.570 ⇒ 00:05:30.849 Josh Crittenden: yeah, that’s awesome. Steve’s a great guy. I know he’s I think this is his 3rd stint since he left
41 00:05:31.110 ⇒ 00:05:32.880 Josh Crittenden: 3 Cloud. But
42 00:05:34.270 ⇒ 00:05:42.969 Josh Crittenden: yeah, I even tried to get him in on the vendor side here at snowflakes. I know he was interested in Snowflake and Databricks at one time, but I think he likes being on the services side.
43 00:05:43.800 ⇒ 00:05:50.919 Uttam Kumaran: Yeah, I think so, too. I think he likes the relationships and things like that. But yeah, so I kind of mentioned a little bit. But
44 00:05:51.150 ⇒ 00:06:17.340 Uttam Kumaran: you know, Brainforge is a company. I started about a year and a half ago. We’re about 15 people now. My background again is in data engineering. But I’ve worked on Snowflake since 2018, and I’ve probably implemented Snowflake like 30 or 40 times now. So we’ve done a lot of snowflake business just in my personal career. And then that’s part of brain Forge as well. So that’s really was our sort of the foot in the door, and a lot of things is our expertise on Snowflake. And then
45 00:06:17.430 ⇒ 00:06:34.690 Uttam Kumaran: beyond that, we we started doing a lot of work across the entire data stack. So basically, like we? We come in as like a fully as like a forward deployed data team. Everything from data Eng to data, warehousing to modeling and Dbt to then
46 00:06:34.690 ⇒ 00:06:54.110 Uttam Kumaran: standing up data, you know, data analytics like Bi, and then of course, we’re also doing insights and kind of the difference for us is we’re not. We don’t do any sort of staff on most of the stuff that we do. Is all where we own the relationship, where we come in at the top sort of identify that there is a problem. And that problem can be solved
47 00:06:54.270 ⇒ 00:07:16.813 Uttam Kumaran: well by analytics, if not partly. And then we work with the executives sort of get their buy in, and then work to deploy everything. So we’ve deployed Snowflake for tons of clients. Very, very familiar with the product. But yeah, that’s sort of like our background and part of like what I’m spending a lot of my time thinking about is sort of partnerships. And how do? How do we sort of
48 00:07:17.190 ⇒ 00:07:35.639 Uttam Kumaran: you know, how do we be a win win for the companies that we implement? Right? So we. We tend to manage a lot of procurement decisions for a lot of these firms. I mean, these are mid market to large firms where they may have already selected a series of software software ecosystem
49 00:07:35.899 ⇒ 00:08:00.050 Uttam Kumaran: or they may. There may just be like looking or shopping, or they have some legacy thing that some someone inherited. And so we always come in, and they always want us to take a look at like, Hey, do we have the right tools? We when we do everything, we review everything from cost to licenses? Looking at like, hey, do these tool play tools play nice with each other? Hey? Are you still stuck on like stitch, and you need to move to something like way better than that.
50 00:08:00.350 ⇒ 00:08:08.369 Uttam Kumaran: things like that. And so Snowflake is really like always top of mind for us. And I honestly don’t like everybody. A lot of people
51 00:08:09.030 ⇒ 00:08:19.670 Uttam Kumaran: tell me like, oh, so if it’s expensive, I’m like, it’s not that expensive, for, like what the work it does and so I don’t know. That’s that’s a little bit about us. I mean, I’m happy to share more. If you have any questions.
52 00:08:20.070 ⇒ 00:08:22.680 Josh Crittenden: No, I think I think all that makes sense.
53 00:08:23.180 ⇒ 00:08:28.021 Josh Crittenden: one thing I wanted to ask you, cause I I did spend some time on your website was,
54 00:08:28.460 ⇒ 00:08:35.010 Josh Crittenden: are you all involved in the Snowflake partner network, currently? Or are you trying to get engaged there.
55 00:08:35.500 ⇒ 00:08:50.039 Uttam Kumaran: Yeah, so we are involved there. We haven’t done like a lot of our deals. Either we inherited snowflake or moving clients into annual contracts. But we have. We have been on the partner network, and I wasn’t on. I was part of like the partner network for even the last
56 00:08:50.481 ⇒ 00:09:03.889 Uttam Kumaran: company that I was at. So. Yeah, I’m very familiar with that. We haven’t really grown a lot there. But that’s something that I’m I’m really curious about how do we get on the radar and and like, how do we
57 00:09:04.020 ⇒ 00:09:06.029 Uttam Kumaran: sort of continue to grow with y’all?
58 00:09:07.150 ⇒ 00:09:18.099 Uttam Kumaran: Okay, we are. We are on there signed up. I don’t know. I haven’t talked to anybody on your side about it, or I don’t. I don’t. I think I tried to register one or 2 deals, but I sort of didn’t go. The distance really.
59 00:09:18.810 ⇒ 00:09:41.299 Josh Crittenden: Yeah, I think the so. Where where I’m going with this is it will be a tough sell internally to snowflake account team. So obviously, I’m I’m on the Se. Side. I actually synced with some of my, my Texas colleagues internally to see if they’ve met with you at any of the user group groups. It sounds like you may already be synced with Aaron.
60 00:09:41.300 ⇒ 00:09:50.009 Uttam Kumaran: Yeah. So yeah, I go to the Austin user group a bunch. And I think now that we’re doing more, I’ll probably try to hit the Dallas one. But yeah, I’ve seen for Aaron.
61 00:09:50.370 ⇒ 00:09:52.150 Josh Crittenden: Yeah, awesome. So just
62 00:09:52.468 ⇒ 00:10:02.990 Josh Crittenden: just a little bit about me. So I’m on the Enterprise expansion team. I’m on the Ohio Valley team. So what that means is, we’re kind of in between commercial and vertical segments.
63 00:10:03.691 ⇒ 00:10:13.330 Josh Crittenden: But I still work with primarily like fortune. 500 customers so I have. I have big, large 100 year old enterprise
64 00:10:13.789 ⇒ 00:10:37.539 Josh Crittenden: customers, and then we we have all the way down to that smaller and midsize segment as well that that you described and then I so I know you know Aaron. Chris Hancoop is another peer of mine. He I believe he’s in Dallas, and then, my, all of our Se. Director, Josh Chakona, is actually in Austin. I believe he’s on the.
65 00:10:37.540 ⇒ 00:10:38.150 Uttam Kumaran: Right.
66 00:10:38.560 ⇒ 00:10:41.710 Josh Crittenden: The west side of town. I don’t know if you’ve met Josh Chacona before.
67 00:10:43.450 ⇒ 00:10:44.520 Uttam Kumaran: No, not yet.
68 00:10:44.790 ⇒ 00:10:46.829 Josh Crittenden: Okay, so what? Where? I’m going with this.
69 00:10:46.830 ⇒ 00:10:49.600 Uttam Kumaran: Unless he, unless he’s come to the Austin user group. Yeah.
70 00:10:50.040 ⇒ 00:11:02.149 Josh Crittenden: Yeah, I don’t. I? I don’t think so. But where? Where I’m going with this is, I think, that the logical step forward for you all is to at least achieve that that minimum partner status which would be our select here.
71 00:11:02.150 ⇒ 00:11:04.140 Josh Crittenden: Oh, but it’s like, I don’t
72 00:11:04.140 ⇒ 00:11:08.190 Josh Crittenden: know the requirements off the top of my head. But I believe you’ll have to have like.
73 00:11:08.190 ⇒ 00:11:08.869 Uttam Kumaran: So, maybe.
74 00:11:08.870 ⇒ 00:11:12.329 Josh Crittenden: 10 consultants certified in the snow pro core.
75 00:11:12.330 ⇒ 00:11:12.820 Uttam Kumaran: There’s like.
76 00:11:13.064 ⇒ 00:11:22.369 Josh Crittenden: And then there’s like a there’s some kind of annual fee. I’m looking at an old document, but it’s 4 years old, so I don’t wanna quote anything and and misstated, but there’s gonna be some kind of.
77 00:11:22.370 ⇒ 00:11:24.449 Uttam Kumaran: Think I think I have it. Yeah. So
78 00:11:25.340 ⇒ 00:11:27.949 Uttam Kumaran: so mainly just hit. Try to hit the select.
79 00:11:28.240 ⇒ 00:11:57.049 Josh Crittenden: At a minimum hit the select. Because then what you can do is you can then go into these user groups. You can sync with account teams, if you wanna like, be more regional in in the Texas area in the Southwest, or if you want to branch out to the Midwest, or either of the coast, and then you can at least dangle that carrot to snowflake account teams and say, Hey, we’re we’re a snowflake, partner. We’re a boutique partner right now, but we want to grow into becoming one of those premier and elite partners over time.
80 00:11:57.050 ⇒ 00:12:05.879 Josh Crittenden: and and then obviously working for a boutique part as a boutique partner. You have the advantage of saying like, we’re not. We’re not staff Aug. We’re gonna
81 00:12:05.880 ⇒ 00:12:23.980 Josh Crittenden: we’re gonna come in and truly establish rapport with customers and build that relationship and handle everything from A to Z. And I think that’s gonna to me I would be more eager and talking to partners your size than I would be talking to the Ph. Datas and Hakotas of the world, or or even the
82 00:12:24.370 ⇒ 00:12:36.149 Josh Crittenden: like, Deloitte and accenture, because obviously your goal is probably to to grow someday. And you know, maybe maybe you’re open to to selling someday, or you want to grow and and.
83 00:12:36.850 ⇒ 00:12:39.380 Josh Crittenden: And and become one of those elite partners. But
84 00:12:39.650 ⇒ 00:12:51.500 Josh Crittenden: I I always have had greater success working with boutique partners because they’re not. They’re not a staff fog. They they know where they excel at. They’re not afraid to turn down work.
85 00:12:51.932 ⇒ 00:12:59.499 Josh Crittenden: If it’s not in their wheelhouse, whereas, you know, when I work with Deloitte they’ll they’ll take anything and everything. And then, when the project.
86 00:12:59.500 ⇒ 00:13:00.730 Uttam Kumaran: Yeah, yeah.
87 00:13:00.730 ⇒ 00:13:03.200 Josh Crittenden: True. It’s like, well, why’d you guys even take the sound.
88 00:13:05.073 ⇒ 00:13:26.500 Uttam Kumaran: Yeah, I guess. Can you like, can you maybe double down on that, too? Because I’m trying to think about, you know, part of it is like, yes, exactly. You know, I’m talking to some. We even partner with some other larger development partners, where by the time they even get to talking about the project and get a proposal over. It’s like 4 weeks. I’m like, we’ll turn something around and like
89 00:13:26.570 ⇒ 00:13:47.939 Uttam Kumaran: one week, you know, with the proof of concept and something. And and that’s where we excel. Because I can have. I can just increase amount of surface area. But you’re right in that. What I. What I gain in that I don’t have in sort of like all the development workforce, like hundreds of people. But what we’re doing is sort of partnering with them, saying, like, Hey, we’ll handle
90 00:13:48.130 ⇒ 00:14:00.110 Uttam Kumaran: bringing those deals in if they end up being large, and we’ll work with them. And so we’re we’re signing some of those partnerships as well in not only in data analytics, but also in AI. So one thing I kind of failed to mention is, we’re also doing a lot of work.
91 00:14:01.058 ⇒ 00:14:04.600 Uttam Kumaran: Building AI workflows, agentic workflows
92 00:14:05.050 ⇒ 00:14:10.010 Uttam Kumaran: specifically larger part of our business. I know
93 00:14:11.646 ⇒ 00:14:25.099 Uttam Kumaran: we model everything, and we make available agents available to our lens.
94 00:14:25.390 ⇒ 00:14:41.869 Uttam Kumaran: But you’re right now, even or so. That’s where we.
95 00:14:42.020 ⇒ 00:14:45.029 Uttam Kumaran: That’s what we we can do. You know, that’s the pay solution.
96 00:14:46.580 ⇒ 00:14:49.170 Josh Crittenden: Yeah, yeah, 100%. I I think,
97 00:14:49.750 ⇒ 00:14:54.615 Josh Crittenden: I think that makes sense. And so if you were, if you were talking to oh.
98 00:15:08.310 ⇒ 00:15:09.489 Josh Crittenden: hey, you back.
99 00:15:10.380 ⇒ 00:15:12.179 Uttam Kumaran: Hey? Yes, I’m back. Sorry just just a minute.
100 00:15:12.180 ⇒ 00:15:13.477 Josh Crittenden: Yeah, no worries.
101 00:15:14.110 ⇒ 00:15:28.079 Josh Crittenden: I think all that makes sense. And if you were to talk to, you know, let’s say, my, my district the Ohio Valley team. So it’s we. We cover Michigan, Indiana, Ohio, and then Western PA. So pretty much Pittsburgh
102 00:15:28.830 ⇒ 00:15:45.730 Josh Crittenden: And and this goes even for, like the Texas team or some of the other Midwest teams in my expansion territory. We’re all working with existing customers. So where I’m going with this is a lot of our customers already have their foundation in place with Snowflake. They’ve already adopted
103 00:15:45.770 ⇒ 00:16:00.089 Josh Crittenden: Sigma power tableau. And so the conversations that we’re having is primarily around Gen. AI, and even even to a lesser extent, traditional machine learning, like helping with inventory forecasting use cases.
104 00:16:00.090 ⇒ 00:16:00.470 Uttam Kumaran: Yes.
105 00:16:00.470 ⇒ 00:16:08.999 Josh Crittenden: Not so that that’s gonna resonate more with, I think, my team. And then even the verticals teams where we’re working with established customers.
106 00:16:09.140 ⇒ 00:16:09.680 Uttam Kumaran: Yes.
107 00:16:09.680 ⇒ 00:16:16.389 Josh Crittenden: I think, what you were telling me at the beginning is gonna resonate more with the acquisition teams where you can partner with them to help
108 00:16:16.933 ⇒ 00:16:27.430 Josh Crittenden: bring Snowflake to prospects and and help them help lead the prospect to water to eventually becoming a snowflake customer and then building building that foundation.
109 00:16:27.866 ⇒ 00:16:49.820 Josh Crittenden: But yeah, for for my team, we’re all about Jenny I because we’re obviously competing with Microsoft with databricks with Palantir. And we we all have very similar gen AI capabilities. We all host Llms in a very similar manner. We all have co-pilot experiences. We all have. Agentic AI experiences.
110 00:16:49.820 ⇒ 00:17:01.420 Josh Crittenden: But that’s that’s what’s gonna ultimately help us build our moat and our customers and and keep our competitors at at arm’s length. And so one other suggestion I had for you is
111 00:17:01.810 ⇒ 00:17:09.199 Josh Crittenden: build, build repeatable accelerators. If you’re not already doing this where that way you can come in and you can say, hey?
112 00:17:09.210 ⇒ 00:17:31.829 Josh Crittenden: You know, potential customer, we’ve we’ve already done this 5 times, and we can align a customer reference, or, if you want but, like, here’s how we go from ideation to production in the next 4 to 12 weeks. These are the things that we expect we’re going to run into. So the this is where we’re going to need your help, but trust our engineering team to, you know, to actually bring this into production over the next 2 to 3 months. Here.
113 00:17:32.590 ⇒ 00:17:49.479 Uttam Kumaran: Yeah, that makes sense. I mean, I think we we do. I mean, that’s basically what we’re pitching is you know, the fact that we’ve done this so many times in 4 week span we’ll send up Snowflake. Stand up, Etl, have data landed and already have something for you to look at in like some manner right? And like, that’s faster.
114 00:17:49.570 ⇒ 00:18:11.979 Uttam Kumaran: You, you know again, if we’re competing with bringing on someone full time like it’s certainly faster than that. And then at that point, we’re competing just with, you know, other analytics firms. And then we can compete on price and and a couple of other things. So that’s what I’m thinking. Can you talk a little bit about for the Gen. AI capabilities? What in particular? Because you guys have released a lot. You know everything from
115 00:18:12.240 ⇒ 00:18:27.759 Uttam Kumaran: like having it right in sitting queries to the co-pilot experience to hosting Llms like, what in particular are you seeing? Are you guys wanting to push? And then what is actually getting adoption? If you can like, share a little bit about that.
116 00:18:28.060 ⇒ 00:18:42.770 Josh Crittenden: Yeah, I can. Even I can even email you a quick 15 min demo that I recorded for our customers. But long, long story short, like, if if you were at your desk, I would I would share, like the high level cortex slides. But
117 00:18:43.320 ⇒ 00:19:05.709 Josh Crittenden: Essentially, cortex is our umbrella term for all of our capabilities. And so within cortex. What what customers should be aware of is that this is fully serverless, where Snowflake is hosting all of the frontier Llms. Natively within the serverless layer on their behalf. So there’s no infrastructure they have to manage.
118 00:19:06.363 ⇒ 00:19:15.649 Josh Crittenden: There’s your your prompt. Your data is never be leaving the walls of their snowflake account. So it’s never being sent to Openai, to Anthropic, to Meta.
119 00:19:15.650 ⇒ 00:19:16.130 Uttam Kumaran: Query.
120 00:19:16.519 ⇒ 00:19:34.140 Josh Crittenden: And so everything’s fully secure, and it’s it’s fully serverless. And so a good example is. Just yesterday we announced support for Claude, sonnet 3.7 we have Openai models coming to cortex. I’m assuming at this point. We’re just waiting for our Snowflake summit next month.
121 00:19:34.200 ⇒ 00:19:35.230 Uttam Kumaran: Okay.
122 00:19:35.230 ⇒ 00:19:45.869 Josh Crittenden: We have the the llama suite of models. And so a good example was when when Llama, or when Meta tried to crash the Google next party, and they announced, like Llama 4.
123 00:19:45.870 ⇒ 00:19:46.190 Uttam Kumaran: Yes.
124 00:19:46.190 ⇒ 00:19:54.150 Josh Crittenden: On a Saturday that that very next day on Sunday, we already had those models in public preview within cortex for customers to use.
125 00:19:54.150 ⇒ 00:19:54.570 Uttam Kumaran: Nice.
126 00:19:54.899 ⇒ 00:20:06.119 Josh Crittenden: So we we are always at the bleeding edge and looking to add in the latest and greatest models for our customers to use. And then, once those models are in cortex, we.
127 00:20:06.230 ⇒ 00:20:22.959 Josh Crittenden: we use them to power. A number of Jenny I experiences. So, for example, we have, we have like task specific functions that will do very specific things like, hey? I have text here in Spanish or Italian. I want to translate it to to English or a different language.
128 00:20:22.960 ⇒ 00:20:23.520 Uttam Kumaran: Hmm.
129 00:20:23.650 ⇒ 00:20:40.720 Josh Crittenden: I have a big block of text. I want to summarize it. So we have a translate function. We have a summarize function. We have a sentiment function for basic sentiment scoring. So those are some of the building blocks. And then we have our complete function which can be used for traditional, prompt engineering. You basically
130 00:20:40.720 ⇒ 00:20:58.950 Josh Crittenden: pass in what model you want to use. And then you pass in your prompt and so you could just keep it generic and ask it questions as if it were, you know, Chat Gpt, or you can actually pass in a column from a database table and have it do some kind of work on on your actual organization’s data.
131 00:20:58.990 ⇒ 00:21:10.690 Josh Crittenden: So that’s that’s the foundation. And then when it when we move up the you know the pyramid, if you will. That, that middle layer is our agentic AI, where
132 00:21:10.890 ⇒ 00:21:26.440 Josh Crittenden: Agentic AI is obviously the buzzword of the year. But what I like to tell customers is like it’s an actual, tangible thing in Snowflake. You are literally building agents, and all snowflake agents are or cortex agents is essentially a
133 00:21:26.580 ⇒ 00:21:29.779 Josh Crittenden: an orchestrator that will take someone’s
134 00:21:30.183 ⇒ 00:21:52.470 Josh Crittenden: prompt or request or question, and it will route it down to one of 2 underlying services. And so when you have structured data. That’s where you would use cortex analyst, where basically, you’re building a semantic model to improve the accuracy and precision of the responses that the Llms generate. But basically, it’s all about, hey? I have structured data in in my snowflake account. How can I?
135 00:21:52.852 ⇒ 00:22:15.799 Josh Crittenden: Ask questions about it? You know what are, what are sales by brand over the the last 4 years? What are you know? The demo? I’ll show you will actually focus on fictitious IoT like plant telemetry data. And so what are average temperature readings in this part of this plant on a specific date? So that’s that’s all for structured data. And then, for traditional.
136 00:22:15.830 ⇒ 00:22:26.579 Josh Crittenden: you know, retrieval, augmented generation. And you have unstructured data. And you want to ask against that. That’s where cortex search comes into play. And then in the demo you’re going to see, we have a number of
137 00:22:26.730 ⇒ 00:22:30.239 Josh Crittenden: unstructured processing capabilities like the ability to.
138 00:22:30.240 ⇒ 00:22:30.680 Uttam Kumaran: Mistaken.
139 00:22:30.680 ⇒ 00:22:36.549 Josh Crittenden: Parse entire documents, the ability to to chunk that text. And then cortex search will actually handle
140 00:22:37.420 ⇒ 00:22:39.300 Josh Crittenden: the vector embedding embeddings.
141 00:22:39.300 ⇒ 00:22:45.220 Uttam Kumaran: Yeah, and then for for the
142 00:22:45.970 ⇒ 00:22:59.609 Uttam Kumaran: yeah. So for the for the document, structures like, are these like pulling from S, 3, or are these? Are we uploading this directly to Snowflake? Like, how are you guys pulling the the actual like flat, like files like this? Pdfs or images, or.
143 00:22:59.820 ⇒ 00:23:09.690 Josh Crittenden: Yeah, great question. It could be both. So the the other thing I always tell customers is snowflake is is just a managed service. So when you load data in the snowflake.
144 00:23:09.690 ⇒ 00:23:10.450 Uttam Kumaran: There’s like.
145 00:23:10.450 ⇒ 00:23:23.539 Josh Crittenden: If you’re snowflake on aws, we’re storing your data in in Snowflake, managed S. 3. If you’re snowflake on azure, it’s just Snowflake managed as your blob. So what that means is you can either you’re gonna load it into a stage, is the short answer.
146 00:23:23.540 ⇒ 00:23:24.190 Josh Crittenden: Cool.
147 00:23:24.190 ⇒ 00:23:26.630 Josh Crittenden: It can either be an internal stage, or it can be
148 00:23:27.260 ⇒ 00:23:29.040 Josh Crittenden: internal. S. 3, or as a blob.
149 00:23:29.040 ⇒ 00:23:54.183 Uttam Kumaran: Great. And then so basically, you know, this is really helpful to hear those, the products actually explain that way. So I sort of only seen it piecemeal, like feature by feature. But I think hearing as in that manner, is actually really helpful. I. So basically what comes top to mind is that anyone we implement. So for we should just have, we should just already set up the cortex analyst immediately. Basically
150 00:23:54.570 ⇒ 00:24:16.000 Uttam Kumaran: right? Because and then I think, what what will be helpful for me to understand is what what should we provide at the semantic layer in terms of table descriptions or column descriptions, or whatever, and and basically handling that. But I mean for every customer that we have on so like, I should just turn this on and my, I guess my my question is, can is that available as an Api like if they want to do that
151 00:24:16.110 ⇒ 00:24:26.819 Uttam Kumaran: within slack, for example? Or if they want to handle that in another ui like can is that, can that be made available external to the snowflake platform?
152 00:24:27.100 ⇒ 00:24:49.049 Josh Crittenden: 100%. Yeah. So I’ll I’ll actually, I’ll literally just copy and paste my my email campaign that I had sent to several customers and I’ll send that to you that way you can. You can see that because what what I always tell customers is that Ui is kind of a choose your own adventure. If if you want to use slack, if you want to use teams use that if you want to use
153 00:24:49.150 ⇒ 00:24:59.370 Josh Crittenden: Microsoft co-pilot or some homegrown solution, use that, and then from a native from a native snowflake capabilities. We also obviously have have Streamlin snowflake, which is great.
154 00:24:59.370 ⇒ 00:24:59.870 Uttam Kumaran: Yes.
155 00:24:59.870 ⇒ 00:25:17.610 Josh Crittenden: Typing. But then we also have, and I don’t know if you’re aware of this, but we also have Snowflake intelligence that we’re announcing at Summit in a few weeks. And so Snowflake intelligence is, think, think a chat Gpt interface. But for for organizations using Snowflake.
156 00:25:17.960 ⇒ 00:25:24.349 Uttam Kumaran: Nice. I mean, I saw the right sidebar experience. And I was like, okay, this hopefully, this experience is something bigger.
157 00:25:24.600 ⇒ 00:25:49.799 Josh Crittenden: Yeah. So it’s already like the right now in private Preview. It’s it’s part of snowsight. But we’re already decoupling this from Snow site where it’s going to be ai.snowflake.com, and then you can still sso into that URL. Using your credentials. So it’s going to be fully integrated with your snowflake account, but it will be completely decoupled from snowsight.
158 00:25:50.840 ⇒ 00:26:05.580 Uttam Kumaran: And then can you explain the pricing on cortex, by the way, so like for the AI analyst and then for for the new, you know, chatgpt style, experience like, what’s the pricing like? And for me, what? What that’s helpful to explain to clients is like.
159 00:26:05.620 ⇒ 00:26:28.920 Uttam Kumaran: I will then look at the alternatives in market. Right? There’s a lot of like text to sequel type, competing pro competing products. There’s a lot of bi tools that are adding this. I am not interested in doing this in a bi tool. I would rather do this in the warehouse. Entirely right. So I think I have a good pitch on the product side. I I know what to say like, but can you give me a sense of of how the pricing works.
160 00:26:29.130 ⇒ 00:26:41.320 Josh Crittenden: Yeah. So court cortex analyst is. And and here’s here’s the thing like snowflake is never going to be the cheapest, hard, the cheapest option when it comes to hard costs on the market. But what you can say is.
161 00:26:41.750 ⇒ 00:27:09.909 Josh Crittenden: Hey, bi tool over here is going to be cheaper from a pure, hard cost perspective. But this is these are the soft costs that you’re saving like you don’t. You don’t need 5 engineers to build this and maintain this. You just need one person to to update this from time to time. So from all of all of our Jenny capabilities are just token based pricing just like any any other platform. And then those tokens get translated into snowflake credits. Okay?
162 00:27:10.260 ⇒ 00:27:24.909 Josh Crittenden: And so when you are like, if you’re using the complete function. If you’re using our summarize function like, that’s all. I’ll link this in the email to you, too. But that’s in our credit consumption table. And you can see the number of snowflake credits per 1 million tokens.
163 00:27:24.910 ⇒ 00:27:25.370 Uttam Kumaran: Okay.
164 00:27:25.370 ⇒ 00:27:46.769 Josh Crittenden: When it comes to like our agentic AI experiences cortex analyst has right now a current cost of 67 credits per 1,000 messages. And and then you know what you need to do is work with the customers to understand what their current snowflake discount is, and and then arrive at their price per credit.
165 00:27:46.840 ⇒ 00:27:47.430 Uttam Kumaran: And then.
166 00:27:47.430 ⇒ 00:28:10.929 Josh Crittenden: You can do the math there. We, when you work with snowflake account teams us have calculators at our disposal where we can. We can actually calculate, you know, based on their anticipated usage, their anticipated growth patterns like, you know, if this Chatbot interface has a hundred people today, but maybe 10,000 people a year from now we can. We can bake that in.
167 00:28:10.930 ⇒ 00:28:11.330 Uttam Kumaran: Yeah.
168 00:28:11.330 ⇒ 00:28:17.289 Josh Crittenden: And then the other thing that I’ve recently learned is as these chat interface use cases.
169 00:28:17.290 ⇒ 00:28:17.630 Uttam Kumaran: Why?
170 00:28:18.125 ⇒ 00:28:18.620 Josh Crittenden: Expand.
171 00:28:18.620 ⇒ 00:28:19.370 Uttam Kumaran: With that.
172 00:28:19.370 ⇒ 00:28:27.149 Josh Crittenden: The Snowflake account team can actually work with our deal desk internally to to basically try to unlock additional discounts and rebates.
173 00:28:27.150 ⇒ 00:28:27.810 Uttam Kumaran: Nice.
174 00:28:27.810 ⇒ 00:28:33.870 Josh Crittenden: So, for example, maybe maybe our retail price is 67 credits per 1,000 messages. But if they have.
175 00:28:34.000 ⇒ 00:28:44.199 Josh Crittenden: you know, if they’re, I’m just using round numbers. If they have 100,000 messages per month, maybe we can lower that, you know, to a to a much more attractive price for them.
176 00:28:44.200 ⇒ 00:28:53.019 Uttam Kumaran: Okay, no, this makes sense. I mean, look for we. We deal with a lot of unstructured data, use cases like documents, images. I think.
177 00:28:53.210 ⇒ 00:28:54.710 Uttam Kumaran: part of those
178 00:28:55.080 ⇒ 00:29:11.570 Uttam Kumaran: you know there, there isn’t a structured component, and so it makes it hard to pitch snowflake as the best route for that, because we could just use S. 3. I think this is where I’m seeing where you guys marry is like, if we have both right for clients that have both, and they wanna make they want to give both.
179 00:29:11.850 ⇒ 00:29:28.180 Uttam Kumaran: They want to shovel both into an Llm. And they don’t want to have like sort of they don’t want. They have tighter security constraints where they don’t want to use an external service. This is perfect right? Because we’re shoveling not only documents and like things from rag, but also structured data from like Hubspot or things like that, to to basically
180 00:29:28.590 ⇒ 00:29:54.249 Uttam Kumaran: just slap an Llm on top of like the right context. So this makes a lot of sense. And again, for me, I think the biggest thing I’ll go learn is, you know, we don’t typically like maintain a ton of semantic documentation unless it’s like a really big client, because just because it takes time and nobody ends up reading that stuff. But now that it there’s actually a value to it, I think that’s where we will. We’ll actually work a little bit more on making sure that there are there stable metadata call metadata.
181 00:29:54.290 ⇒ 00:30:08.230 Uttam Kumaran: And yeah, I’m curious to read more. I’ve been following along at a high level, and I followed a lot when it when the stuff just came out, but I knew that it would take maybe 6 months to get like to start to get everything right. But that’s really really exciting. Thanks.
182 00:30:08.230 ⇒ 00:30:32.999 Josh Crittenden: Yep, yeah, 100%. So I apologize, I have to drop for another call. But I will. I’ll send you an email here shortly with with the demo and all of the content that I’ve sent customers, and then I’ll I’ll link to our credit consumption table, too, for all things cortex. And and you can you can see, like what how the tokens get translated into credits. And then for you’ll be able to see it by feature, by feature.
183 00:30:33.000 ⇒ 00:30:38.710 Uttam Kumaran: Okay, okay, perfect. Thank you so much, Josh, I really appreciate it. Thanks for taking the time and for all the info.
184 00:30:38.950 ⇒ 00:30:40.789 Josh Crittenden: Yeah, absolutely. Alright. We’ll talk soon.
185 00:30:41.080 ⇒ 00:30:42.200 Uttam Kumaran: Thank you. Bye.