00:00There's something I want to take up with you, Sridhar, and you and I have discussed this before.
00:04You have Project Snowwork, and the idea is that you have this big customer base that goes from
00:09running agents that give you answers to running workloads. And I go back to,
00:16is an agent that only gives answers actually an agent? Is it a real piece of agentic AI that
00:22acts with autonomy? It seems like now we're getting to that part.
00:26Yeah, I mean, agent is a loose term for what it's for.
00:29It's all about, it's fundamentally about AI models using tools and learning from how they
00:35are using it. And for example, our first generation of agents was Snowflake Intelligence, which gave
00:41analytic insights into data. And then CortexCode made acting on Snowflake a lot easier. Snowwork
00:48is a natural culmination. What it's able to do is take the hundred tabs that you have open on your
00:53computer and create a single environment that brings together things like that. So not only
00:59can you get information about something you want, how is a customer doing, but you can also take an
01:03action, send them an email or make a recommendation for them. It's that power that's increasing by the
01:09day as AI models get stronger and stronger.
01:11The idea, you know, every time you go to an event like HumanX, you're trying to say,
01:15what is the thing that people are excited about or they're talking about?
01:18And one of the concepts is you go to bed and you leave your computer to do work overnight and
01:23you
01:23wake up to tangible results. Is that the idea with this?
01:28That can be. It's one of the consequences. We've always had scheduled things like scheduled reports.
01:32I'm sure you get something in the morning that's like a summary of all of the things that you should
01:36know. You can think of agents.
01:38But that summary hasn't taken an action. It hasn't performed a piece of work on my behalf.
01:43That's right. That's right. But with these agents, absolutely. You can. I have people that are
01:47writing code 24-7. They just like set the agent on the task of writing new code to get some
01:53feature
01:53done and they give it some instructions, come back, check in the morning. I think absolutely.
01:58These agents are getting better and better at taking actions on your behalf. But part of the magic
02:04is how do you do that in a governed way? How do you have like a control plane that says
02:09these kinds
02:09of actions are fine, but those have large outside world consequence. They really need to be subject
02:14to human scrutiny. Increasingly, that's going to be what is needed to make to make AI succeed at scale.
02:21Outside world consequences. You are the man to speak to today about, well, the furor, the concern,
02:28the anxiety of the latest anthropic model that is being stress tested by 40 tech companies because
02:34it's so powerful. It can't get into the hands of us mere mortals. Shrita, what do you make about the
02:38step changes that we're seeing in models right now and how powerful they're becoming?
02:42I think the impact that they're having on software is profound. Software is getting easier to create.
02:50And this also means that people can create value faster. I think Anthropic is doing an incredibly
02:56responsible thing by taking a model like that, which has breakthrough capabilities, especially in
03:03code generation, and giving it to a set of their closest partners to ensure that when it comes to
03:09security, for example, we are able to plug gaps. Security has always been a difficult, hand-done,
03:14human-led kind of motion. And I actually think of this as a big advance where we can systematically
03:20make all of our systems that are so critical for our nation, for our own survival, a lot more
03:26stronger. I actually see this as a positive outcome.
03:28You were at the cutting edge of models, of course. You were the co-founder of Neva. You came on
03:31to
03:32Snowflake and you worked for a long time at Google. I'm interested actually in how you're finding the
03:36narrative around Snowflake right now. We actually had the top of the show, Brooke Dwayne on from
03:40Goldman Sachs Asset Management. They love Snowflake. And in particular, they're liking and wanting to see
03:45from application names and from software companies, well, more cost optimization, more capital return.
03:51Is that something you're having to think about when you've got some of your coders working 24 hours
03:56because they're able to just plug and play while they sleep? How is that affecting your costs?
04:02Well, I think the return that we get from investing in AI comes back to us many, many, many fold.
04:08Used to be last year, we had maybe a dozen demos that we would show our customers. They were all
04:13kind
04:13of cookie cutter. But today, I and my sales team is a lot more capable, can generate a demo for
04:19you
04:20in 10 minutes if I want. But and that is a lot more personal. It's much more relevant to your
04:24business. AI is making a real difference in how we show people what is possible with Snowflake.
04:31And we are getting huge returns in areas like support or even our what we call our SRE team.
04:36There's a team that keeps Snowflake, the system up. They're using the power of agentic AI to be
04:41a lot more effective. AI has cost, but the returns we get from it, problems that used to take multiple
04:47days to debug and get to the bottom, we can get answers now in 10 to 15 minutes. There is
04:53cost,
04:54but I think the returns are more than worth it, definitely for us and for a lot of our customers
04:59that are able to do things that honestly were just inconceivable just like a year ago where
05:05companies like United Rentals basically just released an app to their 1,600 branches. And
05:10whatever questions these branches have about the current state of the business, they can just get
05:15it interactively. That is game changing for them. That's Bloomberg Intelligence, our in-house
05:18research arms thesis, that as company use of agents grows, you will benefit. Just very quickly,
05:24where are you seeing that ring true the most?
05:28Beautiful part right now is it's across the data lifecycle. It's increasing migrations because
05:34getting data into a place like Snowflake, where the data is AI ready, has an added impetus. Because
05:39of AI, it's also automating the process of migration. So people are bringing in more data sets. They're
05:44building more with products like Snowflake Intelligence because they can get value faster,
05:49distributed more broadly. Just here, I wanted to look up what exactly is United Rentals doing in
05:55other areas. That was just a single question into my Snowflake Intelligence app that's right on my
06:00phone. That's value when I want it, where I want it. That's the kind of thing that every CEO that
06:07I show
06:07this to wants because they want those insights to be available at their fingertips. So it's actually
06:12a broad acceleration across many, many things that we do. And as models like Mythos come along,
06:19we are able to take advantage of them in our products and get our customers to do even more with
06:24Snowflake.
Comments