Snowflake Build London Keynote

Turn data into your [music] competitive edge with Snowflake on AWS.

One unified, [music] secure platform to transform, share, and analyze data seamlessly.

Over 7,500 customers are using Snowflake [music] on AWS to drive impact across industries.

Advance [music] generative AI strategies, increase insights, develop faster, and simplify your tech stack with Snowflake and AWS.

Every great idea begins just out of reach.

Builders bring ideas into reality, turning data into dreams and code into creation.

Where complexity used to constrain, today's builders are coding without limits. Where your data

limits. Where your data meets your workflows.

Building apps and [music] agents that turn insights into action.

With limitless data for accuracy and a powerful platform for execution, you're building new possibilities, pushing the boundaries, and breaking barriers.

Because AI isn't just the future, it's being built today by you for every

business, user, app, [music] and agent.

Show the world what's possible.

Welcome to Bill.

Please welcome Christian Kleinerman, EVP product management at Snowflake.

>> Good morning, London.

Good morning, London.

[cheering] >> That's better. How's it going? Welcome

to Build.

This has been a long journey of Build events. We did over 60 global snowflake

events. We did over 60 global snowflake and community events, over 120,000 people registered,

and more than 20,000 people that signed up for trainings. all part of this wave of build events that we started in November back in the Bay Area. And

today, of course, we're doing the largest set of new product announcements here in London at the O2. Anyone excited

about some new announcements?

Oh, before we get started, big thank you to the partners that make this event here today. Let's give a round of applause to

today. Let's give a round of applause to our sponsors.

>> [applause] >> And if I have one message for all of you today is that AI is today. And what does that mean?

If you're thinking that AI is something you're gonna do for next fiscal year or for next quarter or for next month or for next week, you're behind

because companies many and many of you in the room are getting value from AI today. And a lot of what we do at

today. And a lot of what we do at Snowflake to innovate is to help you adopt AI more easily with less friction

and faster. And you'll hear from some of

and faster. And you'll hear from some of our customers hopefully you'll realize that the opportunities to make your organization more productive, more

efficient, to have business outcomes sooner, the possibilities are available right now.

At Snowflake, the pace of innovation is relentless.

I have in this screen a few of the last launches that we made. I think I would have to have a tiny little font if I were to include all of the innovation

that's coming from Snowflake. We're

doing roughly a hundred new launches every quarter.

But a lot of what we do is to simplify the experience for all of you. Simplify

the amount of heavy lifting you need to do. We do all of that for you. Hopefully

do. We do all of that for you. Hopefully

all of you get to focus on how do you help your business? How do you get more value out of your data?

Of course, how do you build applications? How do you build agents?

applications? How do you build agents?

And our goal, our no northstar at Snowflake is to help you through the entire life cycle of data. Whether it's when data is

born or created, whether it's an IoT device or a sensor or an application or a database, all the way through transformation and

ultimately through consumption. This is

sort of my internal to-do list. Anything

that we can help you through the entire life cycle of data that's in scope for us as a mission for Snowflake. And how

do we do this? Through the Snowflake platform.

I'll assert Snowflake is the most comprehensive data platform in the market.

Fully managed from day one. We focus on removing as much of the infrastructure as possible and simplifying as much of the journey as possible.

With governance and security at the heart of everything we do, I've shared this anecdote in the past, but I I don't get tired of repeating it.

Our two founders, Benois and Tierry, the first employee they hired was the security architect because security runs deep through everything we do at Snowflake.

And of course, we focus on be able to store, manage data, help you get value out of data. And the other message

you'll hear consistently from us is one of choice.

We don't want to force you down any one specific path. Whether it's the type of

specific path. Whether it's the type of data that you want to store or manage with Snowflake, structured, semistructured, unstructured or whether it's the type of compute that you want to leverage with Snowflake. We have a

warehouse model, we have a container model, we have access to GPU instances, we have of course CPU choice is very important in how you get value out of Snowflake and value out of

your data. And of course, programming

your data. And of course, programming languages SQL is where we started. We

have secure hosting of Python, Java, Scala and through Snowbark container services you can host pretty much any programming language that you want.

So for today the journey is structured in three different chapters and those chapters we've organized them based on the needs that we believe many of you

here in the audience can relate to or face on a regular basis. The first one is how do I remove all the friction from building

like let's take out all the noise and keep it as simple as possible. Second

one is how do I activate data and models? And the third one is how do I

models? And the third one is how do I build applications? How do I build

build applications? How do I build agents? How do I build agentic apps? So

agents? How do I build agentic apps? So

let's get started on the first one.

Chapter number one is I want to be able to build without friction.

Something that we have obsessed at Snowflake is to be able to deliver tools that you will fall in love with.

One of those tools that you say once you have it, once you've tried it, you you're not taking it away from me.

That's what true love for a tool looks like. And we just focus also on the

like. And we just focus also on the notion of tools and technology that it just works. From the beginning of

just works. From the beginning of Snowflake, we've always thought about ease of use, simplicity, and the same thing is for tools, whether it's userbased tools, command line

tools, extensions to IDEs like VS Code, and of course our own first party UI, which is Snowite.

And a big part of Snowite is an introduction that we did in the last few months, something called workspaces.

Workspaces is our re-imagined modern new IDE like interface for creating projects and managing

filebased projects in Snowflake.

Whether it's SQL files, Python files, DVT projects, stream lead applications, notebooks, you'll hear more about it in a second. Git integration, all of it is

a second. Git integration, all of it is part of workspaces. And today we're super excited to share with all of you about a new extension to workspaces

which is called shared workspaces.

What this enables is the ability to do teambased collaboration on workspaces.

For those of you have used workspaces to date, you'll say, "Well, that was only me and it was difficult to share and collaborate with others." Okay, starting

today, that's a thing of the past. You

can be able to have a multiple individuals users collaborating on a single single workspace. There's a draft and publish model. So you can do a draft this version of accessible only to you.

When you're ready, you publish it to the rest of your team. There's a full history of changes. And of course all throughout there's rolebased access control, security and governance at the

center of it.

But it is not just about IDE and tools and editors. We've been focusing on bringing

editors. We've been focusing on bringing the absolute best of AI into the tools that you use to program Snowflake, to

manage Snowflake, to leverage Snowflake.

And probably one of the the biggest most significant launches that we've done at Snowflake since the beginning, for sure in the last few years, is the

introduction of Cortex code.

Let me explain what cortex code is. And

and by the way, cortex code is starting now generally available for all of you.

So I'll tell you first what it is.

I've been asked this question now many times in the last few days. And the

easiest way for me to describe it is what coding assistants have done for programming. And I assume many of you

programming. And I assume many of you are using coding assistance. Anyone here

using coding assistants?

Okay, a few of you good. So for those of you that have

good. So for those of you that have leveraged code assistance for programming and software development and software engineering, we're bringing that same paradigm to database management, database operations,

database um programming.

That is what cortex code brings. But the

most important piece is it knows Snowflake. We went through a very

Snowflake. We went through a very systematic process of taking the knowledge of the entire product and engineering team of Snowflake and that

knowledge is in Cortex code. So it's

vastly superior that pretty much any tool that you take out of the box because we've gone through that effort of making sure that it works. Whether

it's an administrative task, maybe you want to do financial governance. We've

talked with many of of our customers and organizations for many years. If you

have uh need to manage cost it's easy but now you can say cortex code this is what I want to do this is my policy go configure snowflake or maybe you want to do some wrangling of data and create a

pipeline you may know all this syntax to do pipelines but you can just say cortex code please create a pipeline maybe you want to configure data governance cortex

code does that for you or maybe you just want to go and create policies that help customers have access to data All of that is done with Cortex code. I

cannot emphasize enough how significant this is.

I can also tell you that Cortex code shows up in two different environments.

Environment number one is part of a user interface in Snowite. It shows like shows up as the right hand pane where you have contextsensitive

assistive experience. If you're in a

assistive experience. If you're in a notebook, you can go and say quite this code add a cell and do this. But if

you're in the warehouse page, you can say go and create a warehouse and configure it. So it knows where you are

configure it. So it knows where you are based on where you are. It has the smarts to help you and assist you with the task at hand.

We're also been adding capabilities to Cortis code to understand third party tools, things like oh maybe help me orchestrate this with airflow or with

dbt. And our goal is to continue to

dbt. And our goal is to continue to build those skills so that the tasks that you or many of your teams do on a regular basis are simpler and simpler.

But today we're also excited to introduce the concept of Cortex Code CLI. Some of you may may say, "Oh, I'm a

CLI. Some of you may may say, "Oh, I'm a terminal type of person." So we got you covered. And what cortex code cli does

covered. And what cortex code cli does is it brings a command line agentic interface to snowflake. It runs and interfa interacts with a local

environment. So that gives you some

environment. So that gives you some differences relative to what cortis code and snow can do. And of course you can use it from many of your favorite environments whether it's v code or

cursor or other tools.

At the end of the day, a lot of what we're doing with Cortis code is about productivity.

And the stories that we're hearing from some of the early adopters of Snowflake are mind-blowing. When I tell you I I I've

mind-blowing. When I tell you I I I've been on stage many times. I've announced

all sorts of cool and amazing capabilities we launch. This one may be one of the most significant based on the productivity boost that many of you are going to get. And again I will emphasize

this is available for all of you to try it today. Hopefully all of you go to

it today. Hopefully all of you go to your offices and to your homes and say I am now dramatically more productive with snowflake.

But it's not just what customers may say. Any of you wants to see cortex code

say. Any of you wants to see cortex code in action?

Anyone?

[applause] Okay. So this guy that always sneaks up

Okay. So this guy that always sneaks up behind me. Dash, I want to introduce

behind me. Dash, I want to introduce you. Dash Des the Sai developer

you. Dash Des the Sai developer advocate. He's demo master. He's the

advocate. He's demo master. He's the

best at showcasing technology for Snowflake. It is also his birthday

Snowflake. It is also his birthday today.

>> Oh, thank you. [cheering]

[applause] >> I I think he didn't want me to say that.

And with that, Dash, please show us what you have. Take it over. Give it over to

you have. Take it over. Give it over to for Dash.

>> Hello, London.

This is an amazing time to a builder.

Snowflake provides so many options for you to build whether you're working in snowite or you're working in ID for example VS Code. Who's excited to see

some of these demos?

Come on, let's go.

All right. So, what you see up here is basically a Snowflake extension within VS Code. I'm logged into my account and

VS Code. I'm logged into my account and I can run queries right from here within VS Code. I can see what kind of tables I

VS Code. I can see what kind of tables I have. I can look at the actual data in

have. I can look at the actual data in my in my account. Here are some orders, stuff like that. Now, whether you're coding within

UI or CLI, Snowflake has you covered.

Okay. So for example, I'm going to go to terminal window and use no CLI to run a query. Now you can use Snow CLI to also

query. Now you can use Snow CLI to also deploy pipelines, deploy streamlid dashboards and what have you. So the

choice is always your as a builder. Like

I said, this is a great time to be a builder. Now let's go ahead and look at

builder. Now let's go ahead and look at workspaces. For example, if you're not

workspaces. For example, if you're not if your team members are not using any IDE, you can use a lot of these features obviously within Snowite. Now, if I

switch back to Snowite, you'll see that on the left hand side, I have the same GitHub repository open that you saw in Visual Studio Code. This is the exact

same file I have. Obviously, I can run any SQL here that I want and get the results. Now besides getting results. So

results. Now besides getting results. So

let me go ahead and run one of these other SQL queries. So you can see the results down below. You can also see charts just by clicking on that button

right there. Obviously this query does

right there. Obviously this query does not require a chart. So it's not going to build a chart. He's not going to hallucinate, I guess, right? Uh which is a great thing. But check this out. I can

also look at the query history right here down below. Not only that, I can also look at my DBT pipelines and models that I've built out and deployed in my

account. Now, super amazing thing that

account. Now, super amazing thing that uh Christian just mentioned, Cortex Code CLI. Who's ready to see that in action?

CLI. Who's ready to see that in action?

>> Whoa.

>> Come on, let's go. [applause]

All right, so I'm going to switch back.

This is a terminal window. Here I have launched Cortex Code CLI. Now, here you can create skills and there's also built-in skills that come out of the box

that you can use. I've actually built a skill that's going to help me demo so I don't have to type.

Yes. Okay. So, I have four prompts in here. The first one I'm going to run is

here. The first one I'm going to run is as a data analyst, I want to be able to see a query that's going to show me trends of month overmonth growth. Now

all I have done is entered one. So it's going to pick up

entered one. So it's going to pick up the first prompt. Again this is a custom skill I've built out. You can obviously type the prompt as you see fit. But here

you'll see that as a data analyst I can ask show me monthly revenue trends. Now

it's going to think look at all the tables all the things that I have access or it has access to my account to give me a query and the results. Okay. So

here you'll see that it's trying some things out and here you go. So not only the query also the results and our explanation in natural language what

exactly it's doing. Pretty neat. Yes.

It's okay to clap. Okay. As a data engineer now I'm going to ask you to build a DBT model for customer churn risk. Now everything that's happening

risk. Now everything that's happening right now you can also actually just save it to a file or files so you can look at it later. You can review it what

have you. Now where are the files going

have you. Now where are the files going based based on the skills that I have uh developed the files are going in this uh folder right here demo output. So the

very first thing that uh the prompt that I ran was about monthly revenue trends.

So here's a SQL that it wrote out. I can

review that or not. Now the DBT uh model is being created right now. You can see some results being populated and that is also going to be generated into a file

that I can look at later on. From here I can also actually deploy the pipeline if I choose to. Okay, we'll give it a

second. Um, so far pretty awesome. Yes.

second. Um, so far pretty awesome. Yes.

Thank you.

Let's keep it interactive. Now, here's

the here's the DAG that was uh just created. Here's the DBT pipeline or DBT

created. Here's the DBT pipeline or DBT model I should say and also explanation of what exactly this model is for. Now,

as a developer, I can also have it review my code. I

have Snowpipe streaming running in my in my project. I wanted to actually enhance

my project. I wanted to actually enhance the the uh orchestrator so that it has a good uh logic to basically retry if

there's any errors. Okay. Uh give it a second. And then what we're going to see

second. And then what we're going to see is actually launching a streamlined application right from here. Okay. So

let's go back to workspaces and I want to show you guys uh something else here.

All right. So from here you can also basically look at different files Python SQL and also split plane. So what I can do here is click on this split right. So

you can see files side by side. It's

pretty amazing. From here, I can also look at all the changes that have been done, pull down the changes or also push back any changes. If I click on one of these files, it's going to show me the

differences between the two. All right,

let's go back. You know, I'm running out of time. There's so much to show. I

of time. There's so much to show. I

really wanted to show you guys this. So,

here as a developer, how many people are writing stream applications these days?

So, check this out. This is going to write a Streamlit application based on my prompt and also actually launch it so I can actually look at it or see what it looks like. Okay, just give me a second

looks like. Okay, just give me a second and I promise I won't take up too much time because I know CK has a lot more to share and so do our customers. Okay, so

you can clap if uh if you want.

[applause] This is uh created by Cortex code CLI from scratch based on a prompt that I had given. What do you guys think?

had given. What do you guys think?

Pretty amazing. Okay, so I know you guys want to see a lot more demos, but for now I'm going to hand it to Christian.

Back to you, Christian.

[applause] Awesome. Right. Or should I say

Awesome. Right. Or should I say brilliant.

And I I think it's it's quite impressive what you can do and and you just saw a very very small sample of the types of CA capabilities you can build with with

snowflake cortex code. We've been using it internally for months now and on a regular basis I get stories of things that I I think were not possible before

at least not in the time frames that we're seeing. And yet this reality is

we're seeing. And yet this reality is available today to all of us.

So I started with let's eliminate friction and hopefully you see that we're not only helping you eliminate friction but give you material ability

to go faster. So it's not about just don't go slow it's let's go faster.

That's what the productivity tools workspaces cortex code are all about.

But now let's talk about okay I have tools but there's little I will do in the enterprise with AI without data. So,

I want to be able to activate all my data. I want to be able to bring models

data. I want to be able to bring models and connect them to that data. And at

Snowflake, we've been saying now for for a couple of years, there is no AI strategy without a data strategy. And

many of you are now in the middle of living this sentence here, which is I want to do AI, but if my data is fragmented in all sorts of places, if I

don't have clear governance policies, I don't have clear permissions, AI is going to have a hard time giving value to your organization or giving value in the way that you expect it. So

let's talk about what Snowflake can do to give that ability for all of you to activate data and models. And it starts with the introduction of Open Flow which

we brought to market roughly in the June time frame at our summit conference.

OpenFlow is a managed service. It has a number of connectors and it simplifies ingestion and data access from a number of sources. Here you see in the diagram

of sources. Here you see in the diagram databases and structured data streaming applications whether it's a well-known application

Oracle, SAP, Salesforce or a database like Oracle all of that is easily connected from OpenFlow to bring data

into Snowflake. But once you have data

into Snowflake. But once you have data in Snowflake, what you want is to be able to transform that data. Maybe you

want to augment it with other data.

Maybe you want to cleanse it. Maybe you

want to do a data quality pass. And we

give you a choice. You want to use declarative and use dynamic tables or you want to use dbt projects hosted natively inside of Snowflake, we're totally cool with that. But if you want

to write some code, whether it's SQL, whether it's Python, we have Snowpark, Snowark Connect, which implements the

Spark Connect API. If any of you have Spark pipelines and you want to modernize it, you want to go faster, you want to save money, please let me know.

Send me mail directly. I'll I'll I'll connect you with the right person at Snowflake. I'll help you save money.

Snowflake. I'll help you save money.

Snow park. Snow park connect. We have

Pandas API. We have lots of capabilities to just help you transform data with higher ease of use.

But the other question that many of you think before I say oh I'm pulling data and I'm breaking down silos is what is my data architecture and we also want to support the architecture that you want.

We have many customers saying snowflake is only my golden data and analytics layer. Okay

layer. Okay we have customers saying I want to use a data mesh. We're a large organization a

data mesh. We're a large organization a combination of federated and central. We

have all the capabilities to properly help you in that journey in that architecture. And if you say I want to

architecture. And if you say I want to do a lakehouse, we can also help you with that.

But across all three, what stands out the most is the enterprise nature of the architectures that we built. We want to make sure that security,

interoperability, and things like disaster readiness are at the center of whatever you do. No architecture is good

if when an outage of of a cloud provider takes you down and then you say, "Hey, my architecture was awesome, but it's no longer working and my business is not being well served." We bring those enterprise capabilities and we help you

all forget about the details of the infrastructure and again focus on data and getting value out of your data.

We're also introducing at least for for the interoperable storage the ability to have snowflake managed storage so that you can say I may want to choose Apache

ice as a file format as a table format but also store it in snowflake so you have fewer tradeoffs to make in how you choose your architecture.

But if you go back to when I share the life cycle of data at the very far left, you think of how data is born. And

oftent times there's applications and those applications need state. Even if

it's an AI agent, the agent may need state. And back in June, we made an

state. And back in June, we made an announcement that caused so much buzz and excitement from many of you around the introduction of Snowflake Postgress.

Many many of you said, "Oh, we want a database, an old database, and we want Postgress." And today here at Build

Postgress." And today here at Build London, we're extremely extremely excited to announce that Snowflake Postgress is now generally available.

[applause] So what is Snowflake Posgress? Fully

managed posgress experience enterprisegrade security out of the box.

It's very important. It's part of our unified data architecture.

Simple oneclick provisioning. We take

care in stool fashion of all the automative maintenance. So it frees your

automative maintenance. So it frees your team from dealing with upgrades and other type of uh operational activities

and it is prepackaged with a number of common powerful extensions things like PG vector things like postGIS which just increases the capabilities of what you

can do. And speaking of extensions,

can do. And speaking of extensions, we've also made available as opensource an extension that we developed called PG

Lake. And what PGL enables is the tie in

Lake. And what PGL enables is the tie in between that lakehouse architecture that we spoke about and Postgress. What that

allows is Postgress to seamlessly read or write data that is stored in Apache iceberg table formats. And at that point

you have the full synchronization of data in your transactional OLTP database Postgress but it's also available in ice tables which enables the rest of

snowflake to add value.

All in all, this makes Snowflake a data complete offering from structure, semistructure, unstructured, icebreak tables, snowflake

tables, hybrid tables and now with Postgress, we have the full spectrum of needs that you may have for storing your

data. But when we talk about data and

data. But when we talk about data and then we talk about AI, there's something that needs to bridge the gap between those two. And that's why we introduce

those two. And that's why we introduce something called semantic views because semantic information is what makes AI models

work at its best. And we introduce semantic views. All of you said yay. And

semantic views. All of you said yay. And

then you tried to populate it and you said, "Oh, this is a lot of work." So we listened to you. We heard you. And today

we're we're introducing brand new semantic view autopilot. And what this does, it helps you automatically

populate an initial creation, but also iterative maintenance and editing of semantic views. And we leverage

semantic views. And we leverage knowledge that is unique to Snowflake about how you are using your data. We

plug into things like query history and we recommend saying, hey, these are relationships you should be included or these may be some verified queries on your semantic view. The goal is to

accelerate the development of semantic views which helps you accelerate the deployment and leverage of AI in your enterprise.

The last comment I'll say is we did semantic views. Many of you said yay and

semantic views. Many of you said yay and then we said well am I going to be locked in? I thought you snowflake were

locked in? I thought you snowflake were committed to interoperability. And the

answer is yes we are. And we spearheaded an effort called the open semantic interchange consortium where we're coming together with a number of players

in the industry. I have five logos in here. That's only a tiny subset of the

here. That's only a tiny subset of the number of companies that are collaborating with us on this front to make semantic views interoperable and interchangeable between systems.

But when I talked about the life cycle, I said posgress on the far left. Now

think about the far right. What if you need to deliver insights at low latency?

I want a dashboard. It needs to be super fast. Or I want an application that

fast. Or I want an application that needs to be super fast. And that's why at snowflake we have introduced snowflake for interactive workloads

also generally available. And what is this? It has two concepts. One is the

this? It has two concepts. One is the notion of interactive tables. The other

one is the notion of interactive warehouses. And what this enables is

warehouses. And what this enables is snowflake to have optimized configuration both compute and data for

realtime use cases. pre-wararm caches,

low latency analytics, near realtime delivery of data so that you can have very snappy, very uh responsive experiences.

So I I help you figure out how do we bring data, transform it, manage it in snowflake. But now you may be wondering,

snowflake. But now you may be wondering, okay, how do I actually get value of data? And for that we have a speaker

data? And for that we have a speaker that will come onto the stage. I want to welcome Ana Lucy. She is a PM director.

She is the creator of workspaces and a lot of what you've seen here. Please

welcome Ana London. I am so happy to be here with

London. I am so happy to be here with you guys today. Everyone back in California is jealous that Dash, Christian, myself, we all get to be

here. Um,

here. Um, let's get started. I want to talk to you about the journey Christian mentioned and what comes next.

So when we think about um the full life cycle here, right, Snowflake is allowing you to store the data in the format you

want, bring in the types of data that you want, transform it in a seamless way that works for your company. Get it to that enterprise ready stage, right? with

whether it's you know semantic modeling of it to build that quality and accuracy or the real time nature that gets you fresh insights but what about that last

flow of hey we want agentic power we want predictive power that's what we're going to go over and we're going to zero in on Snowflake ML

um this product suite is getting massively upleveled today and I want to talk to you guys about a couple of our most exciting launches here all all the way from training through to deployment

through to online inference. Um we have some things that you guys are going to be really excited about.

So the first thing here in the journey for a lot of folks is notebooks, right?

We start exploring the data. We start

building models, iterating, experimenting.

Um we launched our very first version of notebooks one and a half years ago.

Many of you guys have adopted it. We've

been getting incredible feedback ideas.

We've in parallel been developing a lot of supporting technologies that today I'm very very excited to announce we're taking it to the next level. Uh

notebooks are going in workspaces.

Yes, it is a clappable moment. It's a

clappable moment because developers are telling us that they're shifting their their real place of work into workspaces. Christian just told you and

workspaces. Christian just told you and you saw with Dash workspaces are are now agentically powered here too. So what

does it mean for ML and notebooks? It

means that uh you get that that that feeling of hey all my files are in the same place. I can build these more

same place. I can build these more complex projects. Our uh kernel is now

complex projects. Our uh kernel is now Jupyter compatible. It's more easy than

Jupyter compatible. It's more easy than ever to bring in your notebooks that you may have been working on in other places. Uh we also of course have the

places. Uh we also of course have the container runtime. This is a very

container runtime. This is a very powerful built-in runtime for machine learning and data science containing over a 100 packages right at the click

of the finger. You do not need a PhD in Kubernetes, right, to figure out how to set this up. It's right there. It's

sitting in snowite and of course this is 100% arbback governed. There are no additional permissions to set up.

Um so so cortex code you saw it earlier but we've really really gone deep on this with data science and machine learning. We've built skills into cortex

learning. We've built skills into cortex code that can take your natural language prompts right. Hey, build me a

prompts right. Hey, build me a prediction model for COVID case trends or you know a pricing recommendation algorithm. Taking those prompts, knowing

algorithm. Taking those prompts, knowing your data, knowing the setup of your account and uh basically automating that ML workflow for you, right? You get that

whole flow in your notebook. You can

talk to Cortex code whether it's to update it, try something different. Of

course, you can still manually edit it right in the UI there. And we think this is really kind of the great next leap forward um in machine learning and data science.

So what about after you've developed that model, you've maybe used our experimentation framework which is also

now GA uh to kind of choose what you want to go to production with, right? Uh

today online ML is generally available.

This is huge. So, think about all of our customers, right? Use cases like

customers, right? Use cases like pricing, real time fraud detection, real time recommendations, next best action

models. These require ultra low latency

models. These require ultra low latency serving of predictions. So, what what is low latency? What do we actually mean by

low latency? What do we actually mean by that? Um the the feature store of

that? Um the the feature store of Snowflake can serve features in as little as 30 milliseconds.

And on the prediction serving side, we're down to 100 milliseconds.

That really brings just a whole new set of use cases uh within reach. Again,

it's all within the Snowflake ecosystem.

Um it's all governed by the ML observability suite that we've had and we are really excited to see what you guys are going to build with this. But

before that, we're going to see what Dash has built with this.

>> [laughter] >> Just pretend that you haven't seen me before. Okay.

before. Okay.

All right. So, what you see here is what just Anise uh uh described. It's a

notebook in workspaces. It's a brand new thing that we're launching. It's

Jupiterbased uh which is a big difference between the first version and this one. Now, I wish I had time to run every single cell, but

I can't, okay? Because I don't want to take too much time that you guys have at your hands. Now, what I've done is I've

your hands. Now, what I've done is I've run all the cells in this notebook. I'm

building a product recommendation model on GPU. So, you could have these models

on GPU. So, you could have these models trained on CPU or GPU like Ana mentioned. And it's as easy as clicking

mentioned. And it's as easy as clicking on this and either selecting an existing service or creating a brand new service.

From here you see that this service is a GPUbased service. Now let's go through

GPUbased service. Now let's go through some of the cells. So first we're setting some execution context uh using some uh libraries that we want including

so for Python. We're loading some data.

We're checking what the data looks like.

We have uh quite a bit of data set. uh

so it's easy to kind of train uh split into uh training and test data sets.

Okay, we can do some feature engineering using soapart uh python as well. Here we

are creating some customer statistics using soark of python so that we can train the model based on the results of the feature engineering. Now down here

we're loading the data and we're ready to encode some categorical features. So,

a lot of these things are very basic when they're trying to train a model.

But what I'm trying to get to down here is um let's look at the training first.

Uh we split the data, configure the model parameters, very basic stuff uh that you normally do. And um once you have a trained model, you can obviously

make predictions and look at the accuracy and what have you. Now, this is my favorite part right here. Once you

have a model trained, you can store this model or uh block the model in snowflake model registry. What this allows you to

model registry. What this allows you to do is basically version and manage ML models as a first class snowflake object, which is pretty amazing.

Everything's backed by rolebased access control. So you don't have to worry

control. So you don't have to worry about security. And the other cool thing

about security. And the other cool thing I'm doing here, so let me show you what the uh model registry code looks like.

This right here. Okay, so we calling log model on registry object providing all the information for example the model model name what have you and also all

the metrics the cool other cool thing there's two cool things here I'm creating a service endpoint model inference service what this means is I

can use this model as a deployed model running in either soap container services that can serve as for predictions now Let's see how to use the

service endpoint. For that I'm going to

service endpoint. For that I'm going to bring up uh Cortex code. So before we looked at Cortex code CLI. Cortex code

is also embedded within Snowite. So I

just clicked on the button on the bottom right. I'm going to prompt. What I'm

right. I'm going to prompt. What I'm

going to do here is um give it a prompt to basically create some Python code for me that will allow me to test or measure the latency for this endpoint when I'm

calling from different applications.

What it's doing right now is looking through my code. It's going to generate some Python code, actually put it automatically in a new Python cell right

here and actually run it so that I can see what the endpoint might look like from an application that I'm calling this model from. Now, you could be

spending at least an half an hour about an hour to write some code. You have to figure out what charting libraries you want to use and so forth. Here, you

don't have to do any of that. You prompt

Cortex code. It's generated all the code. It's actually also about to run

code. It's actually also about to run it. So, here we go. It's run the uh

it. So, here we go. It's run the uh added the code to a cell and it's running it right now. So, that I don't have to actually do anything. It's

pretty amazing, right? Yes.

[applause] Which is the results in just a second right here. Including charts.

right here. Including charts.

What do you guys think? Amazing.

[applause] Awesome. So, I know you guys want to see

Awesome. So, I know you guys want to see more demos.

You can't help it, but I have to go for now and I'll come back later. Okay, back

to you, Christian.

I hope all of you are getting ideas of how you could leverage this technology in your organization. But sometimes it's not as interesting to hear from us saying, "Hey, technology is cool." So

that's why I want you to hear from our customers directly. And for the next

customers directly. And for the next segment, I want to invite onto the stage Neha Patel. She runs the machine

Neha Patel. She runs the machine learning team at I A Loyalty. Neha,

welcome.

Okay, Nha, thank you for being here and maybe let's start with tell us about you, your role and I loyalty.

>> Yes, of course. Hello everyone. I'm Niha

Patil. I'm lead machine learning engineer at I A Loyalty. You may have heard of AIOS currency which you spend

redeem at which you spend redeem when you book uh book airlines. IIG loyalty is a home to

book airlines. IIG loyalty is a home to AIOS currency and British holidays.

Today we have 69 million members collecting and redeeming a uh airlines and this generates huge

amount of huge volume of data. Our

machine learning teams use this data, turn this data into personalized actions for our members.

And my role as a machine learning platform lead is to make these teams faster while keeping things reliable, scalable, and govern.

>> Yeah, that's awesome. So maybe tell us a little bit about the use cases that you were facing. What got you into thinking

were facing. What got you into thinking about, hey, Snowflake can help me? Yes,

of course, machine learning is used within entire AOS member life cycle. We

have machine learning models for code loyalty use cases, customer segmentations recommendations personalization, winback

and winback.

These models directly help our members to earn AOS to

redeem Fios and connect with our airline and non-airine partners.

>> Yeah. Okay. So may maybe I want to hear about the architecture. How how do you decide to build this?

>> Yes. Yes. Of course. Our machine

learning platform architecture is by built directly on a snowflake.

When we decided to move our data and analytics workload into snowflake, we made a deliberate decision to also bring machine learning closer to data and

hence the snowflake. What we have is we have a centralized single trusted snowflake account where we are bringing

all customer data, customer interaction data from multiple snowflake accounts into this single trusted snowflake account. In there directly we have built

account. In there directly we have built feature store our model training and prediction pipelines as ML jobs which

runs using our Python projects model registry scoring and output tables and moreover model observability directly

into a snowflake platform a unified platform which help us to move faster while keeping things govern

>> amazing and now Maybe a final question.

What are the benefits? Benefits? What

what are you seeing on why is all of this good for you? What are the the outcomes for your business?

>> Yes, sure. Yes. Snowflake takes away lots of operational overhead and also uh gives us reliability and scale which we

need to run to run loyalty experiences at scale. One of the thing we really

at scale. One of the thing we really value about Snowflake is it provides several first class ML capabilities

together. Feature store, explanability,

together. Feature store, explanability, evolution framework and integrated notebooks.

These are some of the capabilities which are found together in Snowflake out of the box which are hard to find together

at at anywhere else. And this this allows this gives us a seamless development experience, familiar development experience as we work with

Python a lot while keeping our model outputs and tracking actions models are taking trustable. Yeah.

trustable. Yeah.

>> Okay. Nhi, what are you doing with Snowflake? It's very impressive. Soup to

Snowflake? It's very impressive. Soup to

nuts. All machine learning solutions with Snowflake. Thank you for choosing

with Snowflake. Thank you for choosing Snowflake and thank you for sharing your story with everyone. Thank you for being here. Give it up for Nhan.

here. Give it up for Nhan.

>> Thank you very much. [applause]

[applause] >> Okay, so we established we can help you with tools and remove friction. We can

help you get your data ready for AI. Now

let's talk about AI, which is how do I build connected agentic apps? And the

goal here is truly to bring the two sections that we just talked about into truly how do you go and deploy AI? How

do you get it to be in production in your organization? And AI at Snowflake

your organization? And AI at Snowflake comes together through cortex AI. This

is sort of like the umbrella term where a number of technologies and in investments we've made are made available to all of you. This is where

we give access to leading AI models.

Governance, I will not get tired of saying governance and security. Run

through everything we do. It applies to AI. It applies to AI models. It applies

AI. It applies to AI models. It applies

to data that is accessible to AI. And of

course, our goal at Snowflake is to make it easy for you to build applications and agents.

That's why we introduce the Cortex agents API. It helps you build not only

agents API. It helps you build not only agents but enterprisegrade agents. It has the power of all the

agents. It has the power of all the leading models or the reasoning capabilities that that AI provides for all of us these days. But it lets you

orchestrate between structured SQL uh tools, unstructured searchbased tools and of course third party tools by the

ability to call into MCP compatible uh tools which is a good segue to saying also a lot of what we've done with snowflake is if you

have an agent you can also publish it as an MCP server. So you can invoke and prop other tools that you may be using, other agents that you may be using in the enterprise. We're also introducing

the enterprise. We're also introducing here at Build the ability to share agents. In the same way that Snowflake

agents. In the same way that Snowflake pioneered eight years ago, the ability to share data between organizations, now we help you share agents. And last but

not least, we've introduced the concept of Cortex knowledge extensions, which are pre-created packages of unstructured content

available in the Snowflake marketplace ready to for you to plug in. The most

common use case we're seeing here are new sources and they're available. The

data is already vectorized. You plug it in as a tool, as a source in your agent and you get results.

For many years, we've been talking on Snowflake. We will help you secure your

Snowflake. We will help you secure your data, but we'll help you also bring computation into that security boundary.

So your data doesn't have to be copied out. So you don't have to reinvent

out. So you don't have to reinvent security and permissions. That's why we are hosting an ever growing number of frameworks and programming languages and

technologies to make it easier to have strong programming models and or leading programming models with strong authentication and strong security.

That's what we're excited about our partnership with Versel.

Anyone ready for another demo?

Yeah, Dash. Take it away.

Dash. Take it away.

Third time is a charm. Yes. All right.

So, don't forget to clap. Now what you see on my screen right now is uh a UI in Snowite for creating data agents. Okay,

it's as easy as following the prompts after you click on uh create agent which I will right now or not. Let me refresh the page.

Create agent. You select where you want agents to live. You you give it a name and then you follow different prompts to actually add tools. and we're going to look at right now what those tools can

look like. So, I'm going to go ahead and

look like. So, I'm going to go ahead and click on one of the agents that I've created for today. I'm going to click on edit. I'm going to click on tools. Now,

edit. I'm going to click on tools. Now,

Christian mentioned a lot about semantic views. That's exactly what Cortex

views. That's exactly what Cortex Analyst is showing you right now. I have

a semantic view that I've added here for my structured data. It gives context to Snowflake intelligence what my data is all about. Okay, you can add as many or

all about. Okay, you can add as many or none. Uh the way it depends. So choice

none. Uh the way it depends. So choice

is always yours depending on the use case. Now what other tool you can add

case. Now what other tool you can add here is a cortex search service. This is

based on a keyword and vector-based. So

if you have unstructured data in your account, you can create cortex search services and you add them as tools in sulfic intelligence.

I have two services for product reviews and support tickets. What another cool thing you can do here is add a custom tool. So if you recall in my previous

tool. So if you recall in my previous demo I have built a model for uh product recommendations things like those you can add them as custom tools. So if I click on this edit

custom tools. So if I click on this edit you will see that this is the model endpoint that was testing using cortex code. Okay you can on all of these

code. Okay you can on all of these things as tools in sofic intelligence.

Now what does the actual UI look like?

This is the UI that's built in out of the box is a brand new interface for business users. Here I can ask questions

business users. Here I can ask questions in natural language. You can also have some suggested questions as you see down below. I can just click on one and it

below. I can just click on one and it will start thinking and orchestrating the insights that you want to see. Now

this is pretty amazing. Yes.

Okay. So as a builder, if you wanted to use APIs to build custom applications that look maybe similar to but a little bit different than what you see here,

you can do that also. So you see right here, this is built strictly using Cortex agents APIs and React front end.

Okay. So here I can come in and say ask a question and you see it responding.

Now another cool thing here is when you when you write these applications you can actually deploy this in your account in soapbox container services. So pay

attention the URL is actually running in my account but the UI is something that I've built on my own using REST APIs.

Now Christian mentioned Versell. You can

actually build this entire UI or front-end application using the same rest APIs in Versell. So, all I did was give it a prompt saying here's my REST

APIs and I'm already connected to my account automatically through partner integration and it brought all the agents that I have access to within it and created the UI. What do you guys

think? It's pretty awesome. [applause]

think? It's pretty awesome. [applause]

All right, this is a great time to be a builder. The choice is always yours and

builder. The choice is always yours and snowflake is your friend. All right,

thank you so much and back to you, Christian. [applause]

Christian. [applause] Thank you, Dash.

Okay, a few months ago, actually like six months ago, we introduced something called Snowflake Intelligence, which is the marquee offering for Snowflake to

help all of you democratize access to data.

A lot of what you've seen comes to life with Snowflake intelligence. And I would love to share with you a lot of what we're doing, but we have can I call you Miss Snowflake Intelligence? Uh I want

to invite Effie Gwan coming on the stage. She runs product for Snowflake

stage. She runs product for Snowflake Intelligence and is going to tell you what we're doing. You're here. Effie,

welcome.

Hi everyone. Wow, it's my first time at London Build and man, now I know what I've been missing all this all this time. I'm EIE. Uh, I lead product for

time. I'm EIE. Uh, I lead product for Snowflake Intelligence as part of the AI team. I'm so excited to be here to

team. I'm so excited to be here to update you on what we've been cooking.

So, I just want to take the moment of like, wow, that was such a cool demo.

What a time we live in. With AI, with natural language, you're able to build, you're able to create people like me.

I'm a product manager now. I'm able to build end to end useful application that I can actually deploy without snowflake intelligence. Building these agentic

intelligence. Building these agentic apps can take resources and time so much work and time that you could have spent um improving what matters to you and

your business. This is why snowflake

your business. This is why snowflake have invested a lot in snowflake intelligence. It's a standalone app that

intelligence. It's a standalone app that run natively in Snowflake and available to anyone in your business. Business

users can get insights at their fingertips. And you know what's amazing

fingertips. And you know what's amazing really is with natural language, business users don't need to learn how to use SQL, how to write a lot of SQL, write a lot of code. They can just ask

the questions. And we're not talking

the questions. And we're not talking about the simple questions like the what, what happens with my matrix. It's

also the wise. And the wise are what's driving business outcomes. The wise are what's pointing out the root cause and what's what's the um actions that I

should take to actually move the needle in my business. And um we've seen enthusiasm and excitements since we G8

in November. And trust me, we've heard

in November. And trust me, we've heard so much feedback from you all as well as customers. We listen to every single one

customers. We listen to every single one of them and please keep them coming. And

today I cannot wait to tell you all of the improvements we've made since and even more to come.

So before we go to um the announcements, what does production ready look like here at Snowflake with Snowflake intelligence? We are

unlocking four highlevel capabilities to you. First with deep analysis, we're

you. First with deep analysis, we're going deeper than the what? We're going

deeper than the service level insights.

We're actually giving you the wise. The

wise are what's driving the outcome. And

then trustworthiness.

And you've, as you can probably know, um, actions are driven when you trust the insights, when you trust the data.

With Snowflake intelligence, we ground these insights on your company data. We

are not building just a generic chatbot or generic agents. Um, you've all seen so many agents. Sure, everybody is building agents. These are so easy.

building agents. These are so easy.

Everybody is a builder nowadays. But

getting an agent that actually works, that is a lot of work. But Snowflake

intelligence make it even possible with your company data. It's truly your enterprise intelligence agent. And then

what we've heard from you and especially for me personally as a PM, it's really hard to get anything done or like to make a really effective business decisions, you need your team. It's not

something you do in the corner by yourself. It's really about bringing the

yourself. It's really about bringing the team together, working side by side, collaborating on the same insights, and driving real outcome like saving and

sharing information all while respecting the Snowflake security and governance that we all know and love. And then last but not least, I

and love. And then last but not least, I hinted on um secure and trusted. So

everything within Snowflake intelligence, the foundation is the Snowflake are back.

So today I want to focus on collaboration and uh we've been thinking about it for some time and we've heard loud and clear from a lot of you here that it's super important for business

users to be able to save and share insights. I'm especially excited to

insights. I'm especially excited to announce you artifacts. With artifacts

you'll be able to do that just with one click. And in fact this is core of part

click. And in fact this is core of part of my job as a product manager. When we

first started building artifacts and made it available internally, the first thing that I did was ask for snowflake intelligence consumption and usage like monthly active users. And then I share

it as easy as one click with team members, marketing people, salespeople, accounts people, and engineers. So we

are all looking at the same life reference. And guess what? Gone are the

reference. And guess what? Gone are the days when I had to take a bunch of screenshots and then share it over email. People have copies from three

email. People have copies from three days ago. And you know what? I don't

days ago. And you know what? I don't

even know if people have access to that data. But now with artifacts, we are

data. But now with artifacts, we are sharing the life reference of the data.

So, and respecting the user's context.

So, people only see what they are supposed to see. And I know that's a lot, but truly for me, I think it's going to be gamechanging. And this is just the beginning. We'll have a lot

more um to enable collaboration.

And I'm excited about these features because of the impact. Truly, building

an agent is one thing. Everybody does

that. Building an agent that is actually transforming the way we work. I don't

know about you, but for me, I cannot imagine the day without snowflake intelligence because today I can easily get insights at my fingertips. I'm not

blocked and the data team. I don't have to file tickets and data team loves me because I'm no longer bugging them asking for insights and analysis. So

truly from sales, marketing, accounts, product management, I've done product for many years and I've just never seen a time where where everyone is a builder. Truly, we're giving you the

builder. Truly, we're giving you the power to make things happen. And um our final product highlight today is giving you these powerful capabilities

alongside choice and flexibility. Just

like what Christian highlighted, SI works across open data sources. Now you

can get all of this awesomeness from Snowflake intelligence whether your data is on Snowflake or other other sources in with an open format. This is all of

your knowledge with your trust enterprise intelligence agent Snowflake intelligence. With that, I would like to

intelligence. With that, I would like to end with special thanks to um EMIA partners from data providers, technology

partners, consulting partners. Truly, we

can we wouldn't have been where we are right now without you all. I'm looking

forward for amazing partnership. Please

help Snowflake intelligence adoption and help everybody succeed. Thank you so much. And with that, I would like to

much. And with that, I would like to hand it back to Christian.

and we'll continue to innovate for all of you on Snowflake Intelligence, Cortex Code, and other fronts. I want to I want you to hear from one more of our amazing customers

because nothing beats real world experience. Please join me in welcoming

experience. Please join me in welcoming Kieran Kuran from Booking.com. Kieran,

come on in.

Thank you for being here. Pleasure.

>> Okay, tell us about you. I think most people know Booking.com, but tell us about Booking.com.

about Booking.com.

>> Sure. Yeah, I'm sure everyone knows about uh Booking.com. Uh we are on a mission to make it easier for everyone to experience the world. So, I am part

of uh trips data within trips business unit. Uh in Booking.com uh trips covers

unit. Uh in Booking.com uh trips covers everything except stays. Uh we are responsible for flights, attractions, cars, taxis and insurance.

>> What types of problems were you solving when you first said hey Snowflake is the answer?

>> Sure. So one of the uh verticals that I support has over 200 employees and data team is uh like 20 specialists. uh they

were constantly getting questions from business about the data on a regular basis and these questions were uh business critical and time consuming and it was causing lot of churn on their

side. They're not able to focus on more

side. They're not able to focus on more complex and uh strategy analytics work.

Sometimes it has also led to prioritization challenges. So we wanted

prioritization challenges. So we wanted to wanted a way to overcome this by taking data closer to the owners of the data in a self-s served way in such a way that they can talk to their data

while still preserving accuracy and trust.

>> Okay. And now okay you did all of this and what's the impact what do you have to show as the results of this work?

Yeah. So, uh we wanted to uh build a data agent and our guiding principle was to build a agent that's grounded and uh uh auditable and it should live within

the security and uh governance parimeter of our central data platform that's powered by Snowflake and we initially thought of building a uh touchto SQL

agent but building it ourself would have taken months of effort and we had to navigate through security and governance challenge. challenges. So that's when we

challenge. challenges. So that's when we heard about Cortex Analyst being GA and because it's native to Snowflake, we decided to adopt uh a successful PC and

that gave us confidence to roll that out. We started small uh created a

out. We started small uh created a semantic view, couple of custom instructions and verified queries and we iterated it over to gain enough confidence and once we had that

confidence we decided to roll out in phases. So as the business users were

phases. So as the business users were testing it, uh we were continuously watching the audit log just to see how accurately it responds and we use that

as an opportunity to further refine our semantic view overall leading to very positive uh feedback from users. But

there was one limitation uh because there was no visualization support like charts and there was no option to export the results and we were planning to

build a custom web application using cortex and resty APIs but then came the game changer uh last year's uh snowflake summit >> game changer >> yes

>> yeah so snowflake intelligence announced as a private preview feature and that was the agentic a tool that we were aiming for As soon as it progressed to

GA, we decided to adopt. Now

transitioning from Cortex Analyst to uh Snowflake Intagious was very easy and straightforward. All we had to do was

straightforward. All we had to do was create an agent, configure Cortex Analyst as a tool to that agent and that's it, right? We were literally reusing the same semantic view that we had built before.

>> Yeah.

>> Yeah.

>> It's amazing. And by the way, what Kieran is doing is what what we advise everyone. Find small use cases and get

everyone. Find small use cases and get them to production. Don't wait for some big AI initiative. Just go deliver value and results briefly because we're tight on time. What's next for booking and

on time. What's next for booking and snowflake?

>> Sure. Uh so so far we have focused only on structured data. Now we have a huge amount of unstructured data which is set to be tapped. So we are planning to use cortex search for that and uh the other

one so far we were only seeking insights from snowflake intelligence. uh we want to take it to next level where we want to drive insights to action mode uh by

configuring the agents to perform some specific tasks and u we also have some plans to integrate this with the internal knowledge base and also we are

considering uh using CKE uh cortex knowledge extension and uh finally we want to adopt it to other teams and build more agents. Okay. Super exciting

what you're doing, Kieran. Appreciate

the partnership and I appreciate you being here.

>> Thank you.

>> Thank you, Kieran. [applause]

>> Okay, there is one last announcement that we want to share in this keynote this morning. Of course,

there are more sessions in a in a short bit, but anyone knows what the final announcement is? It was in the press

announcement is? It was in the press yesterday.

Most consequential AI company right now.

>> Open AAI. Someone said it. So yeah,

we're we're incredibly incredibly excited about the partnership that we announced yesterday with OpenAI because we are able to bring

the absolute best model in the industry with the security and governance of of Snowflake and the best way for us to give you color on the partnership but

also the state of AI. I want to bring an amazing friend onto the stage. the head

of enterprise and open AAI is here with us today. Please give a very loud

us today. Please give a very loud welcome to Ashley Kramer.

>> Happy to have you here.

>> Yeah, thanks for thanks for the invite to my favorite city in the world.

>> Yeah. And by the way, you come from a data background. You and I have

data background. You and I have connected for many years now. So you

understand AI and you understand data better than most people.

>> Love everything about it. I think the first time we ever connected I was building Tableau's cloud product and Snowflake was just coming to be a big enterprise data solution.

>> Yeah. So so this is a a re reunion for us. So I'm so so happy. Thank you for

us. So I'm so so happy. Thank you for being here. Let let's start maybe with

being here. Let let's start maybe with the transition of the use cases of AI. I

think maybe six 12 months ago everything was oh AI can retrieve and give you a chatbot experience but give you answers but now there's a little bit like oh it can make decisions and take actions for

you. Is that the general trend and and

you. Is that the general trend and and how are people thinking about agents taking actions?

>> Yeah this is this is great because now is the time that's exactly what we're seeing. We're seeing agents go or chat

seeing. We're seeing agents go or chat bots go from answers to true action. And

if you think about this in everyday life, I'll use a very good joint customerbooking.com.

customerbooking.com.

Uh no longer do you have to just go there and say find me a hotel in London.

I in the future with Aentic AI will say I want to go to London, pick the best month of the year, clearly not February

and um put me somewhere that has a great food scene. Go to bed, wake up, and the

food scene. Go to bed, wake up, and the entire trip can be booked and planned for me. And of course that's in your

for me. And of course that's in your real everyday life. How does that translate into enterprises which you all care about right now? We're at a moment where it is about trust not the

technology. The technology is there and

technology. The technology is there and there's responsibility on OpenAI's part, Snowflake's part and of course the customer's part. From open a OpenAI's

customer's part. From open a OpenAI's perspective, we always keep security, governance, privacy top of mind. We will

never train on enterprise data. We

comply with the EU AI act. We have data residency not just for EU but local and specific for UK as well. And so it's all

about having that trust. Our partnership

with Snowflake of course now allows us to bring govern data and the models together in a really secure way so you can be comfortable and trust the technology. And then there's

technology. And then there's accountability on the customer's part.

making sure you write have the right policies and procedures in place. Um, I

can build an agent to write all of my emails for me if I want once I trust it, but I can't do that on behalf of Christian obviously because I can't access his email.

>> That would not be good.

>> I that would be really fun actually. I I

wish I had that access. Um, and then also it's about observability and audit auditability. when it comes to

auditability. when it comes to enterprise agents, every step will be logged. So you can audit that, review,

logged. So you can audit that, review, understand the metrics, and iterate. So

there's really powerful ways to have this come to life in production in the enterprise, but you have to be thoughtful about it. The last thing I'll say is um you said it earlier, don't

start big. Don't introduce this agent

start big. Don't introduce this agent that's going to take care of a huge um action end to end. Start small. You

always start with humans doing the actions. Then you can move to humans

actions. Then you can move to humans maybe approving before it happens. And

then finally maybe the human just audits the outcome and fixes anything that needs to be fixed along the process.

>> So so to to make sure that the message is clear to folks that may have come in today with a little bit of apprehension or fear their agents are going to take over the world and we all watch

Terminator. I think you're saying no.

Terminator. I think you're saying no.

>> It's absolutely not. It's going to it's going to remove all of the mundane, boring things in our lives that we wake up and don't want to do and allow us to go do much more powerful things with our creative minds.

>> Awesome. Okay, let's talk a little bit about the differences we you see between adoption of AI here in Europe and maybe contrasting with the US which is where

both our companies are based out of.

>> Yeah. So, so regulation and compliance is top of mind no matter where in the world I travel. There are some differences and nuances that are really interesting that I've observed over the

past year. I'll start with the US. In

past year. I'll start with the US. In

the US, um, they're much more likely to jump into a pilot first. Uh, then what happens? More people get introduced to

happens? More people get introduced to the process. The organization is messy.

the process. The organization is messy.

There's several owners. In one case, a bank in New York, actually the CIO told me that it was um 17 committees that she

had to go through to get the pilot we did into production took over 6 months.

And so they in America tend to go into the approvals and the regulation at the end of the process, which can actually be highly frustrating because everybody gets excited about AI being introduced

and then they have to wait. When it

comes to Europe, a lot of that is upfront. That's actually not a bad thing

upfront. That's actually not a bad thing because you're bringing security and legal and IT and engineering all together upfront. Might take a little

together upfront. Might take a little bit longer, but then you can go from pilot to mission critical production much faster. And so from my perspective,

much faster. And so from my perspective, Europe is by no means slower or behind.

They just go about the process differently. And I think both sides

differently. And I think both sides could learn something from each other.

>> I fully agree with that. And you you mentioned compliance with the EU AI act.

I think at the heart of it, we're both our companies are very supportive and aligned with the principles which is explanability observability understandability because we want all of

you to leverage AI but do so understanding what's going on and having the right guard rails.

>> We do govern governance is everything and I I love with a data background. I

love the slide that data is the foundation for AI, which the opening was is now is here. And so once again, that's where the criticality of this

relationship really shines. Us being

able to move models closer within your secure data perimeter and really produce those powerful agents that can be trusted immediately um out of the box.

>> So super aligned on that. Okay, let's

talk about our partnership. We put out a press release yesterday. There's been

all sorts of positive feedback. Maybe

tell us a little bit more how you think about it.

>> Yeah, we're really excited for this relationship. We already have a bunch of

relationship. We already have a bunch of joint customers and hopefully future ones here in this room. And so when we think about it, when it comes to agents

and Gen AI, it's all about the context, the deep context that starts with data.

And in enterprises must be secure data.

So this partnership for us allows us to move faster together. It allows for customers to now work with real time data and not stale extracts. It allows

us to observe the security and compliance controls that already exist with Snowflake and really have a tighter feedback loop. When it comes to AI, it's

feedback loop. When it comes to AI, it's all about how you're training the models specific to your organization. And once

again, we never use your data to train our models. When it comes to our

our models. When it comes to our enterprise versions, our APIs, and our chat enterprise, we do not train on your data. That stays private to you. But

data. That stays private to you. But

context is everything. That feedback

loops makes it smarter and smarter. And

by having this tight relationship, we're able to move faster together.

>> Yeah. And and just to give you a color for everyone here in the room of truly the the the relationship and the partnership, a lot of our conversations and our emails are about compliance with

this regulation and ISO 9,000.

>> Oh, I've learned so many words throughout these last uh so many new controls we have to put in place together. But that's the beauty of

together. But that's the beauty of having this partnership.

>> Correct. And and that's what we we all offer and promise to you, which is we want you to adopt OpenAI's models, the the reasoning capabilities are through

the roof good, but do it with the safety and comfort that you know your data that context that Ashley is mentioning is um preserved and maintained.

>> Yeah, that's absolutely right. We're at

a we're at a critical moment right now where model capabilities are not the issue. It is moving so fast that

issue. It is moving so fast that enterprises can't keep up. There's a

huge value gap between what models can provide and the value that enterprises are actually extracting. And so we want to help remove those barriers to get you

to deploy faster. And um the the last thing I'll say on that is the momentum within Europe is huge. So outside of the

US, uh, we have UK, France, and Germany all as our top markets when it comes to our enterprise customers. So we're

already on the right track and I think this partnership will help us gain even more momentum together faster with all of you.

>> Okay, Ashley, I think uh people here get color on what we're trying to do together. Any parting words or final

together. Any parting words or final thoughts on either where partnership is going, where AI is going, where OpenAI is going? Yeah, we're really excited to

is going? Yeah, we're really excited to to really help transform everything when it comes to AI within an organization.

That is helping your teams become AI fluent. That is helping them automate

fluent. That is helping them automate processes via agents. And by the way, a big piece of that is not just taking the process you're doing today and automating it. It's reinventing the

automating it. It's reinventing the process along the way and then deeply infusing it in products or services both for your internal team and external. and

we're already on a great path together.

I cannot wait to see where we take Agentic AI next. And the last thing I'll say is AI is now. It is here. And so if you don't have this top of mind, you are

already behind. But we collectively are

already behind. But we collectively are together here to help you catch up and become a leader within your industry.

>> Yeah. So I want everyone to to to give it up for for Asley and and very sincere. Thank you for traveling, being

sincere. Thank you for traveling, being here with us.

>> Thank you.

>> Give it up.

[applause] Thank you so much.

Super super excited about what we're doing with Ashley and the broader team.

Okay, this is the the end of this initial session. We started with I

initial session. We started with I talked about three chapters and I said, "Oh, things that I wish you could do and that I I want to do." Hopefully based on what you have heard today, what you saw

dash demo, you walk out of this room saying, "Oh, I can do all of this today.

I can build without friction. I can

activate my data, my models, and I can build agents and agentic applications."

And at Snowflake, we're committed to delivering what we call the AI data cloud. It's the combination of great

cloud. It's the combination of great technology platform, but an amazing ecosystem of partners. OpenAI front and center of this relationship and how we think about Snowflake. Everything we do,

we want to make things easy for you. We

want to make it easier for you to connect within your organization with one another. And if the word governance

one another. And if the word governance and security was not said 50 times in the last 75 minutes, we didn't deliver the message because trust is the most

important thing of what we do. And we

want you all to leverage data, leverage AI, but with in a technology that you can trust. I end where I started. AI is

can trust. I end where I started. AI is

happening today. Ashley just said it. If

you are not leveraging AI, you are already behind. Your competitors are

already behind. Your competitors are likely chasing it. The benefits are here today. And I am incredibly excited about

today. And I am incredibly excited about what all of you will build with Snowflake, with OpenAI, and with the broad set of technology that you saw

today. Thank you so very much. Go build.

today. Thank you so very much. Go build.