All right. Welcome everyone. We're going to give a couple of minutes for everyone to filter in before we get started, but we're very happy to have you all here. All right. Welcome everyone to our joint Atlan and Google webinar. We have a lot in store for you today. So let's get started with me here today. I have Chandru, our director of product at Atlan, along with Lou, head of product at Google. Today, we're going to have an interactive session or just some housekeeping. Know that the webinar is being recorded and will be shared after. Use the chat for comments and questions at any time. And we're going to have a Q&A at the end.
We also have a couple of polls during this session that we'll ask that you interact with. Our agenda for today. So we're going to go into how Atlan and GCP work together for all of your data. Then we're going to have a live demo followed by Q&A. So before we get started, if you're able to drop your role in the chat. And before I hand it off to Chandru, we have a poll coming up. What best describes where your organization is with GCP data today? You should see a poll now.
We'll give it 30 seconds to a minute. And then I'll hand it over to Chandru and we'll get started. All right. Chandru, if you want to take it away.
Good morning. Good afternoon, everyone. Hello from Seattle. I can see people joining from London, Dallas, Mumbai, everywhere. So excited. So let's get started. Let's ground on a simple truth. Data isn't the bottleneck. The context is. This is the grounding truth here. You all in the enterprises are scaling your cloud footprint. In GCP, you can think of BigQuery data sets, more looker dashboards, pipelines, Spark, Composer, Cloud SQL, you name it. More AI agents being spun to query all of it. And yet the most AI experiments never make it to production. Hope that resonates with you.
It's not because the models are bad. The models are pretty good. They are remarkable and getting better every week, every day. The wall isn't the model. The wall is that the model doesn't know your business context. It doesn't know what revenue means, what a churn means, the definition of them in your arm. It doesn't know which table is trust or who owns that or what breaks if something changes. That's the AI context gap. Right there. That's the context gap. Experimentation on one side, production on the other. Context is what fills this chasm. So let's make this a little bit more concrete. Let's start with a simple question.
Who are our top customers this quarter? I hope this kind of like you have heard this in the past. You're, you know, someone in the leadership or in the business came and asked this question to you, right? Who are our top customers this quarter? Sounds easy. It isn't. The moment an AI agent tries to answer that, it needs to know a bunch of things. It needs to know who's asking finance or sales because they're different, top, slightly different. What does customer mean here? Is that an individual, a parent company or a partner?
What does even top mean here? Is it by volume of the number of orders or the net ACV margins? And how do we even calculate these different revenue cross margin discounts, refunds? And which table holds that number? Is it in, you know, BigQuery or CRM or both? Every one of those is a context. Miss one, the answer breaks. If it's familiar, just drop a chat like, you know, how many of your AI use cases have stalled because of questions exactly like this one?
This is the reality. And this is what Atlan is solving for. It's not a data problem. It's a context problem. Now, as we have worked with enterprises going through this AI transformation, I know there is a lot in this slide, but let's, let's, let's, let me demystify you. We have basically three challenges. Every time we are able to consolidate and every time talking with, you know, different enterprises, we are able to ground on these three different challenges. First, the cold start problem. What is it? It is a context bootstrapping problem.
Your business logic exists. Its definitions exist, but they're all scattered. It could be in your, you know, let's say your BigQuery or dashboards or Dataplex or Confluence talks or in Salesforce fields that someone wrote in a blog like three years ago, heads of your best analysts, getting all of that into one form that AI can actually use. That's the cold start problem. Hope that resonates with you all. And the second is the testing cycle. Now imagine the team builds a first version very quickly, right? First version is always quick. The analysts build it.
Then they spend months not knowing, is it good enough to ship? Not repeatable? Is it, could it be version? Is the definition aligned across the board? One wrong answer to CFO, the trust is completely gone, right? And the third problem, the pattern is the portability of this context. You're not running an agent, one agent tool. You're running in, could be in Gemini, could be in Lama or something else in next quarter. Every time changes, you know, all these models, you know, get matured and they release different versions every time. And every new tool needs the same context rebuilt from the scratch.
That's not a foundation, that's a tax. Right. And here's the underlying truth across all three. Context is scattered across your enterprise, and that needs portability. Not just for your cloud, but on-premises, SaaS tools, business apps, line of business application.
Could be from finance or marketing or ops that live in there every day. It could be near unstructured docs. Institutional knowledge is never ever written down. That's where meaning lives. That's where the real context gap hides. So what's, how are we solving? I'll just give you a deep dive. Let's read this diagram from the left. You start with the systems of truth. That's the data stack, the warehouses, lake, BI tools, governance and security of those, the metric catalogs, the quality, and then the whole GCPA state and everything beyond it. The on-prem systems, the SaaS, business apps.
Atlan integrates and pull all of that context into one place, which is our enterprise context layer, also known as enterprise context graph. And what lives inside this layer? Your tables and columns, descriptions, their lineage, their metrics, the technical metadata, operational metadata, the definitions, policies, classification, you name it, right? The owners, who used it, when used it, the knowledge tables. And what underneath all of it is what we call it as Atlan's metadata lake house, which is iceberg native, the foundation that makes context queryable, versionable, portable. So we basically have two motions on top of it.
One is our enrichment studio. If you remember the bootstrapping problem we spoke about, which is our AI-powered context creation and enrichment and already exists across your stack. We bootstrap natively with AI. The first 80% of your context is already ready before a human even, you know, look at it, write a single line of text. And then our context studio, which is our human in the loop and on the loop layer, includes the context engineering, generating semantics, the we use simulations, including evals and feedback loop. Your domain expert results conflicts, annotate edge cases, certify the production ready data sets, context stores.
And output of that context is used by our agents, the custom agents, the third party agents, the apps that is consumed via SQL or API or MCP or any other vector stores. That context portability, build the context layer. Once every agent in your enterprise, include your agents, Gemini agents, consume from the same grounded foundation. So with that, I would like to get us grounded, bridging into the next session. Today, we have Lou, I'm going to hand over to her. Dataplex is the native context store for GCP estate. This includes the BigQuery, Locker, AlloyDB, and you will hear from her more.
Atlan connect to it and that extends beyond it into your rest of the enterprise. So when your Gemini asks a question about your GCP data, it's not reasoning in a vacuum.
It's reasoning from the complete grounded context layer in Atlan. Lou, please take it over. Awesome. Thank you so much, Andrew. Do you mind jumping out of the presentation and reproject? There might have been a bit of a refresh. Sure. So in the meantime, hi everyone. Very nice to meet you today. Super excited to be meeting our joint customers and then, you know, sharing more about the better together story between Atlan and GCP. So my name is Lou. I'm a head of product at GCP within our agentic data cloud portfolio. Very excited to be sharing a bit more about, you know, where we come from, from a GCP perspective, as well as, you know, super excited to be sharing more about the integrations, the deeper partnership we have been building together with the Atlan team.
So starting from Google's data cloud stack here, as you can see, Google's data cloud is really built on GCP's unique technical infrastructure, which is powered by Google's latest AI and data intelligence, as well as strong investment on the infrastructure side. It is a converged platform for operational, analytical, as well as all of your AI systems. It's an open and flexible platform as well for meeting our customers with their multi-cloud and hybrid cloud strategies. It's also built for the agentic era, allowing the agents to be able to collaborate seamlessly, allowing customers to build agents and providing the context at the speed of the business.
So onto the next page, please. Now this is one page that double clicks a bit more into Google's unique planet scale technical infrastructure. So as you can see here, this is including our disaggregated compute and storage architecture, which can scale flexibly and independently. This also includes the innovative zero trust security approach, which we have pioneered that really is aimed to help our customers stay super secure. And then there's also the planet scale Google network, as you may be aware of. And all of those technical infrastructural innovations are aimed to help our customers stay secure and then really benefit from a state of the art, you know, system with the greatest cost efficiency, availability, performance, and scalability.
Now onto the next page. As Chengdu mentioned earlier, raw data, even if you provide that at scale, is not going to be enough to serve the need of our enterprises today. Now the real breakthrough would come from connecting the dots, right, from your own data as well as the latest innovations that Google can provide in providing this real-time and planetary scale understanding of the world. So there's a lot of additional, you know, intelligence that can be infused into your data workloads based out of Google's unique data and AI assets. So for example, our customers can get better consumer pulse data sets by connecting your data with Google's real-time signals from the market that is powered by the billions of search queries and trends that are happening on Google every day.
There's also real-world grounding that can be provided, enabling our customers to optimize, for example, logistics based on live traffic and maps information. And most importantly, we can offer the state-of-the-art large language models as well as a series of purposely trained AI models, including Gemini, other types of assets, to really make sure that, you know, those are directly built into your workloads, providing various ways for you to leverage those models, including building experiences to really help you solve your high value problems in the easiest possible way. So overall, you know, we're working very hard at GCP to make sure that we can help our customers to move towards a truly automated and intelligent enterprise.
Now coming back to the data portfolio, if you could move to the next page, if we could advance to the next page. Yeah, so this is again zooming back into the data analytics platform on GCP, which I know we have mostly data practitioners here, right?
So you're probably quite familiar with this stack on GCP. So overall, you know, it is built on this strong and unified foundation of universal runtime and storage, spanning across not only BigQuery native storage, but also optimized iceberg storage, as well as open storage on GCS for self-managed open formats. And then on top of that, we make sure that we have Dataplex, which provides the unified universal context in terms of metadata, so that we make sure that we aggregate AI-ready context metadata from first-party integrations, as well as integrations with our key partners, such as Atlan, and then making sure that we synthesize and curate the context for our customers, and then also make sure that this context is very easily retrievable, whether you're leveraging this through Google's first-party agents or through third-party DIY agents, which we already see a lot of our customers doing.
So that's the essential context fabric, which has been important even in earlier days. But nowadays, with agentic development, right, as Chen Zhu mentioned, this is really becoming the essential critical path. Now, on top of it, as you can see, within the data cloud portfolio, there's various ways to interact with this data, whether it's through BigQuery SQL engine or Dataproc or LODB third-party engines, and then various different customer-facing experiences, as well as catering to different personas. Now, most importantly, agent is one persona that we're doubling down on. Whether it is first-party agents or third-party agents, we make sure that the metadata context that you aggregate in Dataplex in conjunction with our partners are readily consumable and available through the various endpoints.
I think, without further ado, I would like to actually hand it over to Razzi to show you not only what it looks like in slide form, but also what is this integration that we have been working hard on look like in action?
Yeah. Thank you so much, Luke. As I'm handing it over to Razzi, just to ground us, what you will see in the next 15-20 minutes is the whole enchilada of this integration with ATLAN and GCP through Dataplex, the discovery, the lineage, governance, data products, the AI in-picture, the AI in-picture, how you can get the grounded context to start taking your agents from experimentation to production. That's the message that we want to land at the end of this 15-20 minutes. There you go, Razzi. All right. So, now on to our demo portion.
Before we get started, for this, I'd like to put yourselves in the shoes of a business user or an executive trying to answer a simple question. And the question that we're trying to answer today is, what is our best-selling product? With that in mind, we're going to jump into the demo to see how we build that foundational layer with the partnership between ATLAN and Google to help answer questions just like that, even for the most non-technical users.
So, with that, let's jump in. For those that are familiar with Dataplex, you'll recognize this UI. Here I am in Dataplex looking at our BigQuery Assets. And this is where I can add additional information, aspects, glossary terms, et cetera, on our BigQuery tables. So, let's jump into one. I'm going to jump into our DIM customer table here and show that we have added some custom aspects here. So, we have a custom aspect called governance. If I open this up, we can see that this table, I'm actually the steward for this table.
It has sensitive data, along with PII with a retention policy of three years. Now, we have all this information in Dataplex. How do we make this become active metadata that we can use outside of Dataplex, along with our ATLAN partnership and integration. So, let me jump into BigQuery here. In BigQuery, you can see that already on the DIM customer table that we were looking at in Dataplex. And if you look at the bottom right, you can see a little logo. It's the ATLAN logo. And if I open that up, what this does is actually brings all of the information that ATLAN has collected from Dataplex, as well as other sources, now directly in your workflow with BigQuery.
So, here you can see I actually have an announcement letting me know that this table is actually refreshed at 6:00 a.m. daily. But remember that aspect that we saw in Dataplex, the governance aspect? If I click into our Dataplex logo here, we can see that that integration with Dataplex, we're able to bring all that information into ATLAN to then make it usable in different locations as well. Not only do we show Dataplex information here, we can also show additional information like data quality tags. So, tags you can think of as PII, gold layer, et cetera, ownership information, and more.
This doesn't stop at just BigQuery. Let me jump into an example of Looker. So, here I am on a Looker dashboard, and you can see I have that same logo, same symbol, and I can get information about this specific dashboard, just like I saw with BigQuery. And this extends out to many other tools. So, here I have a look at the GCP as well as externally. Now, from here, I want to start to understand this dashboard, that table, where is this data actually coming from? And you're going to see on the right-hand side, I have a tab called Lineage.
If I jump into that, I'm able to then see, in this case, in table format or list format, where is it coming from upstream? Where is it heading downstream? But I actually want to look at this as a graph. So, let me jump into that. And with that, I'm going to jump into Atlan. In the middle of the screen, you're going to see that same DIM customer table. And now we can very quickly start to see, as I open this up, where is this data coming from? How is it being transformed?
As well as, where is it going? So, this one table, the DIM customer table in GCP or in BigQuery, actually originated in GCS, as well as some additional information from Salesforce. So, here, we're seeing not only the end-to-end lineage in the GCP framework, but as well, we're doing cross-system lineage outside of GCP, in this case, Salesforce. And as we go downstream, we can see that Looker dashboard that we were looking at, the customer analytics, as well as, there are two AI applications that exist outside of GCP using information from this table.
Not only does this lineage exist at the table level, but we actually have this lineage at a column level. We're able to see exactly what column is being used where, how it's being transferred, transformed, and moving through the ecosystem. Now, the question really comes up from here. Great. I can see all this information. How can we start to use this to answer that fundamental question that we asked earlier? What is our top selling product?
And again, with all the tools available through the partnership between Google Dataplex and Atlan, we can start to see this pop up with agent development. So, here, I'm going to jump into an agent that was developed using ADK. So, the agent development kit in Vertex. And it's been connected to all these services with Atlan, with Dataplex, et cetera. So, from here, I can start to ask natural language questions to help me answer questions, help me build reports. So, in this case, maybe I want to build a report that will help me answer that question.
I can say, what trusted tables can I use to build a customer report? And again, this is going to actually use the tools available to it. So, in this case, semantic search through both Dataplex and Atlan. And help me start to identify what tables can I actually use to build a report that will help answer that question. And from here, we're going to take this one step further to say, what if I don't want to even build a report? I just want to be able to answer a question. But here, you can see, it was actually able to pull up that same DIM customer table that we were looking at and letting us know that it's verified and part of the gold layer.
So, this is something that we can actually use to build that report. Now, we have all this foundation. We've taken information that was in Dataplex. We've taken information that was in Atlan, integrated it through MCPs into a chat tool.
So, how can now we operationalize this in production for an executive to ask that question that we posed earlier to get their answer? And for that, I'm going to jump into Looker. So, here you can see we have two agents in Looker, one without the integration and one with the integration. And for these agents, we're going to ask that exact same question. So, I'm going to go into the first one without the integration, and I'm going to ask, what is our top selling product? And as we saw earlier through the integrations, this agent is going to be able to call all the information available to it through whether it be Dataplex, whether it be other GCP services, whether it be Atlan, extending out that information into other sources like Salesforce, et cetera.
So, here you can see the one without the integration actually doesn't have all the context to be able to answer that question. What it did here is actually just did a count of rows in a table. Now, if I jump into the integration with, or the agent with the integration, I'm going to ask that exact same question. What is our best product? And you're going to see here that because this agent actually has all of that additional context that we were talking about earlier with that context layer, that context enterprise graph, it's going to start to understand what do we mean by best?
What do we mean by selling? What do we mean by product? Instead of doing a count of rows, it's going to interpret that correctly with all of that information that it has through Dataplex, through the glossary, through Atlan, and say, and say, this is actually our best selling product. And here you can see it was actually able to interpret this question correctly and pulled up the one that actually had the quantity that we sold the most. And this was due to, again, all of the context it was able to gain from the integration between Atlan and Google.
That is it for our demo. Before we move into the Q&A portion, we have one other poll for you, and it's going to pop up on your screen shortly. Are you heading to Google next this month? And we'll have that poll pop up, and then we can jump into Q&A. So again, please drop your questions into the chat, scan this QR code to book a demo with us here at Atlan, and we can take you through a more in-depth look at how we work with Google and the greater GCP stack, as well as how we activate all that information that you have available, not only within GCP, but outside of GCP as well.
