Sponsored by: Salesforce | Turn Databricks Data into Action with Trusted Context from Salesforce

Summary

Salesforce Data 360 uses Zero Copy access to connect directly to a Databricks lakehouse, allowing teams to unify customer profiles and power real-time personalization without moving or duplicating data.
This video demonstrates three practical use cases: building unified customer profiles from lakehouse data, grounding Agentforce AI agents in governed enterprise context, and activating data across sales, service, and marketing channels.
The core problem Data 360 solves is the missing 'trusted context layer'—nearly half of data and analytics leaders report that their AI agents and experiences lack unified, reliable customer intelligence because IT data systems and business data systems do not talk to each other.

Chapters

FAQs

What is Salesforce Data 360?

Salesforce Data 360 is a trusted context layer that connects directly to enterprise data platforms like Databricks using Zero Copy access, unifying customer profiles without duplicating data. This video positions it as the missing intelligence layer between IT data warehouses and the business systems where AI agents and customer experiences are powered.

What is Zero Copy access and why does it matter for Databricks integration?

Zero Copy access means Salesforce Data 360 queries and federates data directly from the Databricks lakehouse without replicating or moving it to another system. This preserves the single source of truth in Databricks, eliminates data drift, and reduces infrastructure overhead while giving Salesforce applications real-time access to governed lakehouse data.

How does Data 360 ground Agentforce agents in trusted enterprise data?

Agentforce agents are connected to unified customer profiles and governed data stored in the Databricks lakehouse through Data 360's trusted context layer, ensuring agents act on accurate, current information rather than stale or fragmented records. This video shows a live demo of an Agentforce agent using Databricks data to handle a maintenance use case with reliable context.

What is the 'missing context layer' problem in enterprise AI?

This video cites that nearly 50% of data and analytics leaders report that their AI agents and customer experiences lack complete context because IT teams manage data in warehouses and lakes while business teams manage data in CRM and operational systems that do not communicate. Data 360 bridges this gap by creating a unified, governed context layer connecting both worlds.

Full transcript

[00:09] Okay, good afternoon everyone. Uh we're going to get started over here. Uh thank you so much for taking out 40 minutes of your Data Bricks Summit over here for us and joining us here to talk more about how Data Bricks gets integrated with Salesforce and how Data
[00:22] Salesforce helps making more value out of Data Bricks data as well. Uh I'm Kunal Goel. I'm on the Data 360 product team and I'm joined by >> I'm Bobby Brill. Hopefully you're all here to see you know how Salesforce and
[00:35] Data Bricks play well together. >> Yes. And uh standard looking standard uh warehouse, I think standard work. We have a forward-looking statement here. Please make all purchasing decisions based on
[00:48] what is currently available. Now, first of all, before we begin, I would really like to thank you all for joining us here for your time as well as for trusted partners and customers who have been helping us in Salesforce
[01:02] journey and within the being part of the Salesforce ecosystem over the past couple of years. With that, how many of you are have invested heavily in data lakes, data pipelines?
[01:15] Any show of hands? How many are using I'm guessing everybody's using Data Bricks. Now, keep your hands raised if your business leaders are still telling you that your agents or your experiences
[01:27] don't have the entire context. Okay, quite a few. I mean, that's not new. I believe that is a very common problem and the problem is not the data, the problem isn't that you don't have enough pipelines or you
[01:40] don't have enough integrations. It's that we don't have intelligence layer building a trusted context for our customers today. And that's not just for you. In fact, almost 50% of data and analytics leaders
[01:53] across the all organizations have the same problem where you have your IT teams and your data scientists and all your analysts sitting in IT warehouses lakes where they have their set of data
[02:06] and your businesses having data in very different systems all together where their data is sitting and not talking to each other. And that's where there is a missing unified context layer. And this is the gap that we try to fill in.
[02:19] And the gap gets filled where we have a product called Salesforce Data 360. We pioneered this 5 years ago with the inception of customer data platforms and this is not
[02:32] just a data platform or yet another integration tool that is there in the industry. This is the trusted context layer if you may which is unified, governed and has an
[02:45] intelligence layer which helps connect your businesses to your IT warehouses and lakes like Databricks as well which allows you to get more value out of the data that is already sitting in data
[02:57] lakes without moving it. And this is where we have something called zero copy integration between data lakes like Databricks and other lakes in the ecosystem where you don't have to
[03:10] reinvest in those data integrations all over again. Your data can sit wherever it is today and you can still integrate it and create that unified context layer with the help of Data 360. Data 360 allows
[03:24] you to act activate on your data. It may very well be stored in multiple different locations and helps you create business outcomes, makes your agents much more smarter at the same time.
[03:37] Now, how what what does this look like overall as an agentic enterprise? A lot of our customers today are on the journey to make themselves an agentic enterprise and this is the Salesforce
[03:49] way of the entire enterprise architecture that we have today. At the bottom, if you think uh see, there is a data 360 layer, which is where we bring in all the context, where the secret
[04:02] sauce lies, where we have created additional capabilities, which allows you to tune your raw data sitting in different places, harmonize it together, unify it at the same time to create a trusted
[04:15] context so that your app layer, now which Salesforce is most known for, our sales cloud, our service cloud, marketing cloud, commerce, our app layer in many ways, or our system of record,
[04:27] can act a lot better with the unified knowledge created together and unified trusted context created at one place. On top of it, now we also have Agent Force, which is the AI layer. You can create any sort of agents to power any
[04:40] of your apps. Slack as an interactive agent, Tableau, and headless 360. Now, headless 360 is something new that we announced a few months ago, which is nothing but
[04:52] a set of MCP servers and skills that Salesforce is creating, opening up all our ecosystem, allowing you to integrate with that MCP server and headless way in a headless way across all the functionality of data 360.
[05:06] It's as simple as you going and talking to an cloud agent, trying to configure Salesforce, and getting the best out of it with the trusted context created in data 360 and the apps built on top of
[05:18] it. Now, you may wonder what does data 360 have and what all different parts of journey does it have today? Now, within uh we have five jobs to be done within data 360. First is connect any data.
[05:33] You can bring in all your structured, unstructured data through different ways into data 360. Now, this is where you can bring in through batch, streaming, real time, in case you want to ingest directly from your source, and
[05:46] store data into Data 360 directly. Or you can work with zero copy connections with data bricks and our other lake houses in the industry where you can bring data
[05:59] only at run time when you need it via zero copy integrations which we'll talk more about later. That's first part. Number two is where you prepare this data to make it ready for
[06:12] your app layers and for your businesses. This is where you unify it, you harmonize it. Both structured and unstructured data needs to be harmonized. You For structured data, it's very simple. Different schemas, we need to
[06:24] harmonize it into one common data model. But even for unstructured data, you may have data your PDFs, conflicting knowledge bases in sitting in three different locations
[06:38] which now need to be harmonized. At the same time, you need to have a knowledge graph created on top of it which tells you that, "Hey, these three documents are related to each other." So that when your agent is trying to find the right information, it's able to find it
[06:52] at speed with right accuracy as well. So that's all of harmonization and unifying capabilities that we have as job number two. Job number three is very important. You need All of this is built on a
[07:05] trusted governed layer in Salesforce metadata platform. You need to create your tags, your policies, and attach those policies to give your users access to that trusted context that you've created in Data 360.
[07:19] This trusted context is and the governance layer is built to scale as your data comes in. So we have AI-based tagging and classification capabilities built in. So if you define your logic as
[07:31] new data comes in or any new data comes via zero copy also, those policies will automatically act upon it. Fourth is us enriching all of this through either creating new models,
[07:43] we also support bring your own model, creating new segments, actions, searches. Basically, allowing you to tune your data as well as store it in indexes,
[07:55] which allows your agents and your app layer to act upon this in real-time way at speed with precision through a governed layer. That's That's These are all the things that Data 360 helps you build on top of
[08:08] your data. And last but not the least is the capabilities that we have to activate all of that trusted context that you've built in in one place across different ecosystems, whether that may be ad tech
[08:21] or your marketing cloud campaigns that you may want to run for your marketing journeys or integrating it back into your CRM and other transactional systems where your business logic actually runs.
[08:34] Not only that, once your trusted context has been created in Data 360, you can easily share that back to your user sitting in Databricks as well or any other data lake. So, it's the connection between
[08:47] Databricks and Salesforce via zero copy integration is bidirectional. Anything that you have built in Databricks can be leveraged by your Salesforce users. Plus, along with
[08:59] when collaborated along with everything that you built in Data 360, it creates a trusted context. Once this trusted context is created, it can be shared back with your Databricks users as well. So, your analytical users, your data
[09:12] scientists can access that unified customer's profile that is stored in Data 360 as well without you moving any data. Now, how does that all work? This is where we pioneered something
[09:25] called zero copy architecture. Salesforce started a zero copy network between all the big vendors 5 years ago. And we have query federation and file federation here. This allows you to
[09:37] unify your data among sitting in different investments that you might have already made without actually moving the data across different systems. Your data can sit wherever they are. So, your data might be sitting in Iceberg tables in Databricks
[09:51] or in any other data lake vendor. And your data you have made an investment in Data 360 as well. These two systems can simply talk to each other with us able to do metadata syncing and you can apply your own
[10:03] governance policies on top of it as well. We pioneered initially with query federation and then we also added file federation. Currently, file federation is already available for open catalog of
[10:16] Databricks. We are also adding support for Horizon catalog for any customers who might have moved to Horizon catalog. Now, how does this all work? All of this works behind the scenes with
[10:28] this architecture. Now, if you see on top, there is query federation where the compute is going to be happening on the Databricks side. This is what we started with few years back where anytime
[10:40] anybody wants to query the data or wants to query data from Salesforce ecosystem, we will run in federated query behind the scenes going to the Databricks source and we
[10:52] will query the records from there and bring it into Data 360. With that, we also added push down logic where complex queries that you may be creating with additional joints and etc., all of that complex logic will get
[11:04] computed on the Databricks side itself. And we will only retrieve the records that are the results of those queries back into Salesforce ecosystem to do the necessary logic or to power your agents.
[11:18] On top of that, in future, we also added then cached acceleration on top of those queries. So, you can cache some of those query federation results into Salesforce as well in order to get faster time to value to your user journeys.
[11:33] And then followed by something that we are really really proud of is file federation. Now, Salesforce pioneered this 2 years ago with starting with other data lake partners. And we in fact convinced
[11:46] Databricks to onboard onto Iceberg table as well because Iceberg is the format that allows us to do file federation across multiple different zero copy partners of ours. So, they now support our Iceberg also as
[11:58] of 2 years ago. Uh which was something that Salesforce initiated with them. And with the help of file federation, we now have direct access to those files anytime you are trying to read any data,
[12:11] we will run those queries, run the compute of us pulling the right data from those files in Salesforce ecosystem itself. So, this allows you to get rid of two compute systems and only do the
[12:23] uh the compute of any pulling of files in Salesforce itself. Last but not the least is data sharing back from our Iceberg tables into Databricks as well. Anything that has
[12:36] been created in Salesforce is easily accessible by your Databricks users as well that might be sitting there. Often, we have heard months that here engineering teams, integration teams have spent sprints trying to integrate
[12:49] these two systems together by copying data and that becomes stale very often as well. So, zero copy architecture both with file federation, query federation to bring data from Databricks into Salesforce ecosystem and data sharing
[13:03] using data sharing taking it back into the other system is really a pioneering is a revolutionary technology that we have been using. Many of our customers are already using this. For example,
[13:16] Indeed had a lot of data sitting in Indeed, FedEx have a lot of data sitting in Databricks. They also have a massive Salesforce investment as of today. And they integrate using zero copy
[13:28] bringing these two systems together. A little bit more. So, talking about what query federation is and file federation is. For query federation today,
[13:42] uh there are you can absolutely use it to create your new segments in data 360, to run your precision queries which are not going to be running too often. You can use query
[13:54] federation itself to bring that data into Salesforce ecosystem. However, when you are running petabytes of scale you are trying to query petabytes data and that's a lot of data to query
[14:08] from one system to another system. That's where you hit some limits as well. And this is where file federations come comes into play where you can run queries at any number of scale.
[14:20] You can at the same time there is only one compute engine behind the scenes. So, instead of running compute on two different compute layers where the query is running on in query federation the query is running on Databricks side
[14:34] while the output is being handled on the Salesforce side. With file federation you can get rid of two compute layers and all of that can be consolidated in the source system in the target system where you're trying to pull data from.
[14:47] So, file federation is definitely a pioneering technology. We would highly suggest you to use that whenever moving any data between the two systems. But when you're trying to just run few queries with precision,
[14:59] query federation can definitely do that as well. And you don't have to make the choice right away. Anytime you can choose both, you can choose one or the other for your different use cases as well. But the idea is that you have both options
[15:12] available and you can select use both with federated auth as well which allows you to create these connections in a more secure way. With that I'll hand it over to Bobby to give you a nice demo.
[15:24] >> Awesome. Thank you, Ken. All right.
[15:36] recorded here. Uh there's a couple things that I may have, you know, hand-waved over. I'll talk through that. Something goes wrong, here, you know, I'm fine with that. Uh okay. So, how many people in this room are uh have you
[15:48] know, have to maintain a Salesforce instance as well as Databricks? All right. And usually the reason for that is Salesforce is going to be that uh that application layer. It's got sales, service, marketing, commerce, all
[16:02] those uh different layers where the business lives. And your Databricks is your data lake. You're connecting a lot of data there. Um but if you're a service agent, and I'm not talking
[16:14] about, you know, uh machines, I'm talking about actual humans that have to uh work cases, cuz they still exist today. Uh they're going to they're going to be in uh they potentially in Salesforce Service Cloud. So, in this
[16:27] demo, um it I'm going to play the role of a service agent, and the service agent uh is looking at some cases. This is a B2B company. Uh we build machines for other,
[16:39] you know, businesses or or um industrial plants to run. And so, I have a case here that's talking about the AP7100. They need maintenance. So, there's a couple things I'm going to show you
[16:52] here. So, the person logging the case is someone called Leanne Tomlin. So, on the left here, I get Leanne Tomlin's contact information. Most of the time that lives in Salesforce. Uh but there's a couple
[17:04] things here that it's pulling into that uh contact. I can see Leanne's training and certifications. That data does not live in Salesforce. In fact, Leanne Tomlin is uh her her her profile exists
[17:18] in the training certification database. All that data is going into Databricks. So, I've got my contact in Salesforce. She's logging cases. I got my training and cert information. All that lives in in Databricks. And I also have some
[17:32] engagement. I know where Leanne's on, you know, our website. I know kind of the things that she's doing, calls that she's having. All these things are engagement of Leanne. And the depending on what system is is tracking that, I
[17:45] could have maybe 10 Leannes. So, one of the most important pieces when you're trying to build that kind of context for agents, you want to unify the profile. So, I only want one instance of Leanne,
[17:57] but I want to know everything that Leanne is doing. So, Data 360's identity resolution allows me to do that. So, I've created a unified view of Leanne. And then I've pulled up Leanne's
[18:09] engagement. I've aggregated that so that I can get the view that I need for my service reps. So, all this information is important for them when they're working with a customer to kind of know that information. Additionally, I've got some asset health
[18:22] from Omega systems. So, this is where Leanne works. And be because she logged the case, I can actually surface, this is just doing a federated query into Databricks, on the the asset health.
[18:34] These things are, you know, important. So, I have additional context. Again, this is for a human agent to kind of look at. And then lastly, all of the equipment that they have, all these assets, they've got a lot of signal. And
[18:46] so, I can actually query some of the, you know, last last signal from these systems and I can show it right here. Again, if this person's talking with a customer, these are all things that that person's, instead of having to ask
[18:59] Leanne, we can actually see it right here for the service rep. So, it bringing that stuff into the service console is very important. And I don't want to copy that into Salesforce to do it. I just want to go federate that
[19:11] query out to make it more we want to keep data where it lives. But I also have Agent Force and I can ask Agent Force some some more pointed questions here like I have this AP7100.
[19:23] It's got some maintenance requests. So I might ask what are some of the main issues with AP7100? So if I ask this question,
[19:37] what we want to do is actually it's we call it data Q&A. So I'm asking a question about my data and I want to get an answer. So it's sort of this analytics. And so what we built is the ability to we have we have the
[19:50] metadata. So we're able to connect your data. We understand the metadata and we can actually do natural language to answer. So in this case this is where I kind of hand wave
[20:02] cuz it would have taken a little bit longer. So I hardcoded this. But what we have is the ability to do natural language to answer. And so what I got here is 78% of recorded anomalies in the
[20:14] last two years have been bearing wear. Almost all pumps past 80,000 service hours. Well, it's also giving me a recommendation. It's saying that AP7200 is performing better. All of this is done by
[20:27] by doing some sequel and analyzing the results and it's just passing me the results back here to answer my question. So I've got some information about the AP7200. I should uh
[20:39] it says it says it's performing better. But I might ask a question Can you give me more details on the AP7200 product? So I'm going to ask this question about
[20:52] the product. What I've actually done is I took my product catalog that lives in Databricks and I built a search retriever in in Data 360. So that search retriever is allowing me to do
[21:06] semantic search. So I can you know, it's basically giving me knowledge from this the product catalog I've got a lot of like descriptions and and all kinds of things. So, it's it's taking all that data and it's allowing me to summarize
[21:18] it here. I need my search index here because it's giving me just the key the key information, some specifications, and then it's also telling me what certifications required. So then, the last thing I'm going to ask here is is anyone in
[21:33] Omega Systems certified to use this. So, before I recommend this product to Leanne or her company, I want to understand are they actually
[21:45] certified for it? So, I ask the question again, we have all that certification information. It's zero copy federated query in I can I can answer the question and I can see that some of their employees are
[21:58] certified with the first of the fundamentals, but nobody has that advanced diagnostic. So now, the the agent human can now recommend potential training to
[22:10] to Leanne and her company in order to upgrade to this this path. So, this is where it's not a simple request. She's not just going to go to a website and ask an agent. You need more expertise to to kind of go help. So, that's where
[22:23] we're giving that context to that that human to be able to make those decisions. So, this is a use case. Like I said, some of these things would actually technically take a little bit longer. I don't want to waste the time here, so I hardcoded them. But all this
[22:35] stuff is doable today. So, how did I do this? So, I'm going to I'm going to flip into the data data cloud data 360 application in Salesforce where I start with my data stream. So,
[22:47] if you remember what Kunal shared, it starts with connecting the data. So, we've got a bunch of connectors. I I already have things already connected here, but there's a couple of things that would be interesting to to everyone. So, all of that interaction
[22:59] with Agent Force, all of those logs are already being written into Data 360. So, we have these connectors predefined. There's a lot we call the AI platform connectors. So, I got a bunch of agent
[23:11] stuff. I've got, of course, the things I brought in from Databricks. We've got some ingestion just from other signals that we needed for various applications. And then I've got all of my Salesforce CRM objects. So, I don't
[23:24] know how many of you are trying to get Salesforce data into Databricks today, but you might be thinking, "Hey, this is a I can't do this real time. We can't federate that query because Salesforce runs on something called SOQL that literally no
[23:38] one really understands why why that has to happen." It means that you can't federate that query. But when we sync that data within Data 360, you now get that ability to zero copy Salesforce
[23:50] data. So, Data 360's providing you that that capability because unfortunately, we had to move the data. We don't like to move data, but we had to to make it SQL ready. And you get that, you know, federated query or
[24:02] federated file directly to Salesforce through these connectors. So, I just want to call that out when we show you kind of zero copy back end. So, I've got some connectors. I have my
[24:14] What do I have here? I've got some customer profile. I have their certification information. I have some lead information, but how do I create this? So, it all starts from creating a new data stream. So, if I'm going to pull in another table,
[24:27] I choose my, you know, I've already told them I want Databricks. And so, I have table in main. You're going to see how long this takes, and then I'm going to show you a really
[24:39] cool thing that we just did. I want it to be zero copy, and then I can choose my table in here. I'm not going to do anything here, but notice I have my training catalog. I haven't brought this in. That's going to come up later. I'm going to show you what that means. But I I bring in my my connector,
[24:52] and then I have data. We have data at two levels. We have what's called the data lake object. It's a virtual when you're doing zero copy, it's virtually kind of pointing to that data, but we've created a container, so we know the metadata. And then we have this concept
[25:06] of a data model object. This is where that harmonization happens. This is how I can get data from multiple streams, and I can point them all to a single data model object. So if I've got contacts from multiple places, I can get
[25:20] that into a you know, a single object. So we remember back when, you know, Leanne's in three different systems, well, I created I have a single um individual object where all those Leanne's live. So the next thing I want
[25:34] to do is actually unify them. So I'm not going to go through the process of what unifying profiles look like, but I've I've already created this. I've created a unified contact view. So I've got all my contacts in Salesforce, I've got all
[25:47] my profiles coming from Databricks, and I've unified them. I've I've created some some rules, so some fuzzy matching, so that when it looks at Leanne and her certifications that sort of live with
[25:59] her, all of that stuff is is pulled into the same unified ID, so that I can identify one Leanne. That means if she's done certifications at another company, and she changed companies, it doesn't
[26:11] matter. I've got all that information cuz those we care about who has the certification, not what company they were at. So I can do this. So this is where unifying the ID is very important to build that unified view of the
[26:23] customer. And then we show I showed you how I asked the question about my product catalog. How did I do that? I created this search index. So this search index
[26:36] is a I took all of the the product catalog from from Databricks, and I chunked it. So I created chunks for my semantic search because I needed to power my my agent force. So this this search indexing
[26:50] works on that's we can we think of that as semi-structured data. It's like long text fields, descriptions, but this also works on unstructured data, PDFs, HTML files, connected to like a file catalog, and we
[27:02] can pull all that in and we can index it and now we have that knowledge for for our agents. So, I've created this search index and then we have this retriever on top of it. So, this is the rag pipeline and across all of the areas we have the
[27:16] ability to test these things. So, I asked something about the 80 the AP 7200 product. So, I can just show you from like how you would bring this in and test it before actually giving to the agent. I could say,
[27:29] tell me about the AP 7200 product. And what I get here is it's going through the search index and it's showing me based on the semantic
[27:41] query, what are the files that it's sorry, the chunks that it's going to the pull up to answer the question. So, I can I can see the information of the chunks. I can see the score. So, this lets me further tune how my rag pipeline
[27:54] is working. Again, this is building a more trusted context for that that agent layer. All right. So, I showed you how to what this looks like within data 360, but we just announced
[28:06] headless 360. So, what is headless 360? I'm going to go into my code. So, anyone that anyone here using VS Code to do Salesforce development development today? So, if you're a Salesforce developer,
[28:19] this is like one of the best IDEs to do that in. There's a lot of capabilities through Salesforce CLI. I can you know, do my I can write some code. I can use CLI to to deploy that to you
[28:31] know, my sandbox or whatever instance. Well, I've got Claude code here and remember when we talked about the the catalog data. So, I sorry, not catalog data, my training data. I had
[28:44] training data. I want to be able to build a retriever for that. So, I'm just going to ask Claude to help me build a search index based on Databricks data.
[28:58] Uh actually, search index on Let's tell it a little more context. Search index on my ca- training catalog
[29:11] in Databricks. All right. So, what's happening here? There's a Data 360 MCP server running. It's already connected to my instance. I don't want to walk you through what that looks like. There's a little bit of setup. I've already done that. And what this this found is a skill, a
[29:26] skill to create a search index from external. So, all of our the folks in the the PM org were building skills to help our customers do these things very fast. So, it's going to ask me some questions. So, first off, it found my connection. I
[29:39] already have a connection. It's not going to ask me what connection cuz there's only one that I have access to, so it doesn't have to figure that out. In fact, it found another one that's in air state, so it knows not to use that cuz it's smart enough. And it's going to ask me a question,
[29:51] "Where Where is the data?" Well, if you remember my demo I showed you I went to main.default, so I'm just going to tell it where that table lives. It's going to ask me a few questions, then I'm going to kind of set it and forget it. Um
[30:04] and so it it's looking for the table. So, I told it Look, it found my training catalog, and it's not in use. That means I haven't connected it yet. So, now it's going to figure out all the things that
[30:16] it needs to do in order to create the search index. So, this is thanks to all of our APIs and the MCP server. So, now it's it's got a set of tasks here. So, it's going to first
[30:28] figure out the data, create my federated data lake object, then it's going to map it to my data model object. It's going to do my field mapping and create the search index. It's going to do all this for me. You saw in the application there
[30:41] were some pages I had to, you know, I'm clicking through. Some of the things take some time to load. This is just going to do it. So, while this is working, I'm going to go back to my demo and let's assume that by the end of this
[30:54] demo all that's done. Cross my fingers. Again, this is a live demo. So, I'm going to go into the next piece that we want to show you and that was zero copy back out. So, how does a zero copy back
[31:06] out work? I can create what's called the data share target. So, I already have some created but let me show you from scratch. I can create a I want to share data back to data bricks. And from here I give it a I give it a
[31:19] label, API name, what's my data bricks account? On the data brick side I'm going to create a connector. But the important piece here is I get this core tenant ID and then the endpoint. Data bricks actually needs
[31:32] this in order to connect to be able to do that file federation if I wanted to do that. So, we can leave this up cuz there's two parts to data share. First I've got my target and then I can actually choose what tables are shared
[31:45] specifically with with data bricks. So, this is important. You don't necessarily want everything there but in this case I've I've chosen everything but if you have zero copy in I don't think you want to share that back out necessarily that doesn't make sense. But
[31:58] some of the important things that I might want to share out all of my salesforce data because I want to I want to get that zero copy back into data bricks. My unified profile. So, this is actually created in salesforce it's
[32:10] based on that zero copy data from from data bricks but it is my unified IDs and all that linkage to things. That is important for identifying that single Liam. I can also I can also create some
[32:24] insights. So, some of those aggregations that I saw we call call it calculated insights. So, we're basically materializing, you know, aggregations on data for use of like faster queries. So, I've
[32:36] got some insights here that I can also share back out if I want to. So, all this data is just shared back out to Data Bricks. So, now let me go into Data Bricks. First thing you're going to do to access that data is create a connection.
[32:50] So, I'm going to give it a name. Let's call it zero copy. And there's two types of connectors. So, if you're doing federated query, you
[33:02] choose the Salesforce Data Cloud connector. Or if you want to do file federation, you can choose the file federation. I'm not going to go through the full thing here. But, this is where I would copy in my core tenant. So, I can go back here.
[33:16] Here, I can copy my things. Just kind of giving the other one a little bit of time to create. It's actually really fast, surprisingly. And I can put my tenant ID. And then I can get my
[33:29] endpoint. So, I've got all this stuff. And then I can basically, you know, what's my catalog name and and where do I want to store
[33:42] things? I can set all this stuff up. I'm not going to finish that cuz there's going to be an error because I ran into some problem that I haven't solved. But, this is generally how you set up the file fed. If it's query fed, you set up
[33:54] a connected app in Salesforce. It's all documented out there. But, what I get is, let me go back to my catalog. I've got my data coming from Salesforce. So, this is
[34:06] kind of all that data. And let's just query something here. So, I actually have my GenAI generation. So, when Agent Force generates a response, we store all that stuff in this generated
[34:19] GenAI generation table back in Data 360. And I can run a query on top of that. I'm in Data Bricks and zero copy. I can kind of see what was passed in. I can see the response. If I need to do this
[34:31] in any sort of observability platform or whatever and all that's already connected to Databricks, great. You've now connected your data there. All right, so let's go back to Claude and see where Claude's at. Claude is in the process of creating
[34:43] that search index. So, we're just going to sit here for a minute while it finishes this. Looks like it's done. I can't believe the timing on that worked out so perfectly. But, I've created again, started from I didn't have anything connected,
[34:56] connected it DLO, DMO, mapping. If you've used Data 360, some of those things are maybe a little bit slower than they need to be. We're also working on that. But, with headless, I was able to do that with even not even clicking into any UI.
[35:10] Just let my my agent go do it. So, it created everything. Let's see here. This is my page. I don't see the product. I don't see the sorry, the training catalog. I'm just going to refresh. Fingers crossed we see training. Oh,
[35:23] look at that. Here's my training catalog. So, the next step I would see here is it's it's being submitted to be able to create the search index. So, I basically just went from zero to a rag pipeline in
[35:35] what felt like, I don't know, maybe 5 minutes. Whereas, in the UI, I would have had to read documentation and all these things even figure out. Thanks to headless 360, you're able to do that so much faster.
[35:48] All right, with that, I think did we have we had some finish we had some other slides here. Let's go back into the slides. So, kind of um taking this to the to summarizing what
[36:00] we did, you know, you take your data from Databricks into Data 360 to turn it into that action. Whether it's within marketing for personalized real-time experiences, agent force, we showed you that trusted context, or even
[36:14] just building that 360 view of Leanne. All of that is done thanks to Data 360. Um we didn't talk I didn't show you this in the
[36:26] in the demo, but we have this whole real-time layer. So, I showed you what the data layer looks like, but we are able to capture signals in real time. And I when we say real time, we mean sub-second real time, where we can get
[36:38] the data directly from the website, clicking in so that we can we can take that to action. So, we have the ability that's where that ingest API can can get that information. So, these are just some examples on how we're using the real time.
[36:51] Um some more examples on how we're bringing the uh the trusted contacts to agents. So, we do it across all all industries, all kinds of use cases. Uh building that knowledge base is is usually where
[37:04] uh we see our customers they need they have that data, they just want to get that knowledge uh to those agents. And then building that customer 360. So, these are a couple of our customer stories
[37:17] uh across the travel and hospitality and health and life science. So, we've got some um some cool ways we're we're doing that today. And you know, how you do it? You build that unified profile. So,
[37:31] there's nothing at least you know, Databricks has some other partners they work with for uh identity resolution. Salesforce is one of them, so you Salesforce for that identity resolution. Uh and then activate across whatever channel that you want.
[37:43] >> And just adding to it, like what you saw is all real. In fact, 60% of all data that is getting ingested and processed through Data 360 is already getting coming via zero copy with one of the
[37:55] vendors. Right? 60% So, that's terabytes and terabytes of data that is getting processed via zero copy, and our customers are actively able to leverage any of their data investments that they
[38:07] may have made across any of the data lakes in the industry through zero copy and creating business value today. So, don't wait, and you can really give your business users a lot more access, and
[38:20] get whatever access data that is sitting in Salesforce CRM or in transaction systems via zero copy back and bring it back to your data bricks users as well.
[38:34] So with that I mean we'll end it up. We are at booth 104 as well and Bobby and I both will hang around over here in case you have any questions. So feel free to reach out to us and we would be happy to help you in any way possible and give you more information about this. Thank
[38:46] you.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.