AI in Tableau: Future-Proof Your Analytics Platform | Dreamforce 2025

Okay, we're ready to start and thank you for joining this afternoon in our aiintableau future proof your analytics platform session. Before we jump in, just a quick reminder to please make your purchase decision in what is available today. And also big thank you. I know all thank you to our customers, partners, trailblazers. I know it's a day full of sessions and content and presentations so we really appreciate you taking the time and we hope to make this a valuable use of your time. A quick reminder, our tableau keynote happened today, this morning. It was a great session but if you didn't have opportunity to go, it is available on Salesforce for you to watch at your convenience. With that, my name is Jessica Morguia, I'm in product marketing and here with me is Joe Constantino. He's, he is in product management and he leads all the integrations of MCP across our tableau portfolio. So we are very excited to have him here today in the presentation and with that let's dive in. So you have seen that AI has been evolving many industries and analytics is no exception. We have gone from that traditional business intelligence where it's more of a manual work and it's report based process, all the way to AI assisted analytics, right. Where we have a very helpful copilot that, that help us to expedite our work and even recommend those actions. But it's still the humans who are actually needs to take the action. And today AI is moving from the assistive AI to agentic analytics. Here is where it's acting autonomously. So here is no longer the questions of how can my AI help me, but what can my AI do for me, right? So we are improving, we are trying to provide all, all the tools that organizations need to have more proactive and efficient data driven decisions. But no matter where you are in this agentic analytic journey, we are injecting innovations across the portfolio to help you win in this AI era. So let me talk to you about our AI strategy. Our AI strategy is twofold. Right on your left, we are infusing AI across the whole analytics experience. So basically it's built in intelligence directly in tableau. On the other side, we are enabling you to take the Tableau's analytical engine and connect it to your custom AI solutions. So we're giving you that flexibility but you know, it's the tableau's analytics that are running behind and we're going to cover these two strategies across the presentation. But let's start with AI in tableau and here within the experience Of Tableau, you're going to see that we're infusing AI across every stage of the analytics journey. To explain you how we're doing that, let me take one step back and you're going to see that we have three main products. Tableau Cloud, Tableau Server and Tableau Nix. For Tableau Server and Cloud, the AI experience is assistive. Here we are giving to your business users and your analysts that copilot that help them to expedite their work. But in the case of Server, they can connect it to their own LLM model so they can help them to prepare and visualize the data in their security boundary. In the case of Tablet Cloud, they are going to have the exact same power, but it's fully hosted in Cloud. Plus you are going to have the power and the back of the Agent Force Trust layer. You're also going to have personalized insights and contextual insights for your business users in their flow of work so they can start acting on those insights. Now we have Tableau next Tableau. Next is the evolution of Tableau. Here is where you are still going to have the exact same conversational experience that you have with Server and Cloud, but now we're enriching our agents with business context so they can all not only help you answer questions, but they can also act on your behalf. So we're going to dive deeper into Server and cloud. For that, I'm going to hand it over to yo.

Thanks, Jessica. All right, so a few of our pillars guiding the innovations that we're building in Tableau Cloud and Tableau Server are these three. Here we want to make analysts more productive with all the jobs and the workflows that they conduct within the Tableau portfolio. We want to make it easier for business users to self serve, whether that's jumping in one level deeper on a workbook or analyzing governed metrics that data stewards curate on their behalf. And finally, we want to make our platform far more extensible. We want to make it so that you can take the power of Tableau and plug it into the AI application or framework of your choice. In Tableau Cloud, I'll go through a few of the built in intelligence experiences that we've released over the last year. The first one is Tableau Agent. This is an assistant that sits inside the Tableau authoring experience. It's a conversational assistant that can help you build visualizations and generate calculations and overall just work with the Tableau authoring ui. So if you don't want to use exclusively the Drag and drop experience within Tableau. You can chat with Tableau Agent to conduct some of those workflows on your behalf through Natural Language. Tableau Pulse is our business user centered experience in Tableau Cloud. It's all about making business users who maybe don't have as much technical acumen or technical depth be able to easily grok data as they think about it in relation to their business key KPIs like sales and revenue and active users. You can build metrics with Tableau Pulse and then get natural language insights that are generated for those metrics in a deterministic and always accurate way. A new experience that will be coming out with in beta of this fall is Dashboard Narratives. This is similar to the Tableau Pulse experience in that it creates Google governed and deterministic metrics on your pre built dashboards within Tableau. We do this by leveraging the same insight services which generate things like top contributors, outliers, and we visualize and describe those natural language insights right there in the dashboard for you. In Tableau Server. We're also bringing that Tableau Agent experience to the server platform. So you'll be able to chat with an assistant to build visualizations in Tableau Server. And with that I'll show you what these two experiences look like. The Tableau Agent and the Dashboard Narratives experience. So here I'm looking at an HR Attrition dashboard. Excuse me, this is looking at across my company, how many employees have attrited from the company and when and what were some of the sort of attributes of those of those employees as they were either retained or attributed from my company. So the new capability here we get with Tableau Agent is this way to proactively and deterministically generate statistics against, or insights rather against the metrics that are implicit in this dashboard. So I can just click get Dashboard Insights. And under the hood, what we'll run are deterministic statistics, things like top contributors, bottom contributors, outlier detection, and then we'll use an LLM just to summarize the results of those statistical algorithms. There's no hallucination risk here. We're really just using an LLM to make the information itself more presentable, not actually conduct the analysis on our behalf. This is also governable by a dashboard author. I could click if I want. I could configure this to restrict the types of statistics that I run against this dashboard, as well as how I want the language to sound more verbose, less verbose things of that nature. If I need to go one level deeper I can get into the actual authoring experience of Tableau and use Tableau Agent to actually build visualizations or maybe go one level deeper into this, this data model that I'm connected to more in an unbounded way. So I can see here that I have this visualization for education, the education attainment of folks in my organization. Maybe I want to expand on this or break it down further. So I'll click go to sheet with the help of Tableau Agent we can slice or dice this in a different way. Let's say swap in job role for the education dimension and add a filter for gender. Tableau Agent can help in a scenario where you're working from no vis state or an existing vis state. So it gives you that ability to hop between the visual kind of drag and drop UI that we provide. And as well as the natural language assistant here, it's telling me that it first needs me to add a filter and it's going to guide me through this. We'll go ahead and click all and then it'll use that selection to adapt the viz as I instructed it with that new dimension I'm interested in. All right.

Okay. So yeah, thank you. Let's talk about Tableau next. I briefly mentioned that this is the evolution of Tableau and this is the first agentic analytics platform that is built on Salesforce and deeply integrated into AgentForce and Data360 or formerly known as Data Cloud. And this is extremely valuable because first of all you're going to be able to set the foundation for a more accurate context aware agents that are grounded in your AI semantic layer full of all your business context across all your organization. You're going to able to accelerate time to value with purpose built analytics for every salesforce and industry cloud. You can also empower your team to transform every single insight into business results because you have built in actionability and also you can scale these analytics with the composable assets that are going to be available here in Tableau. Next, we also have what we call Agent Force Tableau. This is a set of agentic capabilities that help you to go from data to insights faster. We have Concierge Inspector and Dataproject, but instead of me showing or talking to you about those, let me show you a demo. So for the purpose of this demo, let's pretend that I'm a sales manager and I need to be on top of my business. So I'm going to click on that notification that I just got and it's telling me that my pipegen is Decreasing comparing to the previous period. So this is the first identity analytic capability from Tableau next, that I wanted to show you. This is Inspector, and Inspector is proactively alerting me at changing my data. And here I didn't have to go hunting for it. So I have it here. And the best part is that I can go from this alert to the metric itself to understand what's the story behind this number. Here is the second identity capability that I wanted to show you. Here is where instead of me slicing and dicing my data to understand what's happening, concierge can help me to understand that just in plain English. So I'm just gonna ask a question. What is the lowest sales? Oh, no. The product family. With the lowest sales. Okay. So I'm gonna give it a quick seconds to my agent to analyze my data, but the idea is to come back with a set of insights along with a visualization to help me understand what's happening. And hopefully it's not going to take a lot of time, but a lot of people probably is demoing. So let's give it a. Okay, perfect. It's telling me that the family with the lowest sales is solar panel. Her. Well, okay, that's a very interesting insight because normally that family is my best performance. So I'm going to take advantage of concierge again and I'm going to ask it to create a notification for me. So notify me if cells for solar panel hardware decrease by 2%. Okay, so here, instead of me manually go and create that notification. Concierge is actually doing it for me. It's acting on my behalf and probably it's going to take a few seconds, but probably at this time, at this point, you're telling, okay, this is great, but how can I trust the AI, right? And that is a very valid question. Concierge is powered by Tableau semantic, and that is the AI semantic layer that is deeply integrated into Data360 and that is the brain behind this operation. So let me show it to you. I'm going to open the semantic model and I'm going to make more. Actually, let me open this. I'm going to clear the history and I'm going to make more space to show you the single source of truth. Here is where we are enriching the raw data with business context so we can enable our agents with more accurate responses. This is why our agents understand what we're asking. When we're asking about win rate or product family or csat, all that business context is baked in in this semantic model. And because we are in the semantic model, I wanted to show you the third identity capability that is datapro. Datapro helped me to visualize, prepare or model my data. So here in this case, I'm gonna ask the agent to create a new calculated field. And just like that, I didn't have to understand very much or deep the data. I don't have to know the syntax of that formula. Agent is gonna be proposing me a new formula that I can use. It's giving me the description of what it's gonna be doing and even the formula that I can use, I just have to click, click on Create and it's going to be adding it to the semantic model just like that in a couple of minutes. I was able to get a proactive alert, understand what is happening with a conversational insights and I was able even to take actions right in the flow of work with the tools that I have available at my disposal. So I can actually go back into the metric that we were discussing and quickly share it with my team over Slack so we can continue that conversation there with the channel that we already use for collaboration and communication. And that is the power of Tableaunics. Now with that we're gonna back to the presentation and we're gonna be discussing like the second side of our strategy, which is integrations. So for that I'm gonna pass it to Joe.

Awesome, right? So we just saw the built in the native experiences inside of Tableau Cloud, Tableau Server and Tableau Next that we're building to make analysts more productive business users more able to self serve. But the third important pillar of our strategy is how we enable developers, our community of people that want to take what they've done in Tableau and extend that into third party applications or into custom build applications. Our focus in this strategy today is around Tableau mcp. MCP stands for Model Context Protocol. Think of this as the glue between agents and third party applications. What the Tableau MCP server does for Tableau Classic and Tableau Next is it wraps a collection of different APIs on our platform as well as different resources, prompts and really business logic, the semantics of what you have in these environments and it makes it available to any agent. So what does this actually look like? So we just saw. Let's get a new chat here. We just saw that demo I did on the HR Attribution Dashboard, right? And that dashboard was built off of a published data source. I've connected a Tableau MCP server here to Claude. Claude's a third party AI application. But because. But through our MCP server, it gets access to those same semantics that our native AI experiences get access to. So I can say, hey Claude, do you have access to the HR attrition data source? And CLAUDE will use the tools that we exposed to it here to identify if it has access to this data source. It found that HR attrition data source, the one that I used in that dashboard demo. And we want this one, not the one that we do evals off of. Great. I'm looking for the boom and let's ask it for one of the insights that was actually available in that dashboard. Like remind me how our attrition varies among age groups through one of our tools called Get Data Source Meditator Metadata. Excuse me. CLAUDE is able to ground on the semantics of the data source that we were just looking at in Tableau and now can construct semantic queries against that data model. It found age group, it found attrition. It looks like the way it maybe executed this query in the first case led to a error, so it recursively improved and now we get that summary of our attrition across age groups, that same same insights that we were seeing in the visualization over in Tableau. And actually to prove that we can just go over here and look at our viz and see that age group analysis. Let's see for attrition 25 to 42, 34. Yeah, 112. Same data, different interface. Awesome. All right, so thank you all for joining us. Great discussing with you all the different experiences that we're building within Tableau to make that the experience more AI rich as well as that extensibility investment. Happy to take questions off to the side and discuss any of this further. Thanks. Thank you.

Thank you.