Learn About Tableau AI

Learning Objectives 

After completing this unit, you’ll be able to:

• Explain the relationship between generative AI and analytics.
  
• Describe Tableau AI's approach to conversational analytics.
  

Explore How Generative AI and Data Go Together

Simply put, generative AI produces an output based on all the information it learns. That’s where generative comes from. You ask; it generates. For example, if you feed a generative AI learning model tens of thousands of pages of poetry, you can be confident that when you ask it to write a poem, it will produce a poem with surprising results.

But notice that data is at the core of generative AI. It only works if it has the right amount of information needed to address your questions. The same goes for business analytics and business intelligence (BI). People can only make informed business decisions when they have the right information at the right time.

So what happens when you put generative AI together with industry-leading analytics and BI tools? Wait, not so fast!

Take On the Challenge of AI Hallucination

When you first think about combining Tableau with something like generative AI, you might think about a conversational interface: “Create a dashboard telling executives how the business did this quarter.” And then you’d expect the generative AI to pull out the insights. While this scenario sounds awesome, the challenge is that even with all the advancements in generative AI, it’s still not at a place where it can do math reliably.

Consider AI hallucinations. It’s the idea that generative AI produces an answer without much certainty on how accurate that answer is. It’s modeling a response based solely on the information it has already learned. This makes generative AI really great for creative applications, but not necessarily for concrete analysis.

At Tableau, we’re taking a thoughtful approach to how we implement generative AI with our analytics platform. We want to make sure the core analytics, the calculations, and the math, are done the way they always have been. That is, with predefined, deterministic code that’s developed by humans in a way that can be fully trusted. We use generative AI to supercharge other parts of the analytics experience.

Balance Generative AI with Trusted Analytics

Let's use a simple example: Say you want to know what profit was made this past quarter. Profit, of course, is a function of revenue minus costs. 

Here’s what’s needed to arrive at a good answer that someone can easily understand and act on.

First, you have to define profit. Profit = revenue − cost. Generative AI is good at connecting things like this. It sees that the data source has columns for sales and understands that sales is probably a reasonable proxy for revenue. It knows the relationship between concepts like revenue, cost, and profit. So generative AI can help build these definitions or this data model.

Second, you must compute the value to figure out what the profit was for that period. Like we said previously, generative AI isn’t so good at actual computations, so Tableau executes these calculations in a more traditional way.

Finally, once you have the answer—once you know what profit, revenue, and costs are—you need to communicate that to the end user. A combination of visuals and natural, conversational language is an effective form of communication. Generative AI can take a set of facts and translate it into something meaningful and natural that a nontechnical user can easily understand.

This is a simple example, but it illustrates the core process of business analytics. It demonstrates where generative AI can be used to enhance and accelerate this process, and most notably, where it should not be used.

Make Analytics Faster with Tableau AI and Tableau Pulse

Tableau AI uses the advanced capabilities of generative AI to simplify the process of data analysis and drive performance, efficiency, and scale. It takes a different approach to conversational data analysis by using prompts within a conversational interface. Instead of requiring the end user to know the right question to ask, Tableau AI guides them by proactively presenting questions. Since Tableau has a deep understanding of the data, it’s able to present relevant and personalized options for end users to react to. 

Think of it like this: It’s dinnertime. You have a personal chef. Do you need to describe the exact meal that you want with all of the ingredients and cooking temperatures and timing? How about which sauce will go best with your entree, what garnishes will make it pop, and how to plate it? That’s a lot for someone who isn’t an expert! 

But what if the chef gave you a few thoughtful options, some you might never have considered? The chef also knows both what you might like and how to pair ingredients. While you have the ultimate say, they give you some choices and recommendations to create the best meal.

For many of the same reasons, a proactive conversational interface with data and analytics makes sense for most business users. 

With Tableau Pulse, data comes to you with proactive insights. It answers common questions to deepen your analysis without you having to formulate them for asking. When you do have questions, choose from a few one-click prompts, or ask your question through the conversational interface.

As for looking at the business as a whole, get AI-generated summary and insights right from the homepage. Drill into the details with suggested prompts, or use hybrid search to uncover the questions that are important to you. 

Resources 

• Blog: How Tableau AI and Tableau Pulse are reimagining the data experience
• Blog: Salesforce Introduces the Next Generation of Tableau, Bringing Generative AI for Data and Analytics to Everyone