How to Create a Report in Tableau with AI
Building reports in Tableau no longer requires you to master its intricate drag-and-drop interface or complex calculation syntax. Thanks to artificial intelligence, you can now analyze data by simply asking questions in plain English. This updated guide dives into Tableau's AI features, showing you step-by-step how to turn simple text prompts into detailed reports and powerful visualizations.
What Exactly is Tableau's AI? Meet Einstein.
If you're familiar with the Salesforce ecosystem, you've likely heard the name "Einstein." This is the branding for all of Salesforce's AI technology, and it's now deeply integrated into the Tableau platform. While it might sound like a single feature, it’s a suite of tools designed to make data analysis more accessible and efficient. For a report builder, the two most important components are Einstein Copilot and the AI features built into Tableau Desktop.
Einstein Copilot: Your Conversational Data Analyst
The main AI feature you'll interact with for report building is Einstein Copilot. Think of it as a smart chat window directly within Tableau. Instead of hunting through menus for dimensions and measures, you can ask questions like:
"What were our total sales last quarter?"
"Show me website traffic by country as a map."
"Which product categories have the highest profit ratio?"
The Copilot understands your question, analyzes your connected data, and suggests an appropriate visualization - a bar chart, a map, a line graph - right in the chat panel. From there, you can refine it with follow-up questions or drag it directly onto your reporting canvas. It’s a completely different way to interact with your data, transforming analysis from a series of clicks into a conversation.
AI-Powered Calculations in Tableau Desktop
One of the steepest learning curves in Tableau is mastering calculated fields, especially Level of Detail (LOD) expressions. These are powerful but a big headache for new users. Tableau’s AI directly helps you with this by generating calculation syntax for you. You can ask for what you need in plain English - like "calculate year-over-year sales growth" - and the copilot will write the complex formula you need. This not only saves an immense amount of time but also helps you learn the syntax by seeing how it’s constructed correctly.
Quick Tip: Preparing Your Data for an AI Assistant
Any analyst will tell you that the quality of your insights depends entirely on the quality of your data. The saying "garbage in, garbage out" is especially true when working with AI, which relies on a well-structured foundation to understand your requests accurately.
Before jumping into building reports, take a moment to eyeball your data source. You don't need a Ph.D. in data engineering, just check for a few basics:
Clear Column Names: Name your columns something a human would understand. Use "TotalSales" instead of "TS_v2_final" and "CustomerRegion" instead of "C_Reg". The AI uses these names to interpret your questions.
Correct Data Types: Ensure dates are formatted as dates, geographic fields (like State or Country) are set to their proper roles, and numbers are recognized as numbers. Tableau is pretty good at this automatically, but it's always worth a double-check.
Address Nulls Sensibly: Empty cells or "null" values can throw off calculations and charts. Decide if you should filter them out, replace them with a zero, or leave them as is, depending on the context.
A few minutes spent on ensuring your data is clean and logically named will make your interactions with the AI far more productive and your reports more reliable.
How to Build a Report with Einstein Copilot: A Step-by-Step Guide
Enough theory - let's walk through how you would actually use Einstein Copilot to build a sales report from scratch. Once your data is connected in Tableau, the process looks something like this.
Step 1: Open the Einstein Copilot Panel
In your Tableau workspace (whether on Tableau Cloud, Public, or Desktop), look for the Einstein icon. It’s typically located near the top right of the interface. Clicking this will open the Copilot pane on the side of your screen. This is your command center for all conversational analytics.
Step 2: Ask Your Initial Question
Start with a broad question to get a feel for your data. Your prompts don't need to be perfectly phrased - the AI is designed to understand common business language. Let's try to analyze sales performance.
Type into the chat panel:Show me my total sales and profit by product category
After a moment, Einstein will process your request. It identifies "Sales," "Profit," and "Product Category" from your data source and automatically generates what it considers the best visualization for this comparison - most likely a standard bar chart showing the sum of sales and profit for each category.
Step 3: Refine Your Visualization with Follow-up Questions
Here’s where the conversational aspect really shines. The initial chart is a starting point, not a final product. You can iterate and dig deeper using follow-up questions. The AI remembers the context of the conversation.
Let's continue our example. We have the bar chart, but now we want to see only our top categories. Try asking:"OK, now show only the top 5."
The chart will instantly update to display only the five categories with the highest sales. Now, let's say we want to visualize this a different way.
"Change that to a treemap."
The visualization in the Copilot window will switch from a bar chart to a treemap, providing a new perspective on the data. You can keep going like this, slicing and dicing your data until you uncover something interesting.
A few more examples of follow-up questions:
"Filter for orders from California and Texas only."
"Break this down by customer segment."
"What's the relationship between discount and profit for these categories?"
Step 4: Add the AI-Generated Chart to Your Report
Think of the Copilot pane as your scratchpad. You can ask dozens of questions and explore different angles without messing up your main report. Once you’ve created a visualization that tells an important story, you'll want to add it to your official dashboard or worksheet.
This part is incredibly simple. You just click and drag the chart from the Einstein Copilot panel directly onto your main canvas. It becomes a standard Tableau worksheet that you can then format, add to dashboards, and publish just like any other manually created chart.
Saving Time by Writing Calculations with AI
Beyond building charts, Tableau’s AI functions as an incredible accelerator for writing custom calculations. This is particularly helpful for creating new metrics that don’t already exist in your dataset.
Imagine you want to calculate the profit ratio, which is (Profit / Sales), as a new field to use in your reports.
Navigate to Analysis > Create Calculated Field, or right-click in the Data pane and select it there.
The calculation editor will pop up. Inside, you'll see a small Einstein icon. Click it to open the AI assistant.
In the prompt box, describe the calculation you need in simple terms:
Create a calculation for profit ratio as profit divided by sales.4. The AI will generate the correct syntax, which in this case would be SUM([Profit]) / SUM([Sales])5. You can review the formula, name your new field "Profit Ratio," and click Apply.
Just like that, you've created a reusable calculated field without worrying about proper syntax, aggregations, or brackets. This functionality turns what used to be a point of friction into a moment of productivity, especially for complex formulas like LOD calculations or date functions.
Tips for Writing Better AI Prompts in Tableau
To get the most out of Einstein Copilot, it helps to be mindful of how you phrase your questions. You don't need highly technical prompts, but a little clarity goes a long way.
Be Specific About Metrics and Dimensions: Instead of "show sales," try "show the sum of sales by region." Including the "what" (sum of sales) and the "how" (by region) leads to more predictable results.
Include a Time-Frame: Use clear time filters in your prompts, like "in the last 6 months," "for 2023," or "by quarter last year." This is one of the most common filters you’ll need.
Ask for a Specific Chart Type: If you know what you want, ask for it! Add phrases like "as a line chart," "show it on a map," or "display as a table" to the end of your prompt.
Start Simple, Then Layer on Complexity: Begin with a foundational question and then use follow-up prompts to add filters, change dimensions, or break down the data further. This iterative process is often more effective than trying to write one perfect, long-winded question.
Rephrase If It Fails: AI isn’t perfect. If Copilot misunderstands your request or gives you an odd result, try rephrasing your question in a simpler way. Sometimes just changing one or two words is all it takes to get the right output.
Final Thoughts
Tableau's Einstein features fundamentally change report creation from a manual drag-and-drop process to a fluid conversation. By using natural language prompts, anyone can explore complex data, build visualizations on the fly, and even generate complicated calculations without facing the traditional steep learning curve of a BI tool.
If you're in marketing or sales and find a full-scale BI tool like Tableau a bit daunting, this conversational approach to data is revolutionary. We built Graphed for exactly that purpose. It allows you to ask similar plain-English questions - like "Show me my top-performing Facebook ads this month" or "What is our sales pipeline conversion rate?" - and get immediate dashboards that connect all your live data sources in one place, hassle-free. It skips the steep learning curve of traditional reporting and gets you straight to the answers you need to grow your business.