How to Create an Analytics Dashboard in Looker with AI

Building a powerful analytics dashboard in Looker no longer requires hours of clicking, dragging, and tweaking complex settings. By bringing artificial intelligence into the process, you can now describe the charts and reports you want in plain English and watch them come to life in seconds. This article will guide you step-by-step through using Looker's AI capabilities to create insightful dashboards, making your data more accessible for the entire team.

Why an AI-Powered Approach is a Game Changer

For years, creating a dashboard in Looker (or any business intelligence tool) followed a familiar, manual process. You'd log in, choose a data source, select specific dimensions and measures, apply filters, and meticulously configure a chart’s visual settings. While powerful, this process required a significant learning curve. You needed to understand your data schema, know exactly where to find each metric, and have the patience to build reports one component at a time.

Enter generative AI. Instead of being a builder, you become a director. You tell the system what you need, and it handles the tedious mechanics for you. The difference is transformative:

Drastically Reduced Learning Curve: You no longer need to be a BI expert. If you can ask a clear question, you can get a clear answer in the form of a chart or table. This opens up data exploration to non-technical team members, from marketers to sales managers.
Unbelievable Speed: What used to take 15-20 minutes of navigating menus can now be done in under a minute with a single sentence. This frees up countless hours for actual analysis rather than report building.
Conversational Exploration: The process becomes a dialogue. You can start with a broad question, see the result, and then ask follow-up questions to drill down without starting from scratch.

Getting Started: Your Pre-Flight Checklist in Looker

Before you can start prompting your way to a perfect dashboard, you need to ensure Looker has access to your data. The effectiveness of any AI model depends entirely on the quality and structure of the underlying information it’s working with. This initial setup remains a crucial step.

1. Connecting Your Data Source

Looker excels at connecting to a wide variety of SQL databases, data warehouses, and services. Common sources include:

Cloud data warehouses like Google BigQuery, Snowflake, and Amazon Redshift.
Transactional databases like PostgreSQL or MySQL.
Google Analytics data, often piped into BigQuery.

The connection process typically involves providing credentials and configuring database access so Looker can query the data live. This stage is often handled by a data engineer or analyst, as it’s the most technical piece of the puzzle.

2. The Role of LookML

Looker’s secret sauce has always been LookML, its proprietary modeling language. A LookML model acts as an abstraction layer that sits on top of your raw SQL database. Your data team uses it to define all of your business logic in one place.

What does this mean for you? The LookML model is where core definitions are stored. It tells Looker things like:

"Revenue" is calculated by summing the price column from the orders table.
"User" and "Order" tables are joined on the user_id field.
Dimensions like "Country" and "Device Category" exist and can be used for grouping data.

This organized semantic layer is what makes natural language queries possible. When you ask the AI to "show revenue by country," it consults the LookML model to understand what "revenue" and "country" mean and how to properly query the database to get you that information accurately.

How to Use Looker's Duet AI to Create Dashboard Tiles

Google has integrated its Duet AI directly into Looker to handle conversational analytics. This feature lets you build visualizations (which Looker calls "Tiles") through a simple, chat-like interface. Let’s walk through how to create a useful visualization from scratch.

Step-by-Step: Prompting Your First Chart Tile

Imagine you're a marketing manager and you want to quickly see which campaigns are delivering the most website traffic. Here’s how you’d use Duet AI to find out:

Start on Your Dashboard: Navigate to the Looker dashboard where you want to add a new report. Click the edit button, then find the option to add a new tile. Looker now provides an option like "Generate tile with AI".
Initial Prompt: A conversational box will appear. Here, you'll type your request. Be as specific as you can. A good prompt includes the metric, dimension, and timeframe. For example: Show me website sessions from Google Analytics by campaign as a bar chart for the last 90 days.
Generation & Review: Duet AI will interpret your request, query the underlying data (via the LookML model), and generate a suggestion for a bar chart. It will appear on your screen for review.
Refine with Follow-up Questions: This is where the conversational part becomes powerful. You might look at the chart and realize it’s too cluttered. You can refine it with another prompt: Okay now only show the top 10 campaigns, sorted in descending order.
Change the Chart Type: Perhaps a different visualization would tell a better story. You can easily switch it up: Change this to a horizontal bar chart.
Add to Dashboard: Once you're satisfied with the visualization, you can officially add it as a new tile to your dashboard. From there, you can resize and position it alongside your other reports.

In just a few moments, you’ve gone from a question to a finished, data-driven visual without manually touching a single settings menu.

Best Practices for Writing Effective Prompts

The quality of your output depends on the quality of your input. While Looker's AI is designed to understand amorphous, human language, clear prompts get you better results faster. Here are a few tips:

Be Specific About Metrics and Dimensions

Avoid vague terms. The AI relies on the named fields in your LookML model.

Instead of: "Show me some sales data"
Try: "What was our total revenue and average order value by month for 2023?"

Always Define a Timeframe

Data without a timeframe is often meaningless. Tell the AI exactly what period you're interested in.

Instead of: "How are our new users trending?"
Try: "Show me the number of new users per week for the last 6 months as a line chart."

Specify the Chart Type You Want

While the AI can often infer a good chart type, you can guide it for better results.

Instead of: "Breakdown traffic sources"
Try: "Show me a pie chart representing the percentage of total sessions by marketing channel last quarter."

Iterate from Simple to Complex

Don’t try to craft one perfect, overly complex prompt. Start with a simple request to make sure the AI understands the core components, then add layers of complexity with follow-up prompts.

Start: "Show me total orders for this month."
Follow-up: "Now break that down by product category."
Refine: "Filter out returns from the totals."

Beyond Building Charts: Report Summaries & Deeper Insights

AI's role in Looker doesn't stop at building individual charts. It's also being integrated to help with analysis and interpretation. A key feature is the ability to generate automatic summaries of dashboards and reports.

At the top of a Looker dashboard, you might see an option to "Get a summary." Clicking this prompts Duet AI to analyze all the tiles on the dashboard and provide a concise, natural language written summary of the key findings. It might give you bullet points like:

"Overall revenue is up 15% this quarter, driven primarily by strong performance in the 'Outdoor Gear' category."
"While website traffic from paid search is high, conversion rates are lower than organic search."
"The West region is currently the top-performing sales territory, exceeding its goal by 8%."

This feature is a powerful tool for stakeholders who need a quick, high-level overview without having to interpret every single chart themselves. It saves time and helps surface the most important insights hidden in the data.

Final Thoughts

The integration of AI into platforms like Looker represents a major shift in how we approach business intelligence. By allowing users to create visualizations and get summaries using natural language, it dramatically lowers the barrier to entry for data analysis. This shift empowers every team member to harness the power of their data without needing extensive technical training, leading to faster, more informed decision-making across the organization.

At Graphed , we are building the future on this simple premise: data shouldn't be hard. While getting a tool like Looker ready for AI can still be a technical hurdle requiring data engineers to set up connections and maintain LookML models, we provide a turnkey solution. You can connect all your marketing and sales tools - like Shopify, Google Analytics, Facebook Ads, and Salesforce - in seconds, with no technical setup required. From there, you can talk to your data and build entire cross-platform dashboards in simple English, getting immediate, real-time answers that help you actually grow your business instead of just reporting on it.