How to Create a Report in Looker with AI

Creating detailed reports in Looker has traditionally involved a steep learning curve with its own modeling language, LookML, and a structured development workflow. But that process is changing. With the introduction of generative AI, you can now build reports and dashboards by simply describing what you want in plain English. This completely transforms the process, shifting the focus from technical report-building to getting answers quickly.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

This tutorial will walk you through exactly how to create a report in Looker using its new AI features. We'll cover the step-by-step process of using conversational prompts to generate visualizations, along with practical tips to get better, more accurate results.

What is Looker and Why a Quick Refresher Matters

Before jumping into the AI features, it’s helpful to understand what makes Looker different from other BI tools. Looker, part of the Google Cloud ecosystem, is known for its powerful data governance layer called LookML. Think of LookML as a centralized rulebook for your company's data. A data team defines all your business metrics once (e.g., "what counts as revenue," "how we define a new user"), and everyone in the organization uses those same trusted definitions.

This approach ensures consistency and reliability, preventing situations where the marketing team's revenue number doesn't match the finance team's. It's this structured foundation that makes the addition of AI so powerful. The AI isn't just guessing at your data, it's working with a well-defined and trustworthy data model, leading to more reliable outputs.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

Looker’s New AI Playground: Duet AI

The main engine behind Looker's AI capabilities is Duet AI, Google's generative AI assistant built into its cloud products. In Looker, Duet AI acts as an on-demand data analyst you can talk to. Instead of navigating menus and manually dragging fields to build a chart, you can simply tell it what you need.

This works in a few key ways:

• Conversational Queries: At its core, you can ask questions directly. Prompts like "Show me our top 10 products by sales in the last quarter" are understood and translated into reports on the fly.
• Generative Visualizations: Duet AI doesn't just give you raw data, it intelligently selects the right kind of chart - bar, line, pie, etc. - to best represent the information you asked for.
• Report Refinement: The conversation doesn't stop after the first prompt. You can ask follow-up questions to drill down or modify the report, such as "Now break that down by country" or "Change this to a weekly view."

This conversational approach significantly lowers the barrier to entry, making data accessible to team members who aren't Looker experts.

How to Create a Report in Looker with AI: A Step-by-Step Guide

Ready to build your first AI-powered report? The process is refreshingly straightforward. Once your data team has set up your LookML model and connected your data sources, you can get started.

Step 1: Open Looker and Find the AI Assistant

When you navigate to a Looker Explore or dashboard, you will typically find a chat icon or a prompt dialog for Duet AI. This is your starting point. It’s an open invitation to start a conversation with your data. The goal is to make data interaction as natural as messaging a coworker.

Step 2: Write a Clear and Specific Prompt

The quality of your report depends entirely on the quality of your prompt. Vague requests lead to vague results. The key is to be as specific as possible. Include the metric, dimensions, chart type (if you have a preference), and time frame.

Here are some examples of effective prompts:

Show me total revenue by marketing campaign for the last 30 days.

Create a line chart visualizing new users from Google Analytics on a daily basis this month.

Which three countries generated the most sales in Q4? Show it as a bar chart.

Common mistakes to avoid:

• Being too generic: A prompt like "Show sales" is too broad. Sales of what? Over what period? Broken down by what?
• Using vague timeframes: Instead of "recently," use "in the last 7 days" or "this quarter."
• Assuming context: Always state the metric you want to see (e.g., "revenue," "sessions," "conversion rate").

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

Step 3: Review and Verify the Generated Report

Once you submit your prompt, Duet AI will process your request and generate a visualization. Treat this as a first draft. It’s your job to review it and ensure it accurately reflects what you asked for. Ask yourself these questions:

• Did it choose the correct metric and dimensions?
• Is the time frame accurate?
• Does the visualization make sense for this data? (e.g., a pie chart isn't great for showing trends over time).
• Do the numbers look reasonable based on your general business knowledge?

Remember, AI is a powerful assistant, but it’s not infallible. A quick sanity check is always a good idea.

Step 4: Refine and Iterate with Follow-Up Questions

This is where conversational AI really shines. Your initial chart is just the beginning. You can now tweak and expand on it with simple follow-up prompts instead of starting over. Think of it as a continuous dialogue.

Starting with the report from "Create a line chart visualizing new users from Google Analytics on a daily basis this month," you could follow up with:

• "Okay, now add mobile traffic to the chart for comparison."
• "Can you filter this to only show users from the United States and Canada?"
• "Change the visualization to a summary table."
• "What was the total number of new users last month?"

This iterative process allows you to explore your data naturally, uncover insights, and gradually build the exact report you need without touching a single settings menu.

Step 5: Save Your Report or Add it to a Dashboard

Once you're satisfied with your finalized chart or report, you can save it as a "Look" or add it to a new or existing dashboard directly from the interface. It becomes a live, interactive element that will continue to update as new data streams in, just like any other Looker report.

Tips for Getting More From Looker's AI

Using Looker's AI effectively is a skill that improves with practice. Here are a few tips to help you get reliable, insightful reports every time.

• Learn the Lingo: The AI works best when you use the exact names of the fields, dimensions, and measures defined in your company's LookML model. If the term for revenue is "Total Gross Sales," using that phrase will give you better results than just asking for "sales."
• Build Incrementally: If you need a complex dashboard with multiple charts, don't try to draft it all in one-long, complicated prompt. Build it one chart at a time. Perfect the first one, then ask the AI to "add another chart showing..." This approach is more manageable and easier to troubleshoot.
• Experiment with "Why" questions: Don't just ask for metrics. Try to get at the relationships in your data. For example, after seeing "top product sales," you might ask, "which marketing channel drove the most sales for our top product?" to uncover valuable connections.
• Focus on a Clear Business Question: Before writing a prompt, be clear about the business question you’re trying to answer. Are you trying to understand campaign performance? Identify top-performing sales reps? Find the cause of a drop in traffic? A focused question will always lead to a more focused and useful report.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

The Shift: From Report Builder to Data Strategist

The rise of AI in tools like Looker represents a fundamental shift in how we approach data analysis. The tedious busywork of report creation - the endless clicking, dragging, and configuring - is rapidly becoming automated. This frees you up from being a report builder and allows you to become a data strategist.

Your time is no longer spent figuring out how to build the chart, it's spent understanding what the chart means and deciding what to do next. It democratizes data, allowing product managers, marketers, and sales leaders to get answers on their own without creating a ticket for the data team and waiting days for a response. Decisions can be made in hours, not weeks, based on real-time information.

Final Thoughts

Leveraging AI within Looker transforms report creation from a technical chore into a simple conversation. By crafting clear prompts and iterating with follow-up questions, anyone can quickly visualize data, uncover insights, and build the reports they need to make smarter business decisions. This makes a historically complex BI tool far more accessible and accelerates the entire analytics workflow.

At Graphed, we've built our entire platform around this AI-first, conversational experience. While powerful tools like Looker require a significant setup of a semantic data layer, we help teams connect their data sources - like Google Analytics, Shopify, Facebook Ads, and Salesforce - with a few simple clicks. We've eliminated the technical complexity so you can go from data connection to asking questions and getting real-time dashboard answers in just a few minutes, turning a half-week of manual reporting into a 30-second conversation.