Getting actionable insights from your business data shouldn’t require deep familiarity with SQL or your company’s LookML model. It should be as simple as asking a question. By pairing Looker’s trusted semantic layer with generative AI, Conversational Analytics empowers everyone to ask complex questions in natural language, instantly receive answers, and visualize trends.
Today, we are announcing updates (version 26.6) to Conversational Analytics in Looker which focus on delivering more accurate, faster and intuitive data experiences.
New features in Conversational Analytics in Looker
2026 has seen significant upgrades to the core functionality of Conversational Analytics.
-
Improved Reasoning: We’ve enhanced the agent’s ability to process and interpret natural language questions, leading to sharper, more accurate insights.
-
Thought Message Visibility: We’ve pulled back the curtain on the agent’s ‘thinking’ process, offering full transparency into exactly how your answers are derived.
-
Smarter Filter Value Resolution: We introduced better tools for resolving filter values. Now, even if a user doesn’t type the exact data value in their question, the Conversational Analytics agent can intelligently figure out and apply the proper filters.
- Sample Data Tool allows the agent to get the top 100 values to find a specific data point or learn patterns from the data
- Sample Data Tool allows the agent to get the top 100 values to find a specific data point or learn patterns from the data
-
Fuzzy Search Tool allows the agent to execute searches against AI-generated terms to find potential matches for user-terms in the database.
-
Tip: both of these tools use Looker’s suggestion API under the hood. This means you can make these tools work faster by hardcoding suggestions in LookML, or using suggest_explore to point to an even faster query.
-
Advanced Analytics (preview): Looker’s Conversational Analytics Agent can execute Python code to do deeper analysis on top of data retrieved from Looker. This allows the agent to do complex calculations, forecasting and machine learning. To try this feature, turn on the “Code Interpreter” toggle in the “Gemini in Looker settings”, then toggle on the “Advanced Analytics” setting when you create an agent.
Smarter, Faster, and Clearer Conversational Analytics
The Conversational Analytics updates in the Looker 26.6 release focuses on optimizing performance and clarity, setting the stage for even more complex analyses down the road:
-
Disambiguation: To ensure you get exactly the data you need, the Conversational Analytics Agent will now ask clarifying questions when a request is ambiguous.
-
Fast Thinking Mode: If you need answers in a hurry, users can leverage a new Fast Thinking Mode for quicker response times. (Note: While this reduces overall thinking time, it may impact transparency and accuracy for complex queries).
-
Latency Improvements: We have further optimized the system to significantly reduce overall response times, through better use of Looker’s cache and running tools in parallel.
-
Improved Debug UI: Developers now have an enhanced interface with better insight for understanding and troubleshooting the agent’s output.
Tip: Use the summary view to see exactly where the agent is spending its time, then head into the details for granular information.
-
Charting Improvements: We have improved the reliability of the charting experience, including fixes where the agent may previously have returned a blank chart.
-
Upgraded Documentation: We’ve expanded our resources to give you better visibility into how things work behind the scenes, alongside an improved best practices guide for crafting LookML and building meaningful agent instructions.
What We’re Focused on Next
We anticipate expanding Conversational Analytics’ power and integration capabilities. We will be heads down on enhanced observability, supporting conversational analytics in new Looker surfaces, and giving it new sets of capabilities. Join us at Google Cloud Next to learn more.