From Dashboard to Dialogue: How AI Accelerates Data Analysis in Looker

The Strongest AI Features at a Glance

Google has introduced a series of AI features into its BI tool Looker, that continuously make working with corporate data easier for users:

• Conversational Analytics: This feature allows users to perform ad-hoc analyses by asking questions to the tool in natural language. Thanks to Looker’s semantic layer, the AI-generated answers are ensured to use the correct business logic and are quickly verifiable.
• Formula Assistant: With the help of the Formula Assistant, analysts and business users can quickly and easily have complex formulas generated while exploring data in Looker, which significantly reduces manual effort.
• LookML Assistant: This assistant automates code generation for LookML, the language used for creating semantic data models in Looker. This accelerates development and reduces the burden on data engineers.
• Visualization Assistant: The Visualization Assistant enables Looker users to quickly generate and modify custom visualizations of their corporate data using natural language.
• Slide Generator: This feature generates slides directly from dashboards, currently only available with the Looker Studio Pro license. The generated text narratives explain the data in the charts and highlight the most important insights. Since the presentations remain linked to the underlying reports, they are always based on the most up-to-date data.

Although the features currently still reach their limits with highly complex queries, they already provide a measurable added value for clear, precisely defined tasks.

For users of the Google Cloud, further synergies emerge. Thanks to the native integration of Looker into the Google Cloud Platform (GCP), the powerful AI tools of the cloud, such as Vertex AI and AI in BigQuery, can be seamlessly connected.

Focus on Conversational Analytics: You Ask, Your Data Answers

As an AI function with particularly great potential, Conversational Analytics (CA) in Looker enables users to gain insights from data using natural language. This is particularly valuable for generating insights faster outside of fixed dashboards and reports. Instead of delving deeply into filters, drill-downs, or segmentations in the Explore view, users send clearly defined prompts to Looker.

CA thus functions somewhat like an intelligent chatbot that you ask questions. Thanks to the integrated Reasoning Engine, users can track the logical process behind the output at any time. This transparency is crucial for trust in the results.

Conversational Analytics makes it easier for many employees to get started with the tool, reduces the inhibition to work data-driven, and makes corporate data accessible even to less tech-savvy employees. At the same time, in our workshops we have repeatedly noticed that a basic technical understanding of the semantic layer and the functionalities of Looker leads to better and more precise results when using Conversational Analytics.

No Secret: We Work with Agents

For users, the interaction in Conversational Analytics is often supported by so-called Agents. These Agents can be continuously developed by the company and tailored to specific target groups to answer particular queries in a targeted way. What's special is that it is possible to provide the Agents with corporate meta-knowledge. For example, if a company defines the rule that "successful locations" should automatically be interpreted as "ROI per branch over the last fiscal year," the AI is enabled to professionally assess non-specific prompts from users.

In this way, even employees outside the specific department can access the collective corporate knowledge. Complementary to this, there is another valuable feature (currently in preview) that promotes teamwork: the ability to share conversations with Agents to quickly pass on gained insights within the team.

An essential aspect when using Agents is data security, which is ensured by the comprehensive governance management in Looker. Your corporate data always remains secure within the Looker and Google Cloud environment and is not used for external development of the underlying AI models.

Furthermore, you retain complete control over the Agents:

• The Agents are trained and iteratively improved exclusively by the company, not by Google.
• The company precisely determines which Agents may access which data and which users can utilize the respective Agents.

These strict governance rules ensure that your data is protected and the use of the Agents complies with internal security and compliance requirements.

(Meta) Data Quality as the Key to AI

For the use of all AI functions in Looker, the same rule applies as always: the better the data quality, the more reliable the resulting analyses. The focus now broadens to the context, which is crucial for the semantic classification of the results. If the company’s business logic is not cleanly anchored in the data model, neither users nor AI Agents can derive reliable analyses. After all, it is of little use if relevant specialized knowledge resides exclusively in the minds of departmental staff; the AI then risks not using the correct KPI to answer a question. Thus, beyond mere data hygiene, the semantic layer gains new, additional importance as soon as the analysis interface is utilized by intelligent systems.

Our recommendation for companies wishing to test Looker's AI options is therefore: First, clarify where the relevant user knowledge currently resides and how it can be transferred to the model and the Agents. Subsequently, we select a suitable use case with good raw data quality to implement best practices and jointly test the functions.

What Comes Next? A Look into Our Crystal Ball

The development of Looker is far from over, and the next steps promise an even deeper integration of data analysis into the entire daily workflow. A major focus is on the de-limitation of Looker: with the Conversational Analytics API, companies will soon be able to embed the semantic capability of AI analysis into external applications. This includes out-of-the-box integration into Google Workspace and common chat applications, which dramatically improves collaboration between different tools.

Generally, it is expected that Looker features will become increasingly accessible outside the Looker interface (for example, via the Google Cloud Platform (GCP) Toolbox or Agentspace) and will also be continuously expanded and optimized within Looker to enable data-driven work and decisions in every workflow.