Google’s BigQuery and Looker get agents to simplify analytics tasks

Google has added new agents to its BigQuery data warehouse and Looker business intelligence platform to help data practitioners automate and simplify analytics tasks.

The data agents, announced at the company’s Google Cloud Next conference, include a data engineering and data science agent — both of which have been made generally available.

[ Related: Google Cloud Next ’25: News and insights ]

The data engineering agent, which is embedded inside BigQuery, is designed to help data practitioners by delivering support to build data pipelines, perform data preparation, and automate metadata generation.

According to Google, the data engineering agent will simplify and accelerate analytics tasks as data practitioners, typically, have to spend a majority chunk of their time creating a pipeline for data and preparing data to get actionable insights from it.

Another capability of the data engineering agent, which is currently in preview, is the ability to detect anomalies in order to maintain data quality.

The data science agent, accessible via the company’s free, cloud-based Jupyter notebook service — Colab, is designed to help data scientists automate feature engineering.

Feature engineering in data science refers to the process of revamping raw data into features that a model can use to make predictions.The agent is also capable of providing intelligent model selection and enabling scalable training along with faster iteration cycles.

This would allow enterprise data science teams to focus on building data science workflows instead of having to worry about revamping data and managing infrastructure.Google has also added a conversational analytics tool to Looker, currently in preview, to help enterprise users interact with data using natural language.

The tool, essentially an agent, shows the reasoning behind its response to a query to help the end user understand and monitor its behavior — a requirement for most enterprises deploying agents in order to escape pitfalls around hallucination.

However, Google pointed out that the tool is powered by Looker’s semantic layer and that is expected to improve its accuracy.The cloud services provider has also made an API of conversational analytics available to developers to help them integrate it into applications and workflows.

The agents made available via BigQuery, Looker, and Colab comes at no additional cost, the company said in a statement.

BigQuery gets a new knowledge engine and an AI query engine

As part of the updates to BigQuery, the cloud services provider is adding a knowledge engine to help enterprises analyze autonomous data — datasets inside the warehouse that exists independent of any applications.

According to Google, the knowledge engine will use Gemini to analyze schema relationships, table descriptions, and query histories to generate metadata on the fly, and model data relationships.

This engine would act as the foundational layer for enterprises to ground AI models and agents in business context Google said.

In order to assist data practitioners further, Google said that it was adding intelligent SQL cells to BigQuery notebook.

The cells can understand data’s context and provide suggestions as scientists write code while enabling them to join data sources directly within the notebook, the company said.

Other updates to the notebook includes features, such as the ability to share insights across an enterprise and build interactive data applications.

In its efforts to help data practitioners analyze structured and unstructured data together inside BigQuery, Google has added a new AI query engine.

“This engine enables data scientists to move beyond simply retrieving structured data to seamlessly processing both structured and unstructured data together with added real-world context,” Yasmeen Ahmad, managing director of data analytics at Google Cloud, wrote in a blog post.

The AI query engine co-processes traditional SQL alongside Gemini to inject runtime access to real-world knowledge, linguistic understanding, and reasoning abilities, Ahmad added.

Other updates to BigQuery

Expanding its efforts to help data practitioners analyze unstructured data further, Google is adding multimodal tables to BigQuery.

Multimodal tables, currently in preview, will allow enterprises to bring complex data types to BigQuery and store them alongside structured data in unified storage for querying.

“To effectively manage this comprehensive data estate, our enhanced BigQuery governance provides a single, unified view for data stewards and professionals to handle discovery, classification, curation, quality, usage, and sharing, including automated cataloging and metadata generation,” Ahmad wrote.While BigQuery governance is still in preview, automated cataloging, as a feature, has been made generally available. Other updates to BigQuery include the general availability of Google Cloud for Apache Kafka to facilitate real-time data streaming, analytics; and the addition of serverless execution of Apache Spark workloads in preview.