Google Cloud widens conversational analytics tools

Google Cloud has made BigQuery Conversational Analytics and the Conversational Analytics API generally available. Conversational Analytics in Databases is available in preview.

The move expands Google Cloud's conversational data tools across analytics and database products as companies seek tighter controls over how generative AI interacts with business data.

General availability now covers BigQuery and Looker, while preview support extends to AlloyDB, Cloud SQL and Spanner. Users can also analyse data stored in Lakehouse Managed Service tables, Apache Iceberg REST catalogues and federated AWS S3 Unity Catalogs.

That gives organisations a single conversational layer for querying data in Google Cloud and other cloud environments. The tools are integrated into BigQuery Studio, BigQuery Data Canvas and Database Studio for data teams, and into Looker, Data Studio and Gemini Enterprise for business users.

Broader reach

Data teams can publish conversational agents created in BigQuery, Looker, AlloyDB, Spanner and Cloud SQL into Gemini Enterprise, giving staff a single interface for data queries across multiple systems. The API and Model Context Protocol tools also let companies embed the service into custom applications, multi-agent systems and workplace chat tools.

The announcement reflects a wider push by large cloud providers to move generative AI beyond general chat interfaces and into governed business workflows. Google is positioning conversational access to data as part of its broader data cloud offering.

Controls added

The service includes Customer Managed Encryption Keys, Private IP and Virtual Private Cloud controls. It also provides data residency at rest and machine learning processing within multi-region endpoints in the European Union and the United States, alongside HIPAA compliance.

Access controls are tied to existing database and analytics permissions. According to Google, role-based controls and parameterised secure views in AlloyDB for PostgreSQL let users see only the data they are authorised to access, down to row- and column-level permissions.

Administrators are also getting more tools to monitor usage and spending. Users can set limits on maximum query sizes, track usage through BigQuery query labels and Looker system activity logs, and export health, latency, tool usage and token consumption metrics through OpenTelemetry standards.

Grounding data

A central issue for conversational systems connected to enterprise databases is reducing incorrect or invented answers. Google said it designed these agents alongside the data platforms they query, rather than placing a generic large language model in front of business systems.

The agents use Knowledge Catalog for data discovery, glossaries and context enrichment such as table joins and descriptions. BigQuery Graphs and Spanner Graphs are intended to help agents query relationships across structured and unstructured data, while Looker's semantic layer grounds answers in governed metric definitions.

Google also described an API-first approach using what it calls Golden Queries, intended to let agents retrieve verified business logic rather than infer SQL joins. In Looker, agents can navigate high-cardinality datasets with dynamic filtering, enforce row-level security during chat sessions and suggest context-aware questions.

From queries to alerts

Google is also extending the product from question-and-answer interactions to automated analysis. With Agentic Workflows, in preview, organisations can schedule reporting routines within chat workflows and run anomaly detection on key metrics.

Agents can send daily or weekly summaries to teams and launch automatically when a metric moves outside baseline thresholds. The system can also run multi-step investigations into metric changes by analysing between 10 and 20 contributing factors.

Alongside standard querying, the tools can work with multimodal data through BigQuery object tables and use functions including ai.search, ai.generate_embedding, ai.classify, ai.score, ai.forecast and ai.detect_anomalies. Google added that ai.key_drivers can be used for contribution analysis to identify causes behind unexpected changes in data.

Developer access

The Conversational Analytics API is available with software development kits for Node.js, Java, Go, Python, PHP, Ruby and .NET. Google said this is intended to let developers add conversational analytics functions to existing software environments rather than requiring separate user interfaces.

As an example, Google pointed to a supply chain orchestrator agent querying a financial data agent to calculate the margin effect of a shipping delay in real time. That suggests a market in which conversational analytics tools are increasingly being tied into broader agent-based software systems rather than used only for stand-alone data queries.

Google said conversational analytics is now available across more data sources, more user interfaces and more governed enterprise settings than before.