Home / AI News / BigQuery Agent Analytics

BigQuery Agent Analytics and Looker: Enterprise AI Agent Observability and Governance

Google Cloud announced expanded AI agent observability and evaluation capabilities across BigQuery and Looker on November 14, 2026, unifying telemetry for 1P and custom ADK or LangGraph agents.

Spark5X · · Google Cloud

Google Cloud announced expanded AI agent observability, governance, and evaluation capabilities across BigQuery and Looker on November 14, 2026. Authored by Google Cloud Product Managers Kate Grinevskaja and Snigdha Jha, the release introduces a unified telemetry framework across Google Data Cloud that captures execution traces, conversation logs, and token usage from first-party (1P) Conversational Analytics agents in BigQuery and Looker and custom third-party (3P) agents built with Agent Development Kit (ADK) or LangGraph.

Key takeaways
  • Google Cloud rolled out a unified telemetry pipeline on November 14, 2026 that lands logs from first-party BigQuery and Looker agents and custom ADK or LangGraph agents into customer-owned BigQuery tables (agent_events).
  • Now generally available (GA): agent-to-agent (A2A) telemetry routing in Gemini Enterprise, managed BigQuery conversation logging, the Conversational Analytics API, BigQuery Studio agent monitoring dashboards, and Looker System Activity dashboards for Conversational Analytics.
  • Privacy boundaries separate sanitized Cloud Monitoring and Cloud Trace metrics from per-agent conversation logs, with fail-closed checks that skip logging when row-level security, column-level policy tags, or dynamic data masking are present.
  • FinOps and evaluation tools add per-query BigQuery byte scan caps (big_query_max_billed_bytes), daily token quotas in preview, native BigQuery evaluations in preview, and the open-source Prism benchmarking tool.

1. What is BigQuery Agent Analytics and Google Cloud's new agent observability framework?

Announced on November 14, 2026, Google Cloud's agent observability framework is a unified telemetry, governance, and evaluation system across Google Data Cloud. It captures operational metrics and conversation logs from first-party BigQuery and Looker agents and custom ADK or LangGraph agents, landing them in customer-owned BigQuery tables.

Running production agents requires solving three operational challenges: tracing multi-step reasoning while protecting sensitive data, controlling token consumption and query spend, and evaluating answer accuracy before prompt or schema changes reach users. Google Cloud addresses all three in a three-step workflow:

  • Capture telemetry from every agent: Out-of-the-box Conversational Analytics agents emit telemetry automatically. Queries routed to BigQuery data agents in Gemini Enterprise through agent-to-agent (A2A) workflows (now GA) send Cloud Monitoring metrics, Cloud Trace spans, and BigQuery conversation logs into the agent's project. Custom ADK or LangGraph agents stream structured telemetry into BigQuery via single-line SDK instrumentation.
  • Land telemetry in one governed BigQuery home: Managed 1P conversation logging (GA) writes prompts and responses to customer-owned BigQuery tables. The Conversational Analytics API (GA) includes always-on admin audit logs and optional data-read audit logs in Cloud Audit Logs, with VPC Service Controls, customer-managed encryption keys (CMEK), and regional and multi-regional data residency.
  • Analyze with dashboards: Teams get prebuilt Agent monitoring dashboards in BigQuery Studio (GA), System Activity dashboards in Looker (Google Cloud core) and Looker (original) (GA, with the Responses & feedback tab in preview), an open-source Data Studio template, and a Looker Block for Agent Analytics.

2. How does Google Data Cloud trace agent reasoning while protecting sensitive data?

Google Data Cloud separates environment-wide operational telemetry in Cloud Monitoring and Cloud Trace—which exclude prompts, responses, internal thoughts, and generated SQL—from opt-in per-agent conversation logging in BigQuery. Built-in privacy guardrails enforce fail-closed checks that skip logging whenever protected data policies apply.

Admins can enable Cloud Monitoring and Cloud Trace at the project or organization level, using Cloud Trace observability scopes to consolidate telemetry across agent projects without exposing customer core content. For deeper transcript auditing, agent owners and editors can opt in per agent:

  • Governed agent_events logging: A Google-managed service agent streams prompts, responses, and turn-level telemetry into a customer-owned BigQuery table (agent_events) via the BigQuery Storage Write API, inheriting dataset IAM permissions, CMEK, and VPC Service Controls. Fail-closed checks skip logging whenever a knowledge source uses row-level security, column-level policy tags, or dynamic data masking, while fail-open execution keeps chat sessions uninterrupted if a logging error occurs.
  • Full trace trees for custom 3P agents: Developers building custom agents on ADK or LangGraph can add the open-source ADK plugin (BigQueryAgentAnalyticsPlugin) to stream full tool-invocation execution trees, time-series telemetry, and multimodal payloads into BigQuery.
  • In-warehouse root cause and drift analysis: Teams can run BigQuery AI functions (AI.GENERATE to summarize error patterns and AI.EMBED with VECTOR_SEARCH to cluster recurring issues) or use the open-source bigquery-agent-analytics Python SDK and Colab starter notebook for semantic drift clustering and LLM-as-a-judge sentiment analysis.
Capability or componentStatusWhat it provides in Google Data Cloud
Conversational Analytics API & A2AGARoutes queries to BigQuery data agents in Gemini Enterprise with Cloud Audit Logs, CMEK, VPC Service Controls, and data residency.
Managed 1P conversation loggingGAStreams prompts, responses, and turn-level token and latency metrics into customer-owned agent_events tables with fail-closed privacy checks.
BigQuery Studio & Looker dashboardsGA (feedback tab preview)Tracks usage, active sessions, knowledge sources, and latency in BigQuery Studio, plus engagement and token usage in Looker System Activity.
BigQueryAgentAnalyticsPlugin & SDKOpen-sourceSingle-line ADK/LangGraph instrumentation plus a seven-page Data Studio template, Looker Block, Streamlit app, and Colab notebook.
FinOps byte caps & token quotasCaps live; quotas previewConfigures big_query_max_billed_bytes dry-run query scan caps and sets daily input and output token quotas per project, user, or agent.
Prism & native BigQuery evaluationsPrism OSS; native previewTests agents against BigQuery tables and Looker Explores via Prism, streaming benchmark runs into day-partitioned BigQuery tables.

3. How do FinOps teams control token consumption and BigQuery scan spend?

Google Cloud provides granular turn-level token logging in BigQuery and Looker alongside active spend guardrails. Administrators can configure per-query BigQuery byte scan caps using dry-run estimates to block oversized queries before execution, and enforce daily input and output token quotas per project, user, or agent.

  • Granular token and latency tracking: Each turn in BigQuery conversation logs records input, output, thought, cached, and total token counts alongside total turn latency, while Looker System Activity reports estimated input and output tokens across BI workloads.
  • Turnkey dashboards: Teams can review projected token usage in BigQuery Studio and Looker System Activity, or query agent_events via BigQuery SQL using the seven-page Data Studio template and self-hosted Streamlit app.
  • Active query and token caps: Setting big_query_max_billed_bytes uses dry-run estimates to block oversized queries without incurring scan charges, while daily input and output token quotas (in preview) can be enforced per project, user, or agent in the Cloud console and gcloud CLI.

4. How do Prism and native BigQuery evaluations benchmark agent accuracy?

Google Cloud's open-source evaluation tool, Prism, tests Conversational Analytics agents grounded in BigQuery tables or Looker Explores using reusable test suites and regex assertions, then streams results into day-partitioned BigQuery tables. Native first-party evaluations in BigQuery are also available in preview.

  • Benchmarking with Prism: Prism (GoogleCloudPlatform/conversational-analytics-prism) evaluates agents via the Conversational Analytics API using reusable test suites, regex assertions, and multi-region agent discovery behind Cloud IAP, streaming completed runs into day-partitioned BigQuery tables alongside agent_events logs.
  • Native BigQuery evaluations: Built on the Gemini Enterprise Agent Platform evaluation service with bring-your-own-service-account support, native first-party evaluations in BigQuery are available in preview via API and a guided Colab notebook.

5. How can enterprise data teams get started with BigQuery and Looker agent observability?

Teams can start in three steps: enable Cloud Monitoring, Cloud Trace, and opt-in per-agent conversation logging for first-party BigQuery and Looker agents; add the BigQuery Agent Analytics SDK to custom ADK or LangGraph agents; and configure Prism benchmarks alongside byte scan caps and daily token quotas.

Alongside Google's Gemini agent, Google AI Search with Gemini 4 Hydrogen, Google Cloud credits in Google AI Studio, and Claude Data, data teams can connect agent telemetry with analytics and attribution, AI SEO, and AI News workflows.

Sources. This article is based on the company's own announcement:

Frequently asked questions

What did Google Cloud announce for AI agent observability across BigQuery and Looker?

On November 14, 2026, Google Cloud announced expanded AI agent observability and evaluation capabilities across Google Data Cloud. The unified telemetry framework captures logs from first-party Conversational Analytics agents in BigQuery and Looker and custom third-party agents built with Agent Development Kit (ADK) or LangGraph, landing them in customer-owned BigQuery tables.

Which BigQuery and Looker agent observability features are now generally available?

Google Cloud states that agent-to-agent (A2A) telemetry routing in Gemini Enterprise, managed first-party conversation logging in BigQuery, the Conversational Analytics API, prebuilt Agent monitoring dashboards in BigQuery Studio, and System Activity dashboards for Conversational Analytics in Looker (Google Cloud core) and Looker (original) are now generally available, while Looker's Responses & feedback tab is in preview.

How does Google Data Cloud protect sensitive data in agent traces and conversation logs?

Google Data Cloud separates Cloud Monitoring metrics and sanitized Cloud Trace spans—which exclude user prompts, model responses, internal thoughts, and generated SQL—from opt-in per-agent conversation logging in BigQuery (agent_events). Built-in fail-closed privacy guardrails skip logging whenever a knowledge source uses row-level security, column-level policy tags, or dynamic data masking, while fail-open execution keeps user chats uninterrupted.

How do FinOps teams control token consumption and BigQuery query spend for agents?

In BigQuery conversation logs, each turn records input, output, thought, cached, and total token counts alongside turn latency, while Looker System Activity reports estimated input and output tokens. Administrators can configure per-query BigQuery byte scan caps (big_query_max_billed_bytes) using dry-run estimates to block oversized queries and enforce daily input and output token quotas (in preview) per project, user, or agent.

How do Prism and native BigQuery evaluations benchmark AI agents?

Google Cloud's open-source evaluation tool, Prism, tests Conversational Analytics agents grounded in BigQuery tables or Looker Explores via the Conversational Analytics API using reusable test suites, regex assertions, and multi-region agent discovery, then streams runs into day-partitioned BigQuery tables alongside production agent_events logs. Native first-party agent evaluations in BigQuery are also available in preview via API and a guided Colab notebook.

More AI news

All AI news →

Get started
Put this to work
Talk to us about using new AI tools in your marketing and growth systems.
    0/1000
    Prefer to pick a time? Book a call →
    🔒 We'll only use your details to reply. See our Privacy Policy.
    Keep reading

    Related AI news

    Request a quoteBook a call