BigQuery makes conversational analytics in Connected Sheets generally available

Google Cloud has moved conversational analytics in Connected Sheets to general availability. The BigQuery release notes entry dated October 27, 2026 describes a Gemini side panel in Google Sheets that takes questions written in plain English, translates them into optimized SQL, runs those queries against BigQuery, summarizes the main trends in the results and can insert live, refreshable database objects into the workbook.

The Google Workspace side of the launch was announced on October 19, 2026, when Google said that users connected to BigQuery through Connected Sheets could ask Gemini about that data in natural language. According to that announcement, Gemini queries the full BigQuery dataset rather than only the rows visible in the spreadsheet, supports follow-up questions and can build a new tab with connected pivot tables, charts and database formulas after showing the user a plan to review.

For BI teams the change touches a familiar pressure point. Connected Sheets has long been the way business users reach warehouse tables without writing SQL, and Google presents the new panel as a way for those users to answer more of their own questions while data teams handle fewer one-off pulls. The practical questions are who gets the feature, which permissions and billing project it relies on, and how an administrator can see what it runs.

··········

WHAT CONVERSATIONAL ANALYTICS IN CONNECTED SHEETS DOES.

The BigQuery documentation describes a short workflow. The user opens the Gemini side panel in a spreadsheet that is connected to BigQuery data. On first use the panel may ask the user to authorize BigQuery access through a Get connected prompt. The user then types a question, can keep context across several questions, asks Gemini to build live objects, and clicks Insert to place the results in the spreadsheet. Those objects can be refreshed manually at any time or on a schedule.

Google's Sheets help center adds three details that shape how the feature behaves in daily use. Gemini runs read-only queries on the connected BigQuery data source and does not modify or delete data. Before acting, Gemini shows a plan card that the user reviews and confirms with Apply, and the action can be undone until further changes are made. Charts, pivot tables, formulas and extracts created this way stay connected to BigQuery, so they can be refreshed later, while charts that Gemini builds from ordinary spreadsheet data do not update when the source cells change.

The connected objects are the main difference from asking a chatbot about an exported file. A pivot table inserted by Gemini remains a Connected Sheets object that reruns its query against the warehouse, which keeps the result tied to the table instead of a static copy. Readers who have used Gemini on regular sheets will recognize the side panel from the features described in Gemini for data analysis and chart generation in Google Sheets; the new element is the generated SQL running on BigQuery.

··········

BIGQUERY CONNECTED SHEETS AVAILABILITY BY WORKSPACE EDITION.

The Workspace announcement lists the editions and add-ons that include the feature, and it states that the rollout is gradual for both Rapid Release and Scheduled Release domains, starting on October 19, 2026, with up to 15 days before the feature is visible.

........

Item	As published by Google
Release status	Generally available (BigQuery release notes, October 27, 2026)
Business editions	Business Standard, Business Plus
Enterprise editions	Enterprise Standard, Enterprise Plus
Consumer plans	Google AI Pro, Google AI Ultra
Education	Google AI Pro for Education add-on
Other add-ons	AI Expanded Access
Admin control	None for this feature
Rollout	Gradual from October 19, 2026, up to 15 days

........

Google has not published a separate price for conversational analytics in Connected Sheets on the pages that announce it. The queries Gemini writes are BigQuery queries, and Connected Sheets requires a Google Cloud project with billing enabled, so the warehouse cost follows the project and pricing model the organization already uses for Connected Sheets.

··········

PERMISSIONS AND BILLING PROJECT FOR GEMINI IN CONNECTED SHEETS.

The feature sits on top of the normal Connected Sheets prerequisites. Google's help center lists access to Google Cloud, BigQuery access and a project with billing set up in BigQuery, and it notes that a Workspace administrator must turn on Google Cloud for the organization. Access to individual datasets and tables is controlled in BigQuery itself, so Gemini can only reach what the user is already allowed to query.

The permission model also decides what colleagues see in a shared workbook. People without BigQuery permissions can still view saved results, but running new queries or refreshing objects requires BigQuery access. The BigQuery documentation states the same principle for Connected Sheets in general: users who only have access to the sheet can analyze the data already in it but cannot refresh it manually or schedule refreshes.

Administrators should note one point from the Workspace announcement. There is no admin control for this feature, so in an eligible edition the panel becomes available to users who already have a connected sheet and the rights to query its source. Teams that want to limit its use have to rely on BigQuery permissions, project billing settings and the organization's existing Gemini policies.

··········

CONNECTED SHEETS LIMITS THAT STILL APPLY TO GEMINI RESULTS.

Gemini's output lands in standard Connected Sheets objects, so their documented limits carry over. The sheet previews 500 rows while pivot tables, formulas and charts work on the full dataset. Data extracts of 50,000 rows or fewer have no cell limit, extracts between 50,000 and 500,000 rows are capped at 5 million cells, and pulls above 500,000 rows are not supported. For pivot tables the BigQuery documentation currently gives two different maximums for returned results, 200,000 rows in one section and 100,000 in another, so analysts planning large pivots should test against their own data.

........

Area	Documented behavior
Queries run by Gemini	Read-only, no edits or deletes
Sheet preview	500 rows
Extracts up to 50,000 rows	No cell limit
Extracts 50,001 to 500,000 rows	Up to 5 million cells
Extracts above 500,000 rows	Not supported
Editing BigQuery data from Sheets	Not allowed
Create, edit or refresh on mobile	Requires a computer

........

Conversation history in the panel is also fragile. The help center says it is lost when the browser is reloaded, the spreadsheet is closed and reopened, or the user goes offline, so the guidance is to insert useful output into the sheet. For the generated SQL more broadly, the article How ChatGPT writes SQL queries from natural-language prompts covers the same class of translation errors that analysts should watch for here.

··········

GOVERNANCE, VPC SERVICE CONTROLS AND AUDITING.

Organizations that protect BigQuery with VPC Service Controls need to plan ahead, because the BigQuery documentation states that VPC Service Controls does not support Sheets. Connected Sheets queries can be allowed with an access level or ingress rule for trusted identities and devices, plus an egress rule that lets results be copied into spreadsheets, using the bigquery.vpcsc.importData permission against the Sheets-owned project listed in the documentation. Scheduled refreshes do not carry end-user context, so perimeters that rely on that context will see scheduled refreshes fail.

For auditing, queries from Connected Sheets carry the job ID prefix sheets_dataconnector, and Cloud Audit Logs record access to BigQuery data. The BigQuery documentation provides a project-level query over INFORMATION_SCHEMA that aggregates Connected Sheets usage by user, which administrators can run to see who generates the most billed bytes after the rollout:

SELECT
  user_email,
  COUNT(*) AS total_queries,
  SUM(total_bytes_billed) AS total_bytes_billed,
  SUM(total_slot_ms) AS total_slot_ms
FROM
  `region-REGION_NAME`.INFORMATION_SCHEMA.JOBS_BY_PROJECT
WHERE
  creation_time >= TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 30 DAY)
  AND job_id LIKE 'sheets_dataconnector%'
  AND state = 'DONE'
  AND (statement_type IS NULL OR statement_type <> 'SCRIPT')
GROUP BY
  user_email
ORDER BY
  total_bytes_billed DESC
LIMIT
  10;

REGION_NAME is replaced with the project's region, for example region-us, and the query must run in that same location. On privacy, the Sheets help center states that conversations with Gemini in Sheets are not saved to Gemini Activity, and that Gemini content may be visible to people outside the domain if the file is shared with them.

··········

HOW ANALYSTS SHOULD CHECK GEMINI ANSWERS ON BIGQUERY DATA.

Google's own help text warns that Gemini features may suggest inaccurate or inappropriate information, and it points users to the Analysis steps view to see how Gemini reached an analysis or chart. For warehouse data the useful habit is to read the plan card before applying it, confirm which table and filters the question was mapped to, and compare one headline figure with an existing certified report or a hand-written query before sharing the new tab.

Ambiguous business terms deserve particular care. A question about revenue, active customers or last quarter can map to several columns or date definitions, and the SQL that Gemini generates will follow whichever interpretation it chooses. Data teams that already publish curated views for Connected Sheets users will get more consistent answers than teams that expose raw tables, and the same discipline applies to the warehouse agents described in OpenAI launches Data agent in ChatGPT Work.

··········

CONVERSATIONAL ANALYTICS IN CONNECTED SHEETS: WHAT DEPENDS ON YOUR SETUP.

Google confirms that conversational analytics in Connected Sheets is generally available, that it generates and runs SQL on BigQuery through the Gemini side panel, that its queries are read-only, and that its charts, pivot tables, formulas and extracts stay connected and refreshable. The eligible editions and add-ons are listed in the Workspace announcement, and the rollout started on October 19, 2026 with up to 15 days before every user sees it.

What each organization gets depends on its Workspace edition, the BigQuery permissions of each user, the billing project behind Connected Sheets and any VPC Service Controls perimeter around the data. Google has not published a separate price for the feature and offers no admin switch for it, so cost and access are governed through BigQuery itself. Administrators who want visibility can filter job history on the sheets_dataconnector prefix from the first day of the rollout.

··········

FOLLOW US FOR MORE.

·····

DATA STUDIOS

·····

datastudios.org