Gemini in Google Sheets now queries your whole BigQuery dataset. The spreadsheet became an analytics agent

A regional sales director opens a spreadsheet connected to the company’s sales warehouse and types: which accounts in my region had declining revenue last quarter, and why? A few seconds later Gemini proposes a plan, runs it against the full BigQuery table, and builds a new tab with formulas, a pivot table and two charts. It refreshes on a schedule. The director shares it with the team on Monday.

Nobody in the data team saw it. It may be exactly right. It may also count returns as revenue, or use calendar quarters where finance uses fiscal ones. And it now exists, refreshing every morning, owned by someone who never wrote a query.

That is the practical meaning of a Google Workspace update published on October 19.

What changed

Gemini in Google Sheets can now analyze BigQuery data through Connected Sheets in natural language, with follow-up questions. Google is explicit that Gemini’s queries run against the full dataset in BigQuery, not only the rows visible in the sheet. When building a dashboard, Gemini shares its plan for the user to review first, then produces a new tab with connected, refreshable objects (database formulas, pivot tables and charts) kept current by scheduled refreshes.

The prerequisites are the right ones: the sheet must be connected through Connected Sheets, and the user must have permission to query the underlying data. Gemini does not widen anyone’s access.

Three details deserve attention from IT:

• It is on by default, and Google states there is no admin control for this feature.
• The rollout began October 19 for Rapid Release and Scheduled Release domains, with up to 15 days for the feature to appear.
• It is available on Business Standard and Plus, Enterprise Standard and Plus, and several AI add-ons.

The misconception: “it uses our permissions, so it is governed”

Inheriting BigQuery permissions solves one problem, who may see which rows, and it is the problem most governance programs already handle. It leaves three others open, and natural language makes them larger.

Definitions. A warehouse table holds columns, not business meaning. “Revenue”, “active customer” and “churn” each have an official definition somewhere, and a plain-language question does not carry it. Gemini shows its plan, which is the right design, but a plan is only reviewable by someone who knows what the correct definition is. Teams that have put business definitions where agents can read them are much better placed here than teams whose definitions live in a few analysts’ heads.

Cost. Queries now run on the whole dataset, and dashboards refresh on a schedule. BigQuery charges for queries according to each project’s pricing model, and the announcement does not say how Gemini’s queries are metered or capped. A few hundred people exploring large tables in plain language is a different load from a few dozen analysts writing SQL. Check which project pays, and set quotas there, before the feature is in everyone’s hands.

Ownership. Each generated dashboard is a small analytics product: it refreshes, it gets shared, decisions are made from it. Nothing in the announcement gives IT a way to see how many exist. That is the same pattern we described when AI-built internal apps met live business data: building became trivial, so the hard part moved to knowing what was built and who answers for it.

What to do in the next two weeks

Because the rollout is gradual and there is no switch, the decisions have to be made upstream:

• Review who can query what in Connected Sheets. Datasets broadly granted “just in case” become broadly queryable in plain language.
• Publish the definitions of the five to ten metrics people ask about most, next to the tables, and tell users to check Gemini’s plan against them.
• Put cost controls on the BigQuery projects behind Connected Sheets.
• Name a path to promotion. When a generated dashboard becomes something a team relies on, there should be a simple way to hand it to the data team for review rather than letting it become the official number by accident.

The spreadsheet has always been where enterprise analytics escaped governance. Gemini does not change that. It makes the escape much faster.