Learn  /  Guides

Claude Data explained: dashboards on Snowflake, BigQuery and Databricks from a chat, and the sharing setting to check first

Your colleagues can now ask Claude a question and get back a live dashboard built on the company warehouse. That is useful, and it also changes who can see what. Here is how it works and what to check before 19 November.

Guide8 min readUpdated 16 November 2026
AIData analysisSnowflakeDatabricksWhat's new
All guides and cheat sheets

On 12 November 2026 Anthropic made Claude Data generally available, together with two betas, Claude Sheets and Claude Whiteboards. Claude Data, previously called Claude Dashboards, connects Claude to your data warehouse: it writes and runs the queries, then turns the results into something people can read and share. On Enterprise plans it switches on by default on 19 November, unless an owner has already chosen a setting.

5WarehousesSnowflake, BigQuery, Databricks, Redshift, ClickHouse
19Novemberdefault-on date for Enterprise plans
4Output typesdashboard, data story, video, written analysis
2Betas to decide onlive sharing and certified queries

How it works

AskA question in plain English, in a Claude chat
QueryClaude writes SQL and runs it on the connected warehouse
BuildCharts, a dashboard, a data story or a written analysis
ShareOthers open it; it re-runs the queries when the data is due

Two details matter for anyone who owns the data. First, by default the queries run with the access of the person viewing, not the person who built the dashboard. If your warehouse hides rows or columns from some users, they stay hidden. Second, there is no schedule: a dashboard re-queries only when someone opens it and its data is due, so the numbers are as fresh as the last visit. Each chart shows its SQL and its last refresh time, which makes checking much easier.

The two betas: certified queries and live sharing

Certified queries

Trusted numbers
  • Mark a query as approved, so people know which revenue figure to trust
  • Editing a certified query removes the certification automatically
  • Can be managed in Git with pull requests, or drawn from dbt, Looker or Cube definitions through MCP connectors
  • Anyone with write access to the Data artifact can certify, unless you manage it in Git

Worth switching on, ideally with the Git option so changes are reviewed.

Live sharing

Share results widely
  • People without warehouse access can see a dashboard
  • The queries cannot change without an editor's approval
  • Results run with the sharer's access, so viewers may see data they could not query themselves
  • It is not yet clear whether the Compliance API logs sharing and certification events

Keep it off until you have decided who may share what.

The admin checklist before 19 November

  1. 1

    Decide on the default

    In Organization settings, Artifacts, Templates, choose whether the Data template is on. If you save a setting before 19 November, the default change does not apply.

  2. 2

    Limit who gets it

    Use custom roles to give Claude Data to specific groups first, for example the analytics team.

  3. 3

    Use read-only warehouse users

    Connect with roles that can only read the schemas you intend, never an admin user.

  4. 4

    Leave live sharing off

    Turn it on later, for named groups, once you have a rule for what can be shared.

  5. 5

    Plan certified queries

    Start with the ten numbers people argue about most, and manage them in Git if you can.

  6. 6

    Tell people about freshness

    Dashboards update when opened, not on a timetable. Anything that must be current at 9am needs a proper BI tool or a scheduled job.

How it compares with the platforms' own assistants

Every warehouse now has its own AI analyst, and most of them also reach into chat tools. The choice is less about the model and more about where governance lives.

Where you askBusiness definitions fromBest when
Claude DataClaudeCertified queries, dbt, Looker or CubeYour data sits in several warehouses
Databricks Genie OneDatabricks, Slack, Teams, ChatGPT, Claude, CursorUnity Catalog, Genie Ontology, PagesMost of your data is in Databricks
Snowflake Intelligence and CoWorkSnowflakeSemantic views, Business OntologyMost of your data is in Snowflake
Conversational analytics in BigQueryBigQuery, LookerKnowledge Catalog, LookMLYou are a Google Cloud shop

These are not exclusive. Databricks' new Genie connector for Claude (11 November) lets Claude ask Genie, so the Databricks definitions do the work while people stay in Claude. More in Databricks Genie One's November updates.

Claude Sheets and Whiteboards

The two betas launched alongside. Claude Sheets creates spreadsheets that export as .xlsx or CSV; for now it has only basic spreadsheet functions, so complex models still belong in Excel or Google Sheets. Claude Whiteboards is a shared canvas for planning. Team plans have both on by default; Enterprise owners must turn them on. Our guide to AI in spreadsheets covers the alternatives.

Check your understanding

1/3By default, whose access do Claude Data queries use?

2/3What does the live sharing beta change?

3/3When does a Claude Data dashboard refresh?

More: ChatGPT, Claude or Gemini for data analysis, a security checklist for AI agents on your data and why AI agents need a semantic layer.

Sources: BERI on Claude Data's default-on date and admin controls (November 2026); ThursdAI's November 2026 release list (Claude Data, Sheets and Whiteboards, 12 November 2026); Databricks release notes on the Genie connector for Claude (11 November 2026). Plan names, defaults and beta features may change; check Anthropic's admin documentation for your plan.

Alessandro Ecclesie AgazziWritten by Alessandro Ecclesie Agazzi, analytics engineer in London. Updated 16 November 2026. About me

Found this useful?

New guides and "what's new" explainers on data and AI land here regularly. Follow along on LinkedIn, or share this one with a colleague.