Claude Dashboards

Set up Claude Dashboards beta, inspect SQL and refresh timestamps, check a sales metric, and decide when to keep your BI tool.

Published Updated

Claude Dashboards

Claude Data (formerly Claude Dashboards) lets you ask questions of your company’s connected data and get charts you can inspect and share. Since November 10, 2026, it is generally available, while certified queries and live sharing remain in beta. Start with one metric you can check against a trusted report before using it in a meeting. Anthropic’s Whiteboards, Sheets and Data announcement

What Claude Data Does for Claude Dashboards Users

You connect data, describe a question and get a dashboard built around it. A data warehouse is the central store where your company gathers records for analysis. A CRM holds customer and sales records. Claude turns your question into SQL, the instructions a database uses to select and count those records, and runs it against the connected source.

Think of the chart as a receipt with its calculation attached. You can open the query behind a number and inspect the chart’s last refresh. The visual is useful; the visible calculation is what makes review possible. Anthropic positions this for exploratory questions, with deeper analysis continuing in your existing analytics tools.

The getting-started guide names Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake, plus connected apps such as Salesforce. That is the documented set of examples, not a promise that any CRM or database works automatically. Claude Data getting-started guide

Your existing dashboards keep working under the new name. The added capabilities change how a checked number travels through the business:

  • Certified metric definitions (beta): Your data team can approve queries for reuse, with a certified badge on charts built from them. Editing a certified query removes its certification, so review the revised definition before treating it as official again.
  • Data stories: Ask Claude to turn metrics into a story or narrated short video, then review the result before sharing it.
  • Live sharing (beta): Colleagues in your organization can see approved query results from your connection without having the connector themselves. They cannot use your access to run arbitrary queries.
  • Design systems: Apply your organization’s visual style to the output.
  • Caching and refresh cadence: Set a cadence per dashboard rather than assuming every opening reloads everything. The Data guide says data refreshes when someone opens a dashboard and it is due, not on a background schedule.
  • Tracked changes and rollback: See who edited the work and restore a previous version when a change needs reversing.

The same launch introduced Claude Whiteboards, a shared brainstorming board, and Claude Sheets, a spreadsheet Claude builds and edits with you. Both remain in beta on paid plans; they are on by default for Team and off by default for Enterprise until an owner enables them.

Clay plaques show a data source and a question feeding SQL, then a chart with a last-refresh indicator.
A chart has two things to inspect: its calculation and its freshness. The bars and clock are conceptual, not measured results or a refresh schedule.

Which Plans Include It

Claude Data is available on Pro, Max, Ultra, Team and Enterprise, not Free. It works on web, desktop and mobile and needs a data connector, such as a warehouse or CRM. Dashboard work counts toward your plan’s existing usage limits; it is not a separate unlimited reporting allowance.

On Team, Data, certified queries and live sharing are on by default. On Enterprise, the announcement and Data guide say Data turns on by default on November 19, 2026, unless an owner has already saved a setting. Certified queries and live sharing are separate beta settings and remain off by default on Enterprise.

Anthropic’s pages disagree on one date: the admin guide lists November 20, 2026 for Enterprise’s Data default. Check the actual setting with your owner rather than planning access around that discrepancy. Owners control Data under Organization settings > Artifacts > Templates, with the beta switches under Data. Artifacts admin guide

If you already have an eligible plan, a pilot need not add another software subscription. Its cost still includes setup, checking the query, consumption of your Claude allowance and any costs attached to your data source. Compare the full time spent, including review, with your existing reporting process.

Set It Up From Data Connection to Sharing

Use Claude’s web or desktop app to change sharing settings; the launch also supports using Data on mobile. A dashboard lives as an artifact, Claude’s saved, editable output beside a conversation. You can return to it from the Artifacts tab. Artifacts require Cloud code execution and file creation (called Create and edit files in the updated Settings layout), under Settings > Capabilities for individual plans or Organization settings > Capabilities for Team and Enterprise. Artifacts setup

1. Have the Owner Enable the Right Features

On Team or Enterprise, an Owner or Primary Owner should check cloud code execution, turn on Artifacts, and enable Data under Organization settings > Artifacts > Templates if needed. Enterprise custom roles can restrict access to particular groups. Check the separate certified-query and live-sharing switches under Data. For connected apps inside artifacts, also check Organization settings > Capabilities > Visuals > Enable artifact connectors.

2. Connect the Data You Will Actually Use

Open Customize > Connectors, select +, find the service, review its capabilities, then choose Connect or Install and complete authentication. On Team and Enterprise, an owner first makes the connector available through Organization settings > Connectors > Browse connectors > Add to your team. Users then authenticate, unless the organization uses managed authentication. Enable the service for the conversation through + > Connectors. Connector setup

For a warehouse, involve whoever owns its connection and table permissions. The Data guide names platforms but does not supply one universal warehouse credential form. Follow the selected connector’s requirements. My recommendation for the first pilot is access to one approved reporting table or a view that presents a saved query as a table, with only the permissions needed to read it.

3. Ask a Business Question

Open Artifacts, select Data and describe the question, or ask in chat. Include the table or app, the metric definition and the period. Refine the result through chat or edit a title, caption or chart directly, then open the chart’s query and correct any mismatch.

4. Check the Recipient’s View Before Wider Sharing

Open Share and inspect the available audience and access settings. Without live sharing, connected parts use the viewer’s own connections; missing source access produces an error. With live sharing enabled, open Sources and select Your access next to each MCP connector to let organization colleagues see approved results using your access. Review the rows and fields exposed before doing that.

Internal recipients sign in to Claude. Outside email invitees cannot use parts dependent on Claude or your connectors, and a public link does not make those connected parts work for outsiders. Do not treat live sharing as permission to distribute connected results outside your organization. Artifact sharing rules

First share with one intended colleague and have them confirm what they can see. A successful view in your account does not establish that theirs works.

5. Set Freshness Expectations and Keep a Recovery Path

Choose a refresh cadence appropriate to the reporting question. Cached results make opening faster, but the Data guide says a refresh occurs on opening when the data is due. Do not promise unattended scheduled updates. Check each chart’s last refresh as well as the source table’s freshness.

To undo an edit, open the dashboard’s version history, select an earlier version to preview it, then click Restore. The preview is read-only; restoring changes the dashboard for everyone and appears as its own history entry. View-only recipients cannot restore versions.

Worked Example: Closed-Won Sales by Region

An operator at a small software company wants to see which regions contributed booked sales last quarter. Use closed-won opportunity value, the value of deals marked won, rather than calling it accounting revenue. A signed deal, an invoice and recognized revenue are different measurements.

For this example, assume BigQuery contains a table named crm.opportunities, with one row per opportunity, an is_won field set to true or false, a region, an amount_usd converted to a common dollar basis and a closed_date stored as a date in the company’s agreed business timezone. These are illustrative field names and assumptions, not an Anthropic schema.

The Question You Type

Build a dashboard answering: which regions contributed the most closed-won opportunity value in Q3 2026? Use crm.opportunities. Sum amount_usd for is_won = TRUE, with closed_date from 1 July 2026 inclusive to 1 October 2026 exclusive, grouped by region. Label it “Closed-won opportunity value, USD”, not recognized revenue. Explain how the query handles missing regions and null amounts before I share it.

This prompt pins down the metric, source, currency and date boundary. It also asks for the awkward cases that a polished chart can hide.

The Query You Check

Open the generated query. Under those assumptions, its essential logic should resemble the following illustrative BigQuery SQL, which has not been run against a real dataset:

SQL
SELECT region, SUM(amount_usd) AS closed_won_usd
FROM `crm.opportunities`
WHERE is_won = TRUE
  AND closed_date >= DATE '2026-07-01'
  AND closed_date < DATE '2026-10-01'
GROUP BY region
ORDER BY closed_won_usd DESC;

It selects won deals, keeps close dates inside the quarter and adds their dollar values by region. The exclusive upper boundary keeps 1 October out. It contains no join, so there is no extra table multiplying rows, but duplicates already present in the source could still inflate the sum.

Do not use this unchanged if your warehouse stores opportunity history, multiple line items per deal or timestamps instead of dates. Have the data owner settle the correct record and date rules first. SQL that runs successfully can still answer the wrong question.

What You Verify Before Sharing

  • Count once. Confirm each opportunity appears once in the source. Inspect several deal IDs and reconcile the overall total with the trusted CRM report using the same filters.
  • Use the intended dates. Check deals near the quarter boundary. Confirm that close date, rather than creation date or invoice date, is the intended rule.
  • Use one currency definition. Confirm how amount_usd is produced. SUM skips null amounts, meaning missing values, so investigate them; decide how to display missing regions too.
  • Check freshness at both ends. Read each chart’s last refresh and check when the warehouse last received CRM changes. A recently queried stale table still yields stale numbers.
  • Check access and meaning. Have a colleague open the shared dashboard with the intended access mode, their own connection or approved live sharing. Include the metric definition and any exclusions so a sales total does not get reused as accounting revenue.
Four clay inspection stations labeled Count once, Dates, Currency and Viewer access.
Recommended checks for the sales example. These are review tasks, not Claude interface controls.

If the totals disagree, resolve the discrepancy before adding more charts. Useful follow-up questions include “Which records explain the difference?” and “Does the source contain historical versions of the same opportunity?” The first dashboard earns trust through reconciliation, not appearance.

Six Useful Jobs, Ranked by Likely Payoff

These are proposed uses when the relevant fields already exist in a connected source. They are not reported customer results or additional connector claims.

Rank and operatorExact workflow to tryWhy it could pay
1. Sales operations lead before a pipeline meetingGroup open Salesforce opportunities by owner and stage, then inspect what qualifies as open.Puts a changing meeting question beside the underlying records.
2. Founder with a clean sales tableBuild the closed-won regional view above and reconcile it with the existing sales report.Could remove a repeated export and chart-building task.
3. Support manager whose ticket data already reaches the warehouseGroup unresolved tickets by queue and age, using the team’s agreed unresolved statuses.Could make staffing discussions start with the backlog rather than spreadsheet assembly.
4. Growth operator with signup recordsCompare completed weekly signup periods by acquisition channel.Could expose where a channel deserves investigation without a fresh reporting request.
5. Customer success lead with renewal recordsGroup upcoming contract renewals by owner and month.Could help organize follow-up around an agreed renewal list.
6. Analyst scoping a future BI reportExplore a candidate metric with its stakeholder, then carry the agreed definition into the reporting system.Could reduce rework caused by building the wrong permanent report.

The best first job has a clear source, a known comparison total and someone responsible for its definition. A broad “show business health” request has none of those advantages.

Where It Fits Beside Looker Studio, Metabase and Power BI

My recommendation is to use Claude Data for questions still taking shape. Its certified queries, refresh cadence and version history now make it more useful for repeatable internal reporting too. That does not settle whether it meets your existing reporting requirements.

If your company already relies on Looker Studio, Metabase or Power BI, compare the pilot against that established workflow: who owns the metric, who can view the result, how freshness is checked, and how changes are approved. Keep the existing report until the new workflow meets those requirements. This is a decision rule, not a feature comparison or a claim that Claude exports directly to these products. For a broader reporting shortlist, see Databox alternatives.

What It Cannot Settle for You

It cannot turn an ambiguous metric into an agreed business definition. That is my practical assessment of this workflow: “revenue,” “active customer” and “resolved ticket” need an owner who decides what counts. Nor does drawing a chart repair missing records upstream.

A live connection is not a freshness guarantee. Caching and a per-dashboard cadence do not mean a dashboard refreshes in the background on a schedule. A newly queried stale warehouse table still produces stale numbers. The Data guide does not publish a fixed dashboard, chart or row quota; do not turn an unpublished limit into an unlimited-capacity claim.

Live sharing is a deliberate data-access decision. Ordinary connector access follows the user’s source permissions; shared credentials carry that credential’s access. Live sharing adds a specific exception for approved dashboard results inside your organization. Review the exposed results and audience, not just whether the dashboard opens.

I would keep contractual freshness commitments and audited reporting in a system your team already operates for those requirements until you validate this workflow against them. Certified queries provide a way to reuse an agreed definition, not proof that the underlying records or business assumptions are correct.

Verified against Anthropic’s announcement and help pages on November 13, 2026. This guide uses documented behavior, not signed-in testing.

Two Things Worth Building Around It

The opportunities here are services and supporting assets. The getting-started guide does not establish a developer API for the Data feature, so neither proposal depends on one.

Strongest: A Sales Metric Setup and Validation Service

A small company could pay a specialist to turn one messy sales question into a documented metric, an approved warehouse view, a Claude dashboard and a reconciliation checklist.

The smallest sellable engagement covers one sales metric, a known comparison report, connection setup and a handover to the person who owns the number. The catch is that source cleanup may dominate the work. Scope that separately. This is the strongest opportunity because the valuable deliverable is an agreed, checked number, something a generic chart generator cannot supply on its own.

Second: A Dashboard Review Starter Kit

A consultant could build a reusable kit for operators: metric-definition worksheets, prompt examples, query-review questions and a sharing acceptance checklist. Offer an assisted first review before trying to sell a standalone product.

An initial kit could cover the sales example and a support backlog example with clearly stated data assumptions. Its weakness is easy imitation and free alternatives. Industry-specific definitions and thoughtful review would need to justify the purchase. For packaging repeatable reporting instructions, see Claude Skills for weekly client reports.

How to create a sales dashboard?

Choose the decision first, then define the records, date field and metric. Connect the source, build one chart and reconcile its total with a trusted report before expanding it. Use the setup and worked example above as the pilot.

What are sales metrics?

They are measurements of sales activity or outcomes, such as open pipeline value and won deals. Define the unit, period and exclusions. Closed-won opportunity value and recognized revenue should not share a label merely because both use currency.

What is KPI in a dashboard?

A key performance indicator is a metric chosen to judge progress toward a business objective. A dashboard can contain many measurements; the KPI should connect to a decision and have an owner responsible for its definition.

What is a good KPI dashboard for sales?

One the sales team can explain and act on. Start with the decision your meeting needs to make, use agreed definitions and show enough context to distinguish a real change from a date, currency or data-quality issue.

Your Monday Move

Pick the operator who currently exports the weekly sales report. Pair them with the person who owns its source table. Build only the closed-won view, compare totals, choose a refresh cadence, check the timestamp and have the intended colleague open it. Decide explicitly whether they should use their own connector or live sharing. Record the full time spent, including corrections. Expand only if the pilot makes that reporting job easier without losing a clear definition of the number.

If you want a checked reporting workflow built around your data, we can help with AI automation.

Last Updated
Category
AI
Related Articles
Gemini 4 Neon: What It Does, What It Costs and How to Start

Gemini 4 Neon: What It Does, What It Costs and How to Start

Where Gemini 4 Neon fits beside Carbon, Hydrogen and Argon, who can use it, and how to compare it with your current model on a real coding task.Nov 19, 2026AI
Always-On Agents: Anthropic's Five Primitives, Shown Working

Always-On Agents: Anthropic's Five Primitives, Shown Working

An always-on agent mostly sleeps. Anthropic's five primitives wake it, carry its memory, guard its actions and deliver its work. Costs and limits included.Nov 19, 2026AI
Claude Cowork Pricing: What One Seat or a Team Costs

Claude Cowork Pricing: What One Seat or a Team Costs

Which Claude plans include Cowork? Compare US prices, usage limits, and worked budgets for one person, five seats, or twenty seats.Nov 19, 2026AI
QuillBot Review

QuillBot Review

When does QuillBot earn a seat beside your AI assistant? Compare its editing tools, US prices, team controls and privacy terms for client work.Nov 19, 2026AI
GPT-6.2 Astra vs Sol

GPT-6.2 Astra vs Sol

Choose a GPT-6.2 default by job. Compare Astra and Sol for agent runs, code review and writing, with worked examples and Ultrafast access by plan.Nov 18, 2026AI
GPT-6.2 Explained: Astra, Sol and Who Gets Access

GPT-6.2 Explained: Astra, Sol and Who Gets Access

Find GPT-6.2 Astra and Sol in ChatGPT Work, Codex and the API. See who gets Ultrafast and which jobs justify switching models.Nov 18, 2026AI
Kimi vs Qwen

Kimi vs Qwen

Kimi K3.1 Pro vs Qwen4-Max for coding agents and products: current access, long context, open weights, API tradeoffs and US deployment choices.Nov 17, 2026AI
Qwen 4: Qwen4-Max Explained

Qwen 4: Qwen4-Max Explained

Qwen4-Max explained: production versus preview access, context and output limits, open-model promises, and when to move from Qwen3.8-Max.Nov 17, 2026AI
Newsletter

One letter, every Sunday.Working systems, not hot takes.

Weekly. No spam. Unsubscribe anytime.