Claude Dashboards and Motion Guide: Build Live Data Views and Editable Explainers

Checking a business number often means waiting for an analyst, while turning the same result into an explainer can require a second design workflow. Claude Dashboards and Claude Motion, announced on October 8, 2026, bring those two starting points into Claude Artifacts. Dashboards writes SQL against connected data and builds a view that can stay current. Motion turns reports, charts, shapes, and images into a short, code-based animation.

The two features solve different problems and have different plan requirements. Dashboards is in beta on Pro, Max, Team, and Enterprise. Motion is in beta on Team and Enterprise only. This guide covers the exact entry points, how to review the output, practical examples, sharing controls, and when to hand the work back to a dedicated analytics or video tool.

Dashboards and Motion at a glance

Area	Claude Dashboards	Claude Motion
Main output	A dashboard that refreshes as connected data changes	A short animation that can be exported as MP4
Input	A warehouse or connected app plus a question	Reports, charts, images, audience, purpose, and duration
What to inspect	SQL behind each chart and last refresh time	Wording, numbers, scene order, and timing
Best fit	Fast exploratory questions and status views	Meeting, training, and onboarding explainers
Beta availability	Pro, Max, Team, Enterprise	Team, Enterprise

Dashboards is not positioned as a complete replacement for a governed BI stack. It is designed for quick exploratory questions such as how this week's signups compare with last month's. Motion is not a realistic video generator either. It writes code that animates your text, charts, shapes, and images, which makes a single word, number, or timing change easier to revise without rebuilding the whole sequence.

A human review workflow connecting live data dashboards, editable motion explainers, and controlled sharing — Dashboards focuses on data questions and evidence. Motion focuses on sequencing and explanation. A person still needs to verify the numbers and decide what can be shared.

What to prepare before you start

Confirm the plan and admin settings. Dashboards and Motion are off by default on Enterprise. An Owner or Primary Owner must enable Artifacts and the relevant templates under Organization settings → Artifacts. The organization also needs Cloud code execution and file creation enabled under Capabilities before Artifacts can be turned on. Motion is on by default for Team plans, although an organization can change the setting.
Limit the data connection. Dashboards can work with platforms such as Amazon Redshift, BigQuery, ClickHouse, Databricks, and Snowflake, as well as connected apps such as Salesforce. Claude does not create broader permissions than the connected account already has. A read-only account with non-sensitive data is a safer first test.
Write the decision question first. Do not stop at “make a revenue chart.” Define the time range, unit, comparison, segment, exclusions, and the decision the viewer needs to make. A clearer request would be: “Compare weekly Q3 revenue by region and plan, excluding refunds, and flag regions down more than 15% week over week.”
Decide the audience and prohibited data. Know whether the result is personal, internal, or intended for customers. Anthropic's admin guide says HIPAA-ready organizations can enable these Artifacts after accepting the terms, but they are not covered by the Business Associate Agreement, so protected health information should not be placed in them.

How to build a Claude Dashboard

1. Connect the source

Connect the warehouse or app that holds the data. In an organization, an admin must also enable artifact connectors under Capabilities if an artifact needs to read from or write to connected apps. Each person connects their own apps, including people opening a shared artifact, so do not assume every viewer will see the exact same rows as the author.

2. Start from chat or the Artifacts tab

In any conversation, choose Output → Dashboards in the message box. You can also open the Artifacts tab, select a Dashboards template, and describe what you want to know. Naming the table or app helps when you know it.

Using analytics.orders in BigQuery,
compare weekly order count and net revenue for the last 12 weeks by region.
Exclude refunded orders and flag any region down more than 15% week over week.
Keep the aggregation rule and last refresh time visible for each chart.

Be explicit about whether the date is the order date, payment date, or fulfillment date. Also define whether “customers” means accounts, unique users, or completed orders. Most dashboard disagreements begin with a business definition rather than the chart design.

Official Claude Dashboards example built from bike-share data — An official Claude Dashboards product example, showing metrics, filters, and charts alongside the natural-language request. Source: Anthropic, checked October 15, 2026.

3. Inspect the query before trusting the number

Open a chart's query or select a number to see how Claude produced it. At minimum, check:

the start and end dates and the time zone;
whether canceled, refunded, test, or internal records are excluded;
whether the unit is customers, accounts, sessions, or transactions;
whether a join duplicates records; and
whether the last refresh time is recent enough for the meeting or report.

If a condition is wrong, correct the rule rather than only supplying the expected answer. For example: “Exclude rows where refund_status is completed, but keep pending refunds.” If you do not read SQL, ask Claude to explain which rows the query includes and excludes, then compare a small sample with the source system or data owner.

4. Move deeper analysis to the right tool

Dashboards is intended for fast exploration. A governed metric layer, formal financial report, complex row-level policy, or long-running operational dashboard still belongs in the organization's analytics environment. Anthropic currently lists direct handoff options including Amplitude, Grafana, Hex, Mixpanel, Omni, Perplexity, PostHog, and Sigma. Databricks Genie, Looker, monday.com, and Tableau are listed as coming soon, so do not treat those planned integrations as already available.

How to make a Claude Motion explainer

1. Specify the audience, venue, length, and source material

Choose Output → Motion from the message box or select the Motion template in the Artifacts tab. Instead of asking Claude to “make a video,” define who will watch, where it will play, the target duration, and which approved source material it should use.

Use the attached quarterly report and revenue chart to create
a 30-second explainer for the opening of our all-hands meeting.
Use four scenes and show only the three most important numbers.
End with two priorities for next quarter.
Prefer enough reading time over dramatic transitions.

Motion animates your content as code. It does not create realistic footage or generated people. For a product demonstration, a better workflow may combine an actual screen recording with Motion for the opening, chart explanation, or closing summary.

2. Review scene by scene and make narrow edits

Check the wording, numbers, order, color contrast, and reading time in every scene. Ask for scoped revisions such as “hold the second scene for two more seconds,” “change the Q3 figure to 18% and keep everything else,” or “remove the final call to action.” You can also edit directly, which means a small correction does not require regenerating the entire animation.

3. Separate MP4 export from Claude sharing

Select Export → MP4 to download the result. Use Share when reviewers should work with the artifact inside Claude. Team and Enterprise artifacts stay inside the organization by default. An owner separately controls public-link sharing and external email invitations. Artifacts that use connected apps or ask Claude questions may not be eligible for “Anyone with the link,” even when external sharing is generally enabled.

Three practical workflows

1. Find a change in weekly signup conversion

Goal: narrow down sudden changes in signup and paid conversion by acquisition channel. Inputs: read-only event, billing status, and campaign data. Process: compare the last 12 weeks and flag large week-over-week changes. Review: verify the unique-user rule, the definition of conversion, and the time zone. Use the result to choose where to investigate, not as proof of the cause.

2. Prepare a sales pipeline meeting board

Goal: organize Salesforce opportunities by stage, owner, expected value, and close month. Inputs: a connected Salesforce account and agreed stage definitions. Process: request pipeline totals plus opportunities that have not moved for a set number of days. Review: remember that viewers may have different Salesforce permissions, and do not publish a customer-level dashboard through an external link.

3. Turn a quarterly report into a 30-second briefing

Goal: communicate the most important change and next action at an all-hands meeting. Inputs: approved final numbers, brand assets, concise wording, and a target duration. Process: restrict Motion to four scenes and three key figures. Review: test whether viewers can read each number, confirm captions and contrast, and play the exported MP4 in the actual presentation environment.

Permissions, usage, and security checks

Artifacts start private. A new dashboard or animation is visible only to its author until it is shared.
External sharing has separate controls. “Anyone with the link” and invitations to named external email addresses are different organization settings.
Policy changes can affect existing access. Turning off external sharing can stop existing public links from working until the setting is enabled again.
There is no separate usage pool. Dashboards and Motion count toward normal Claude plan usage. Longer or more complex Motion projects can consume more usage than a typical message.
Use least-privilege connections. Prefer an account that can read only the required schemas and fields, and aggregate customer data before it reaches a shareable view.
Visible SQL is not a guarantee of correctness. Financial, legal, HR, or compliance numbers still need comparison with the system of record and the organization's approval process.

Troubleshooting

The Dashboards or Motion option is missing: Check the plan first. On Enterprise, ask an owner to verify Cloud code execution and file creation, Artifacts, and the specific template in that order. If access is limited through custom roles, verify the user's group and capability.
The dashboard is empty or incomplete: Reconnect the app if authorization expired. Confirm table and field permissions, date filters, and time zone. A person opening a shared dashboard may query through their own connection and see a different scope.
A chart does not match the expected number: Open the SQL and inspect the aggregation level, joins, and exclusion rules. Correct the definition, then compare sample records with the source system instead of forcing a target total.
External sharing is unavailable: Check the organization's external-link and external-email settings. If the artifact uses connectors, export an approved static result or MP4 instead of trying to make the live artifact public.
The animation is too long or expensive to revise: Reduce the audience goal, duration, scene count, and number of messages. Lock the facts and wording first, then refine transitions and visual polish.

Bottom line

Claude Dashboards can shorten the path from a business question to a reviewable data view because every chart exposes its query and refresh time. Claude Motion can turn the same approved material into a short explainer while keeping words, numbers, and timing editable. Neither feature removes the need for data definitions, access controls, or human review. Start with non-sensitive data and an internal 30-second asset, then keep formal reporting and external distribution inside your existing analytics and approval process.

Official sources

Sources checked October 15, 2026.

한국어로 읽기

Search This Blog

Stackfinch