Build a management dashboard for the numbers you review regularly, such as revenue, expenses, profit and orders.
The work has four parts: connect a source, verify what Claude has read, build the dashboard and choose how to keep it current. Check the figures against your source before using them to make decisions.
This guide uses a Google Sheets file, read through the Google Drive or Google Sheets connector. The final section also covers data warehouses and connected business apps.

What you will need
- A paid Claude plan: Pro, Max, Ultra, Team or Enterprise for Dashboards and connectors.
- Claude Desktop if you will work with files in a folder on your computer. Otherwise you can work in the browser.
- The setting “Create and edit files” turned on.
- A connector for your source: Google Drive or the separate Google Sheets connector made by Google.
- A data source containing your metrics, such as a Google Sheets file.
The Google Drive connector reads Google Docs, Google Sheets, Google Slides, PDFs and Office files.
Try a dashboard with sample data
Choose an example and paste the whole prompt into Claude, including the table. Each example uses fictional figures, so you can try it without connecting a spreadsheet or using company data.
Monthly management overview
How are revenue, expenses, profit, orders and new customers changing each month? Use ten months of fictional figures to compare financial results with orders and new customers.
Sign in to read and copy this prompt.
Sign inSales pipeline overview
Which deals are still open, what are they worth after allowing for the chance of winning, and who is responsible for them? Try a fictional list of twelve deals to compare stages, owners and the share of closed deals won.

Sign in to read and copy this prompt.
Sign inWeekly order operations
How many orders are delivered on time each week, and how many are returned? Use twelve fictional weeks to compare order volumes, delivery and returns.

Sign in to read and copy this prompt.
Sign in1. Prepare your data source
Start with one clear data source.
For example, a table:
| Month | Revenue | Expenses | Profit |
|---|---|---|---|
| January | 42,500 | 31,200 | 11,300 |
| February | 47,800 | 33,100 | 14,700 |
| March | 51,400 | 34,900 | 16,500 |
The same example table also carries an Orders column and a New customers column.
Choose metrics that fit your business — sales, projects, customer support, inventory, or anything else. The clearer your column names and the more consistent the structure, the more reliably Claude will understand the data.
2. Connect your source
In Claude, open Customize → Connectors. Choose Google Drive or the separate Google Sheets connector made by Google. For this example, Claude can read your sheet through either connector and build the dashboard from it.
Sign in and grant only the access needed for your source. You can disconnect the connector at any time.
Before relying on the dashboard for regular reporting, check how it refreshes using step 9.
3. Start the conversation
Claude's chat and Cowork are being merged into one Claude. The change is reaching accounts gradually. In the merged version, start a conversation and Claude chooses the tools. If your message box still shows separate Chat and Cowork options, choose Cowork for this guide, because the Dashboard template and the folder work in step 6 both run as Cowork tasks.
Another way to start is to open Artifacts in the sidebar and choose the Dashboard tile, marked Beta. Opening this template starts a Cowork task, even if Chat was selected.
You can also ask for a dashboard in any conversation, or choose Output → Dashboards in the message box. Everything you make is saved in the Artifacts tab in the Claude sidebar.
4. Ask Claude to verify the data first
Do not build the dashboard immediately. First, ask Claude to check what it can actually see in the data source. This is the most important step — it prevents wrong numbers.
Sign in to read and copy this prompt.
Sign in
Only after you have verified what Claude actually read should you create a visualisation.
5. If the source is large
Large files or tables may not be read in full through a connector. So if there is a lot of data, do not assume Claude has analysed the entire file simply because it found it.
Check:
- how many rows exist in the original
- how many rows Claude actually analysed
- whether totals and date ranges match
If Claude cannot access the full dataset, a more reliable path is:
Source → download as CSV/XLSX → Claude Cowork → Python analysis.
6. Process a large file with Python
Download the full data as CSV or XLSX and save it in a folder that Claude Cowork is allowed to access.
Then use a prompt like this:
Sign in to read and copy this prompt.
Sign inThis ensures the full file is analysed rather than only a partial extract.
7. Build the dashboard
Once the data is verified, explain what you want to understand, such as whether growing revenue is improving profit. Then give the metrics, period and layout you need.
Sign in to read and copy this prompt.
Sign in8. Refine by conversation
Refine the dashboard by chatting. Ask for a different chart, a clearer metric or a corrected calculation. After each change, ask Claude to check its numbers and calculations. Compare a few totals with the source yourself.
Use the Sources panel to trace a number. It lists each source's description, formula, result and how long ago it was updated. For a data warehouse source, you can also open the SQL query behind a number and ask for corrections.
9. How the dashboard stays up to date
You can check the dashboard when you open it and add a scheduled task for recurring work. First, establish what kind of source your dashboard uses.
A. When you open it
Before using the figures, check how fresh they are.
Check Updated … ago at the top of the dashboard and the age of each source in the Sources panel. If a figure is older than expected, ask Claude to update it from the source, or to load the new report if your data arrives as files, and check the result against the source. Whether a connected source refreshes on its own is not documented at the time of writing, so check the age before each use.
B. Add a scheduled task
Scheduled tasks are available on all paid plans. Choose hourly, daily, on weekdays or weekly, or set the task to run only when you start it. There is no monthly option. Use a task to check metrics or ask Claude to load a new report into the named dashboard.
- Tasks run in the cloud, even when your computer is asleep or Claude Desktop is closed.
- A task that needs a local file or local app runs only on your computer. Plan for that computer to be available.
- Each run is its own session and has no memory of the last run. Name the dashboard, source and any folder it needs in the task instructions.
- Tasks can use connectors. Claude notifies you if a run cannot use a connector that needs reconnecting.
- Under Scheduled in the sidebar, you can see, pause, edit and run tasks, and look through past runs.
Before using the example in step 10, add the dashboard's name and its source location.
10. Example scheduled prompt
Sign in to read and copy this prompt.
Sign inHow the full workflow works
- ConnectorRead data from the connected source.
- VerifyCheck what Claude actually read against the source.
- Python fallbackWhen needed, follow step 6 to analyse the full file with Python.
- DashboardBuild a management dashboard from the verified data.
- IterateRefine by conversation and check the calculations.
- ScheduledIf useful, schedule a check or ask Claude to load a new report.
Important note about company data
Artifacts start private. Use Share to share one. On Team and Enterprise, sharing stays inside the organisation by default. You can invite people or share with everyone in the organisation. Sharing with anyone who has the link requires an owner to enable external sharing. This option is never available for artifacts that use connected apps.
By default, an artifact using connected apps reads data through each viewer's own connections. A viewer without access to a source sees an error in that part. Check access with a teammate before relying on the shared dashboard.
After the first dashboard
Monthly report files
Keep monthly files in a folder Claude can access and use consistent names, such as 2026-10.csv. Run the prompt when a report arrives, or have a daily or weekly scheduled task check the folder. There is no monthly schedule option. A task using a folder on your computer runs locally. Check the result of the first run before you schedule it.
Sign in to read and copy this prompt.
Sign inReuse the dashboard for another dataset
Sign in to read and copy this prompt.
Sign inUse the brief in a new conversation, naming the new source and a new dashboard. Run the data check from step 4 first. Confirm the columns, periods, row count and totals before asking Claude to build from the brief.
Common questions
How do I use the dashboard on an ongoing basis?
Open your saved dashboard from Artifacts. Check Updated … ago and Sources before using the figures. Follow step 9 to check how fresh the figures are. For file reports, ask Claude to load the new report or use the monthly prompt above. Keep checking key totals against the source and use the conversation to correct calculations.
Can I connect a dashboard to a data warehouse?
Yes. Dashboard sources include Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake, plus connected business apps such as Salesforce. For a warehouse, Claude writes a SQL query for each chart. Open the query behind a number and check the calculation. Ask your data team for a read-only account and agree on metric definitions before building.
How do I share and manage dashboards with a team?
On Team and Enterprise, use Share to invite people or share with everyone in the organisation. Connected-app data uses each viewer's own connections by default. A viewer without source access sees an error in that part. Check access with a teammate, agree who maintains the sources and calculations, and review sharing when team membership changes.
Which Claude plans include Dashboards?
The Dashboards beta launched on 8 October 2026. It is available on Pro, Max, Ultra, Team and Enterprise. On Enterprise, an owner must enable it under Organization settings → Artifacts. Creating artifacts in chat works on every plan, including Free. Connectors and data storage need Pro, Max, Ultra, Team or Enterprise.
Is there a ready-made starting point?
Use the Dashboard template, marked Beta, on the Artifacts page. Then give Claude your source and the questions you want the dashboard to answer. Run the step 4 data check before building.
Can Claude use figures from a website I do not own?
Claude can read public pages. A scheduled task can read a page if you name its address. Before collecting data from a site you do not own, check its terms of use. Use an official export or API where available, because layout changes can quietly break the numbers. Ask Claude to read specific figures, report their period and update the named dashboard. Ask it to flag unavailable figures, and check the numbers against the page.
When a Claude dashboard stops being enough
A Claude dashboard can suit one team reviewing its own numbers. Consider a system built for your work when people need different permissions, several people edit the same records, figures must be written back to other systems, or you need a full history of changes.
For these needs, you can discuss custom software development or a business management system with WyrdIT. If your team wants to build dashboards itself, explore AI training.
Official sources
This guide follows Anthropic’s own documentation. It carries the newest and fullest version of everything below:


