When it comes to chatbots, the chat window traditionally acts as the place where numbers actually lived were separate rooms. Someone asked a question in one, filed a ticket or wrote a query in another, and waited for a chart to come back as a screenshot.
Explaining that chart to a room usually meant a static slide, or a recording someone had to remake every time a figure moved.
Now, Anthropic is putting two related betas in front of that gap.
Claude Dashboards, on paid plans, connects to a data platform or a CRM and answers a plain-language question by writing the query and assembling a dashboard that stays current as the underlying data changes.
Supported sources at launch include Amazon Redshift, BigQuery, ClickHouse, Databricks, Snowflake, and Salesforce.
Each chart shows the query behind it, along with a refresh time, so a number can be inspected rather than taken on trust.
If the question needs a fuller analytics environment, the dashboard can be sent on to tools such as Amplitude, Grafana, Hex, Mixpanel, Omni, Perplexity, PostHog, or Sigma, with Looker, monday.com, and Tableau listed as coming later.
Claude Motion, in beta on Team and Enterprise plans, takes the other half of the same problem.
A quarterly report, a chart, or a product walkthrough can be turned into a short animation and exported as an MP4. The animation is written as code rather than produced by a video model, which means there is no generated footage and no synthesized person in the frame.
Text, numbers, shapes, and timing can be edited in place or by asking again, and the file can be opened in editors including Adobe, Descript, HeyGen, Higgsfield, invideo, Luma AI, and Runway, with Canva and Captions expected to follow.
The point of the code path is revision: a figure that changes does not require a new generation from scratch.
The same announcement moved Claude Docs, Slides, and Design out of beta and onto every plan, including Free.
Anthropic says users have made more than 45 million artifacts since that format appeared. Those three can now be edited by a team and by Claude in the same file, shared inside or outside an organization, and exported to PowerPoint, PDF, or Google Slides with formatting kept. Design, previously a separate site, is being folded into ordinary conversations.
The standalone site is set to close on December 14, and chats and comments from it will not carry over. For enterprises, the newer artifact features start off and have to be enabled in organization settings. Docs, Slides, and Design are scheduled to turn on by October 15 unless an admin does it sooner.
Taken together, the release is less a new chat behavior than a set of work surfaces that sit next to the conversation and hand off to tools companies already run. The dashboard is meant to remain attached to a warehouse. The animation is meant to remain an editable file.
A day earlier, OpenAI began rolling out something aimed at a similar discomfort with plain text, under a different shape. Intelligent UI, shipping with GPT-6, lets ChatGPT compose a reply from prose, images, charts, buttons, forms, and small tools such as a savings calculator, a bill splitter, or a tappable diagram.
The interface is built from a library of native components and a compiler that renders them as the model generates, so the layout can appear while the answer is still forming. Users can turn the visual density down. It reached Plus, Pro, Business, and Enterprise on October 7, and Free and Go tiers the next day.
Whereas the former puts an interface inside answer, the latter is making the model procide objects that outlive the chat itself: a dashboard bound to company data, with the query left visible, and an animation stored as code and exportable to an editor.
One treats the chat as the place where the tool appears. The other treats the chat as the place where the tool is requested, then sends the result toward the systems where the data and the finished cut already live.
