Skip to main content

Artifacts in conversational AI

TL;DR

How conversational AI creates artifacts such as reports, documents, images, and skills that you can preview in a side panel, revise with AI, and download.

Connect

Some questions are answered better by a file than by a chat reply. When you ask conversational AI for a report, a summary document, or a diagram, it produces an artifact: a standalone file that opens in a preview panel beside your conversation, where you can read it, revise it, and download it.

Prerequisites​

Before you begin, make sure you have:

Supported formats​

Conversational AI creates artifacts in several formats:

  • HTML: Formatted reports and summaries. HTML artifacts support styled layouts, tables, and charts, and render in full inside the preview panel.
  • Markdown: Structured documents such as runbooks, onboarding notes, or documentation drafts.
  • Images: Diagrams and other visual output, rendered in the panel.
  • Text: Plain-text output for results that need no formatting.

View artifacts​

When conversational AI creates an artifact, a card appears in the response with the artifact's title and format.

  1. Click View on the artifact card to open the preview panel beside your conversation.
  2. While the artifact generates, the panel shows progress. Content appears once the artifact is complete, so you never see a half-written file.
  3. Read the artifact in the panel, or keep chatting alongside it.

The panel closes when you send a new message, and reopens automatically if that response produces another artifact. If you close the panel yourself, it stays closed until you open another artifact.

Edit with AI​

Artifacts are meant to be iterated on. To change one:

  1. Click Edit with AI in the panel header.
  2. Describe what you want changed. For example, "add a column for asset owner" or "group these tables by connector."
  3. Conversational AI generates a new version, and the panel refreshes in place so you can see the result without reopening it.

Download artifacts​

Click Download in the panel header to save the artifact to your machine. The file downloads in its original format, using the artifact's name as the filename.

Use this to attach a generated report to a ticket, pass a document to a teammate, or keep a snapshot of output before you revise it.

Track versions​

Each edit produces a new version instead of overwriting the original. When a newer version arrives, conversational AI marks the earlier card in the conversation as Revised, so you can always tell which card holds the current version. Click View on the most recent card to open it.

Create skills from chat​

A skill is a reusable, named capability that tells an agent how to perform a task, including the business logic to follow, the steps to take, and the edge cases to call out. You can create one directly in conversational AI, without leaving the chat.

  1. Ask conversational AI to create a skill, and describe what it needs to do.
  2. Conversational AI asks follow-up questions to pin down the steps, business rules, and edge cases.
  3. It drafts the skill and opens it in the panel as a single document: the skill's name and description, followed by its instructions.
  4. Review the draft. Edit the instructions directly in the panel, or click Edit with AI to describe what you want changed and let conversational AI revise it.
  5. Click Save Skill to save it.

Once saved, the skill becomes a governed Atlan asset: versioned, discoverable, and available in your next conversation. You don't need Context Engineering Studio to create or use a skill this way.

For what skills are and how they combine with knowledge and tools, see Skills, knowledge, and tools.

Security and access​

  • Artifacts are scoped to your conversation. Only you can open artifacts created in your own threads.
  • HTML artifacts render in a sandbox. An artifact can't reach external services or load external resources, so nothing inside a generated file can send your data outside Atlan.

For how conversational AI encrypts and retains your data, see Atlan AI security.

See also​