Agents¶
Data use progresses from answering questions to completing tasks. An agent reads platform data, performs multi-step tasks, and writes results to a target system.
Your data |
AI participation |
|---|---|
Data in a knowledge base is read. Produced reports and tables are new data that can be registered in the Catalog and reused. |
An agent follows its task definition to retrieve, decide, and call external systems. Runs leave auditable, replayable records. |
How agents differ from Q&A¶
Q&A |
Agent |
|
|---|---|---|
Initiation |
You ask each question. |
You define a task once and run it on demand or on a schedule. |
Actions |
Retrieve and answer. |
Retrieve, process, decide, and call external systems. |
Output |
An answer. |
A report, table, or record written to an external system. |
Agents are built and run in astra. Their execution is auditable and replayable: each data read and action is recorded.
Outputs are new data¶
Reports and tables produced by an agent are new data. After they are registered in the Catalog, they can be processed again or included in a knowledge base for the next question or task.
The more you use the platform, the more data its AI can use, and the more complex the tasks agents can complete.
Next steps¶
What you want to do |
Go to |
|---|---|
Review how data reached this point |
|
Prepare better data for an agent |
For AI agents: an agent retrieves from knowledge bases. Its actions follow the task definition, and outputs are also registered in the Catalog when written back.