What's an AI Agent in ARPIA?
What's an AI Agent in ARPIA?
An AI Agent answers questions about your data in natural language. You build it from workflow blocks, so users get answers without writing SQL or accessing databases directly.
Use an AI Agent when people need to ask questions. If you need answers grounded in documents, use a RAG AI. If you need to run a repeatable process, use an AI Worker.
What it does
- Interprets a question and works out its intent
- Retrieves data the user is authorized to access
- Returns a summary, structured output, or visualization
- Highlights trends and anomalies in the results
Building blocks
An AI Agent is a workflow made of configurable blocks:
| Block | Role |
|---|---|
| LLM block | Interprets the question, structures intent, and writes the response |
| Generative Query block | Retrieves data from authorized sources |
| Logic & Processing block | Applies filters, transformations, or business rules before the answer is returned |
You combine and configure these blocks to fit a specific use case or domain.
How it works
- Interpret. An LLM block parses the question and its context.
- Retrieve. Generative Query blocks pull data from authorized sources.
- Process. Logic & Processing blocks apply any filters or business rules.
- Respond. The answer is returned as text, structured data, or a visualization.
Data sources
AI Agents read from authorized ARPIA components, including:
- Data Models (Nodes)
- Neural Networks
- Structured datasets in the platform
Access and governance
- Role-based access. The agent only sees data the user's Security Profile allows.
- Read-only by default. Write access requires explicit configuration.
- Scoped by workflow. The agent can only do what its blocks define.
For governance that applies to all AI implementations (human review, logging, change management), see AI-Driven Implementations.
Limits
- Answers are only as good as the underlying data and the queries the workflow runs.
- Review AI-generated summaries before using them for high-stakes decisions.
Related
Updated 8 days ago
What’s Next
Did this page help you?