You did it!¶
Congratulations!
You've built a conversational agent grounded in real documentation, connected it to live automation tools, and validated it through four types of real conversations.
What you built¶
You created a conversational agent in Agent Builder that answers questions about invoices and ServiceNow tickets in real time. You grounded it in UiPath Security documentation using a Context Grounding index, then connected it to both RPA and API tools published in Orchestrator. Finally, you tested it across a data query, an incident lookup, a write operation, and a knowledge question — watching each tool call happen live.
| Component | Role |
|---|---|
| Conversational Agent | Answers questions in real time, scoped to grounded knowledge and connected tools |
| Context Grounding | Anchors responses to UiPath Security documentation, with citable sources |
| Orchestrator Tools | Lets the agent retrieve invoice data and ServiceNow incident details on demand |
| Prompts | Guides users toward the questions the agent is built to answer |
What's next?¶
When to use which agent type?
| Autonomous | Conversational | |
|---|---|---|
| Interaction model | Single-turn, task execution based on an initial prompt | Multi-turn, back-and-forth dialogue |
| Primary use cases | Executing complex tasks from a defined, structured prompt | Real-time user support and assistance, interactive information gathering |
| User input | Single, structured prompt or command | Continuous chat messages, often with ambiguity |
| Core strength | Executing predefined processes and automated workflows | Maintaining conversation, understanding context, and handling nuances |
1. Deploy and share your agent¶
Publish your agent from Agent Builder and share the endpoint with a colleague. Try asking it questions outside the grounded scope — notice how it handles topics it shouldn't answer.
2. Review observability data¶
Open the observability dashboard in Agent Builder. Check conversation logs, tool call frequency, and user feedback to decide what to improve next.
Keep iterating¶
Broaden the knowledge base
- Add more documentation indexes to the context grounding source and observe how the agent's coverage changes.
Add more Orchestrator tools
- Connect additional tools and connections (for example MCP servers).
Refine the system prompt
- Regularly revisit the system prompt based on data and feedback. Tighten the domain constraints, adjust tool usage rules, and see how the agent's behavior shifts.
Learn more¶
| Resource | Description |
|---|---|
| Conversational Agents | Overview of conversational agents in UiPath Agent Builder |
| Getting Started | Step-by-step guide to creating your first conversational agent |
| Agent Design | Prompt configuration, tools, and grounding options |
| Deployment | How to publish and share a conversational agent |
| Observability | Monitoring conversations and tool usage |
| Managing Indexes | Setting up documentation indexes for context grounding |