Agents user guide

Last updated Feb 19, 2026

Conversational agents

Conversational agents are a class of UiPath agents designed for dynamic, multi-turn, real-time dialogues with users. Unlike autonomous agents that execute tasks from a single prompt, conversational agents interpret and respond to a continuous stream of messages while managing conversation context, tool execution, and human escalations.

Use conversational agents when your automation scenario requires:

Ongoing clarification or back-and-forth exchange
Personalized guidance based on user intent
Seamless human fallback when confidence is low

Conversational agents vs. autonomous agents

Feature	Conversational agent	Autonomous agent
Interaction model	Multi-turn, back-and-forth dialogue	Single-turn, task execution based on an initial prompt
Primary use case	Real-time user support, interactive information gathering	Executing a task from a defined prompt
User input	Continuous chat messages	Single structured prompt
Core strength	Maintaining conversation context and handling ambiguity	Executing a plan across tools

When to use conversational agents

Use conversational agents when your automation involves real-time, context-aware interaction:

Self-service experiences: Helpdesk support, HR onboarding assistants, IT troubleshooting bots
Interactive guidance: Multi-step processes, forms, or decision trees
Contextual conversations: Scenarios where users ask follow-up questions or provide information incrementally
Natural language interfaces: Querying applications, systems, or knowledge bases conversationally

Use autonomous agents instead when the task can be fully described in a single prompt with all required inputs provided upfront:

Structured document processing (extracting data from invoices or contracts)
Automated report generation based on predefined logic
Summarization or transformation tasks with clear, one-shot requirements

Conversational agents vs. Autopilot for Everyone

How do conversational agents relate to Autopilot for Everyone?

These two experiences work side-by-side:

Autopilot for Everyone: UiPath's general-purpose assistant, optimized for productivity tasks and interacting with the UiPath platform.
Conversational agents: Specialized agents you build for specific use cases (such as an HR policy assistant or IT helpdesk bot).

You can access your conversational agents directly from within Autopilot for Everyone, making it a central hub for all your conversational needs.

Our recommendation: We recommend building new conversational use cases on the Conversational Agents platform. It provides a comprehensive design-time experience with built-in evaluation, advanced observability, and full API access.

Note:

Conversational agents do not currently support local desktop automation. If your use case requires triggering automations on the user's local machine, this capability is planned for a future release.

The agent lifecycle

Building and operating a conversational agent follows four phases:

Lifecycle diagram showing Design → Evaluation → Deployment → Observability → back to Design

Design

Use Studio Web to define your agent's persona, configure tools, add context grounding for knowledge retrieval, and set up escalation workflows. The low-code designer lets you build agents visually without writing code.

Learn more about designing conversational agents.

Evaluation

Test your agent using the built-in Debug chat to validate multi-turn interactions. Create evaluation sets from real conversations to measure performance across different scenarios, including single-turn responses and multi-turn dialogue flows.

Learn more about evaluating conversational agents.

Deployment

Publish and deploy your agent to Orchestrator and make it available through various channels: Instance Management, Autopilot for Everyone, Microsoft Teams, Slack, or embedded in an iFrame inside third-party apps or UiPath Apps.

Learn more about deploying conversational agents.

Observability

Monitor your agent's performance through Instance Management dashboards, debug using traces, and audit using AI Trust Layer audit. Review trace logs, collect user feedback, and use these insights to iterate on your agent's design—completing the feedback loop.

Learn more about observability for conversational agents.