Autopilot for Maestro

AI-powered design assistance in Maestro, including BPMN process modeling from natural language input and document uploads, and the Autopilot coding agent for Maestro Flow, BPMN, and Case projects.

Autopilot for Maestro is an AI-powered capability that helps you design and evolve BPMN-based processes using natural language and document inputs. You can use Autopilot to accelerate initial modeling, refine existing workflows, and get contextual assistance while working on the canvas.

Maestro projects now offer two Autopilot experiences: the modeling assistant described on this page, and the new Autopilot coding agent, available in Preview.

Autopilot as a coding agent (Preview)

The coding agent extends Autopilot beyond the canvas. Instead of only suggesting model changes, it plans, builds, edits, validates, runs, and publishes a Maestro project from a plain-language description. It is available for Maestro Flow, Maestro BPMN, and Maestro Case projects.

The main difference from the modeling assistant is that the coding agent works through skills and tools. It loads the UiPath skill for the project type you have open, and runs file system, command execution, web, and utility tools that your administrator governs. You can review the enabled skills, tools, knowledge sources, and Model Context Protocol (MCP) servers in the agent settings.

Note:

Support for Maestro Flow, Maestro BPMN, and Maestro Case is experimental. Generated flows, diagrams, and case plans should be reviewed before running or publishing them. The switch button in the chat title switches back to the classic Autopilot experience in any Maestro project.

For details on what the coding agent does per project type, refer to the following pages:

• Maestro Flow (Preview)
• Maestro BPMN (Preview)
• Maestro Case (Preview)

How Autopilot helps you model processes

Autopilot for Maestro helps you move from intent to a working BPMN model faster, reducing the effort required to create, explore, and refine process designs.

Instead of manually building diagrams element by element, you can describe a process in natural language or provide supporting documents. Autopilot generates an initial BPMN model for you.

As the model is created, elements are streamed to the canvas incrementally. This allows you to review structure and intent as it takes shape and adjust early if needed.

Autopilot also supports you while refining a process. You can ask contextual questions about BPMN concepts, UiPath capabilities, or your current workflow to guide modeling decisions.

When working with logic, Autopilot can assist with writing C# and JavaScript expressions, including expressions used in Script tasks.

All interactions happen in a single Autopilot panel that preserves conversation history, making it easier to iterate without restating context.

After you understand how Autopilot fits into your modeling workflow, you can use it in the following scenarios:

• Creating, refining, and evolving BPMN models using natural language.
• Generating BPMN diagrams from supported file formats.
• Writing C# and JavaScript expressions for use in tasks such as Script tasks.
• Asking contextual questions about UiPath, RPA concepts, BPMN standards, and the current workflow.
• Receiving configuration suggestions while modeling.
• Preserving conversational context through chat history.

Creating and editing BPMN models

To get started, select the Autopilot icon under the left toolbar of the modeling canvas (also available on the right toolbar).

This opens the Autopilot chat window, where you can choose from three default templates for common horizontal processes: Purchase to Pay, Order to Cash, and Invoice Processing.

These templates help you understand how to begin prompting and get started with Autopilot.

Model diff view

When Autopilot suggests changes, the canvas will enter a diff view to compare the original model with the suggested model. The diff view includes the following components:

• Side-by-side view: The original model and the suggested model are displayed side by side on the canvas, with color coding to indicate changes (green for added elements, yellow for modified elements, and red for deleted elements).

• Orientation toggle: A toggle to switch between horizontal and vertical orientation of the models.

• Change list card: A card within the chat that displays a full list of changes in the model.

• Accept or reject changes: Options to accept (Cmd + Enter) or reject (Cmd + Backspace) the changes.

  Note:

  On a fresh canvas, the diff view is limited because there is no original model to compare against.

  

Incremental model generation

When Autopilot generates a BPMN model, elements are streamed to the canvas incrementally. This allows you to observe how the workflow is constructed node by node, instead of receiving a fully rendered diagram all at once.

Incremental generation helps you:

• Understand how Autopilot interprets your prompt
• Interrupt or refine prompts earlier
• Build confidence in the generated structure before accepting changes

File-to-BPMN conversion

Autopilot for Maestro can convert supported document and image files into BPMN diagrams. To do this, attach a file to the Autopilot prompt. Autopilot generates a BPMN model on the canvas, which you can then accept or reject.

The supported file types are: .pdf, .png, .jpg, .jpeg, .webp, .gif, .bmp, .tiff, .tif, .svg, .txt, .xaml, .doc, .docx, .xls, .xlsx, .ppt, .pptx, .rtf, .md, .markdown, .csv, .json, .yaml, .yml, .xml, .htm, .html, .css, .vsdx, .vsdm, .vdx, and .vsdx.

The maximum allowed size is 10 MB.

Writing expressions

Autopilot for Maestro helps you write expressions by leveraging existing variables. Access this feature from the expression editor. Autopilot-generated expressions can be used in standard expression fields as well as in Script tasks. Autopilot supports generating both C# and JavaScript expressions, depending on the task context.

Writing expressions

Expressions Diff view

Chat history

The Autopilot panel provides a single, unified workspace for prompts, suggestions, and responses. It retains conversational context across interactions, making it easier to iteratively refine a model without restating previous instructions.

Autopilot for Maestro supports chat history, limited to 10 responses or until the token limit is exceeded, whichever comes first. Autopilot uses any previous chats within the limit as context for future responses.

Autopilot for Maestro currently only supports English.

General Q&A

Autopilot for Maestro can perform general Q&A about topics such as UiPath, RPA, BPMN, and the current model. Some examples of questions you can ask are:

• What is the difference between RPA and agentic AI?
• What is agentic orchestration?
• What are the different types of gateways in BPMN, and when should I use them?
• What are all the possible paths of my model?

Important:

• Always review and test any generated model or expression before using it in production.
• When generating new diagrams, Autopilot adds lanes only if the diagram already contains lanes. Avoid requesting lanes on a blank canvas to prevent malformed BPMN models.