Autopilot user guide

Last updated Oct 31, 2025

User scenarios for the Autopilot chat

The following examples describe how you can leverage Autopilot based on your scenario:

Generating workflows, sequences, or code

What you do: Use natural language in the chat to describe what you want to build.

How Autopilot helps:

Converts your input into a draft automation sequence.
Allows follow-ups such as "Can you change the order of these steps?" or "Use a different activity here."
Maintains full context of your request to incrementally build or adjust the automation.

For example:

"Create a process that reads unread Outlook emails, saves the attachments to a folder, and logs the filenames to Excel."

Follow-up: "Add a step to delete the email after saving the attachment."

Editing workflows, sequences, or code

What you do: Ask for changes to specific parts of an existing automation.

How Autopilot helps:

Recognizes references to named sequences or selected activities.
Applies changes such as "Use a different variable," or "Replace this with a conditional check."
Iterates on suggestions with you to refine the update in-place.

For example:

"Update the Data Extraction sequence to use the invoiceList variable instead of inputData."

Or: "Replace the last three steps with a Try-Catch block."

Explaining workflows, sequences, or code

What you do: Ask for explanations of workflow logic, activity purpose, or sequence behavior.

How Autopilot helps:

Summarizes selected parts of the workflow in human-readable language.
Responds to questions such as "What does this sequence do?" or "Why is this loop used?"
Supports deeper exploration through contextual follow-up questions.

For example:

"Explain what the Send Notifications sequence does."

Follow-up: "Why is it filtering on status = failed?"

Reviewing workflows, sequences, or code

What you do: Request a review or optimization of part of your automation.

How Autopilot helps:

Evaluates logic, error handling, and efficiency in the selected segment.
Suggests improvements such as "Simplify these steps" or "Improve exception handling."
Adapts feedback based on your goals, such as robustness, readability, or performance.

For example:

"Is there a more efficient way to process these invoices?"

Follow-up: "Can you suggest how to reduce the number of Excel write operations?"

Solving design-time errors

What you do: Ask about validation errors or broken configurations during workflow design.

How Autopilot helps:

Analyzes the error in context of the current workflow and variable scope.
Suggests fixes such as "Add this missing argument" or "Update this property value."
Lets you adjust or clarify the fix through chat.

For example:

"Why am I getting a variable not defined error in the email loop?"

Follow-up: "Add the missing variable with default value True."

Solving runtime errors

What you do: Respond to runtime exceptions or failed execution paths during debugging.

How Autopilot helps:

Highlights the step that caused the error and explains what went wrong.
Allows follow-ups such as "What input caused this?" or "How should I fix it?"
Proposes corrective actions or changes based on current context.

For example:

"Why did the Read Range step fail during execution?"

Answer: "The Excel file path is null. Would you like to add a null check before this step?"

Accessing documentation and knowledge base content

What you do: Ask questions about activities, errors, or best practices during development.

How Autopilot helps:

Returns context-aware answers pulled from official docs and KB articles.
Handles queries such as "How does this activity work?" or "What does this error mean?"
Provides links to full documentation while keeping you focused in your context.

Example:

"How does the Deserialize JSON activity handle nested objects?"

Or: "What are the recommended patterns for retrying failed transactions?"

On this page