Designing Autopilot-relevant automations

This page is targeted for automation developers, and guides them through the best practices on building automations which are optimized for Autopilot.

Autopilot can recommend and execute automations based on the user query. It also infers argument values to help setting up the automation. For these reasons, out-of-the-box automations have additional configurations and best practices to help facilitate Autopilot features.

Design best practices

To build automations that can be leveraged later by Autopilot for everyone , we suggest the following best practices:

Project best practices
To allow Autopilot to use the automation on MacOS, build the automation projects with the cross-platform compatibility. Focus each automation on a single task. For example, use one to five activities in a single sequence.

Input argument best practices
Use descriptive argument names and naming conventions, such as CamelCase format. If some arguments are mandatory for the successful execution of the automation, set them as required during the design time. Use the following data types exclusively: String - rendered as a free text field Int32 - rendered as a whole number picker Double - rendered as a whole or decimal number picker Boolean - rendered as a True / False dropdown menu Datetime - rendered as a date-time picker

Input argument best practices

Use descriptive argument names and naming conventions, such as CamelCase format.
If some arguments are mandatory for the successful execution of the automation, set them as required during the design time.
Use the following data types exclusively:
- String - rendered as a free text field
- Int32 - rendered as a whole number picker
- Double - rendered as a whole or decimal number picker
- Boolean - rendered as a True / False dropdown menu
- Datetime - rendered as a date-time picker

Output argument best practices
Use descriptive argument names and naming conventions, such as CamelCase format. Use the following output arguments exclusively: `ResultMessage` - can be either of the following: A string message, which is display back to the user. It usually holds the results of the operation. A string serialization of a JSON or Data Table object. Important: Autopilot can understand JSON arrays with single-level JSON objects or flat Data Tables. `output_html` - A HTML string used by Autopilot for rendering. If you use this output, Autopilot renders the HTML in an iFrame. The automation must not be configured as a pre-response automation.

Output argument best practices

Use descriptive argument names and naming conventions, such as CamelCase format.
Use the following output arguments exclusively:
- ResultMessage - can be either of the following:
  - A string message, which is display back to the user. It usually holds the results of the operation.
  - A string serialization of a JSON or Data Table object.
    Important: Autopilot can understand JSON arrays with single-level JSON objects or flat Data Tables.
- output_html - A HTML string used by Autopilot for rendering. If you use this output, Autopilot renders the HTML in an iFrame. The automation must not be configured as a pre-response automation.

Exception handling best practices
Use business rule exceptions and custom error messages to handle known errors, preventing the full stack trace from being sent to Autopilot.

Remarque :

To use existing automations in Autopilot, make sure they adhere to the previously mentioned design considerations.

Automation properties for Autopilot

Autopilot recommends automations solely based on the descriptions you provide. For Autopilot to suggest the correct automation, make sure each automation has distinct descriptions, to minimize overlaps. Autopilot can also infer argument values needed for the automation, simplifying the process and reducing the need for you to understand all the details. During setup, be sure to provide clear arguments. When you run the automation, Autopilot infers the argument values from the conversation or asks for additional details.

A set of metadata properties assist Autopilot in understanding various aspects such as:

What the automation does
How the automation card is displayed
How to infer the argument values
How the automation should be run
Any necessary context before executing the automation

Learn how to configure Automation properties.

Automations that return large amounts of data

Autopilot chat sessions have a limited capacity for information. Therefore, when building data-heavy automations, such as retrieval tasks, make sure to return only the relevant data.

To do this, consider the following:

Include arguments in the automation for filtering purposes. For instance, if you retrieve events from Outlook, add arguments that could allow filtering the results by the subject or attendees.
Create two automations:
- One automation to return multiple objects with limited details, and
- One automation to return extensive details for a single object.

Autopilot for everyone Studio template

The template UiPath provides has several design-specific recommendations already applied. To use the template, access it from UiPath Marketplace > Studio > Templates page.

After designing the automation, make sure to:

Publish the package to the tenant where Autopilot is installed.
Create and deploy the corresponding process to a folder where the required users have access.
Configure the automation properties.