Skip to content

AI commands

Ask AI

Purpose: a simple OpenAI / ChatGPT connector. Use it to generate unique test data, specific dates, and many other unique scenarios.

Parameters:

  • Question: the prompt you want to ask the AI. For example, "What is the date of the first Friday of next month?"
  • Response format: optional; specifies how the AI responds. For example, "Respond only in date format dd/mm/yyyy with no additional information"
  • Variable: where the output is saved

Skills

By default the AI answers from the question alone. It cannot see the page. The Skills button changes that. It is the lightning-bolt icon at the right-hand end of the Question input, and it stays visible whether or not you have clicked into the field.

Turn on the screenshot skill and it shows as a highlighted pill on the input. When the step runs, the AI captures the screen at the moment the step runs, then answers using what is on it. Use it when the answer depends on something rendered in the test rather than something you can type into the prompt:

  • "Calculate the line total from the PDF preview"
  • "Read the invoice number from the confirmation panel"

The answer still goes to the variable you nominate, so later steps can use it.

Asking versus verifying

Ask AI with the screenshot skill returns a value you can reuse. If you only need a pass or fail on what is on screen, use Verify Using AI instead.

Verify Using AI

Takes a screenshot of the field or the full page and performs a natural-language evaluation.

For example, selecting the Screen option with the prompt "Validate that the GST is included" screenshots the page at the time of the step and checks whether GST is included in that screenshot.

This is useful for:

  • On-screen reports and invoices
  • Mass validation in one step
  • Icon-based fields, the Field option is handy for things like a tick icon

AI Actions

Lets the AI take over the test and perform steps from natural language.

For example: "Enter vendor name, account, group and then click save." The AI will attempt to perform this task for you.

Advanced users only

This command is for advanced users and requires knowledge of GenAI.

Parameters:

Parameter Purpose
Task Description A user prompt
Extend System Message System message
Max Steps Maximum steps the AI should perform, including thinking and interactions

Click (Using Computer Vision)

Purpose: an experimental command that finds a field on the page using an image as reference. For example, take a screenshot of the New button and this command uses that image to locate and click the field on the page.

Parameters: a single parameter: the image the automation should identify and click on the page.

Advanced: click position, click count, force, timeout.