Skip to main content

LLM Generated Test Steps

Overview​

Creating thorough, reliable automated tests can be a tedious task. With LLM Generated Test Steps, you simply describe what the user needs to accomplish on the page using plain business logic, for example, “fill out the form as a QA Manager working at Applitools and submit it.”

Autonomous instantly generates an executable sequence of test steps based on your intent. It analyzes the page, understands the context, and applies intelligent reasoning to choose the right data and UI elements to interact with in order to fulfill your intent. This enables accurate, context-aware test creation using advanced AI techniques while maintaining the precision and reliability expected from traditional test automation.

Authoring Flow​

Autonomous combines natural language understanding with page-specific context to generate steps. Here's what happens:

  1. Connect to the interactive browser to access your application under a test.
  2. You type a step describing your testing intent (e.g., fill out the form as a QA Manager working at Amazon).

  1. Autonomous inspects the page and applies proprietary AI logic, including large language models (LLMs), to interpret your intent.
  2. Autonomous breaks down your intent into verified, executable steps in plain English, and adds them to your test

While you're authoring, you can view your original intent text via the icon. After saving the test, you'll see the executable test steps instead of the original input.

How it Works​

Use business logic to describe what the user needs to accomplish on the page, like filling out a form. Autonomous interprets your intent in real time and instantly generates executable test steps, all within the context of the current screen.

Behind the scenes, it uses advanced AI including Large Language Models (LLMs) to understand your input and translate it into precise test actions, grounded in the live application under test.

The generated steps are then verified and finalized by our Deterministic Language Model (DLM), which ensures that:

  • Steps are accurate and grounded in the UI
  • Execution is fast and reliable
  • There are no hallucinations or invalid actions

With LLM Generated Test Steps, you get the best of all worlds:

  • No-code authoring with natural language input
  • AI-powered understanding of user intent
  • Deterministic execution for trusted results