This session explores the integration of artificial intelligence into test automation through a novel AI Agent built on top of Robot Framework.
The agent introduces intent-level automation, allowing testers to describe what to test instead of how to test it.

With new intent-based keywords such as Agent.Do and Agent.Check, the framework interprets high-level testing goals, transforming them into concrete test actions and assertions in real time.

At its core, the agent combines the reasoning capabilities of Large Language Models (LLMs) with visual understanding models.
It can interpret a tester’s intent, identify and interact with GUI elements, and verify expected outcomes visually without relying on locator-based definitions.
This enables a more resilient, self-adaptive testing approach suitable for rapidly evolving user interfaces.

• LLM Client Layer:
  A modular interface supporting multiple LLMs to interpret and execute test intents while staying fully compatible with Robot Framework logs.

• VLM (Visual Language Model):
  Merges a vision model with an OCR to extract visual context, semantics, and element coordinates from screenshots.

The roadmap explores advancing the Agent toward higher levels of autonomy and adaptive decision-making in test execution.

This session includes a live demonstration of the current prototype, focusing on:

• Reproducibility of the approach
• Scalability across diverse applications and test environments
• Quantitative and qualitative improvements in test robustness and maintenance

The session will show concrete results and a live demo of the current prototypes, emphasizing reproducibility, measurable gains, and practical AI-in-testing outcomes.