Agentic Test Execution: a new use case for Agentic QA?

We are starting to hear a lot about Agentic QA. But that term can cover very different things: agents can help analyse requirements, design tests, generate automation, investigate failures, or support exploratory testing.

There is one use case I find particularly interesting: Agentic Test Execution.

The idea is quite simple.

Instead of asking a tester to execute a test manually, or first translating that test into automation code, an AI agent reads an existing test case written in natural language or Gherkin and executes it directly on the application.

It interprets the instructions, interacts with the UI, checks the expected results and documents what happened so that the tester can review the evidence.

This creates a new opportunity that can help accelerate test cycles.

I see Agentic Test Execution, scripted automation and manual execution as complementary approaches.

Scripted automation remains very effective for stable, deterministic and frequently executed checks. Human testers remain essential for exploration, risk analysis, UX and situations requiring judgement.

Agentic Test Execution is particularly interesting in the space between the two: functional tests that teams still execute manually because scripting and maintaining automation may not be worth the effort.

At Smartesting, we have been working specifically on this problem with Lynqa. One thing we've learned is that the human role doesn't disappear. The agent executes and documents. The tester remains responsible for deciding what to test, reviewing the evidence and deciding what to trust.

For me, this raises questions that are more interesting than whether an AI can click buttons.

Does “Agentic Test Execution” make sense to you as a distinct use case of Agentic QA? Where would you use it rather than manual execution or scripted automation?

author : Bruno Legeard