🔥 [Question of the Week] What's the ONE thing you wish AI could do in test automation that it can't yet?

Welcome back to Question of the Week — the AI edition!

AI tools have come a long way, but there are still plenty of gaps. What’s the feature or capability that would genuinely change how you test if AI could do it?

This week’s question: What’s the #1 thing you wish AI could do in test automation that it still can’t do reliably?

Be specific — the more detail you share, the better the discussion!


How to participate: Reply below. Upvote answers you agree with. Best response gets featured in our weekly digest!

Right now, AI is incredibly good at syntax generation (e.g., “Write a Playwright script to click this button”). What it cannot do reliably is look at a complex, end-to-end user workflow and determine if the system actually behaved correctly from a business logic standpoint.

What AI can do today:

  • Identify selectors and element IDs.

  • Generate boilerplate test scripts.

  • Catch basic visual regressions (e.g., “this button turned blue instead of red”).

What AI fails at reliably:

Knowing if the data behind the UI makes sense after a complex series of actions.

The Dream Feature: “Intent-Driven Oracle Testing”

I wish I could give an AI a high-level business requirement, and have it independently verify the system’s state across multiple layers (UI, API, and Database) without me having to hardcode every single assertion.