The QA role is evolving, not disappearing
I think the guys at Qase said it best, “QA: My death has been greatly exaggerated”.
Testing doesn’t directly deliver an experience, but rather ensures that the desired experience is delivered. In order to do that, as a QA you need to understand how a day to day user thinks, you need empathy, you need to know what the software should feel like and you need to be aware of the risks it has.
AI is unable to truly grasp the feelings of a user whether it’s frustration, satisfaction, or joy. And it’s unlikely to gain that ability anytime soon; we’re still a long way off, in my opinion. What matters is how we use AI as it exists today. At the emotional level, only a human can genuinely understand another human. On a technical level, though, AI can still play a role and make an impact, even if it can’t connect with how something feels.
The QA role is becoming more dispersed, but it doesn’t mean it’s disappearing. The truth is simple: AI may change how QA works, but it can’t replace the human understanding which makes testing meaningful.
Why Human Judgment Is Still Critical
As mentioned earlier, AI can’t interpret human emotions. A feature might function perfectly from a technical standpoint, yet still leave users feeling frustrated or confused. AI can validate that the “Submit” button is working correctly and the request is sent with the correct payload, but our “Submit” button is half hidden behind another element on mobile on a certain resolution and it makes the user experience frustrating.
The AI can also have biases. What we have access to is trained by humans with data that already exists on internet. Let’s say a model is mainly trained with desktop bug reports instead of mobile issues. Because of this, it can “forget” about checking issues on other platforms, so the result yields false negatives (the things he forgot to check) and false positives (raising false issues for desktop). This leads to wasted time on both sides, QA and DEV.
Another very important issue with focusing on AI as an independent agent instead of using it as a tool is accountability. A critical bug slips into production making the company lose millions of dollars and thousands of users, who’s accountable here? The AI? The company that has cut the QA team from 10 people to 2? Those 2 QAs that can’t possibly check what a team of 10 QAs would’ve checked? In my opinion, it’s the company’s fault.
The Future: QA as AI Supervisors
We can all agree there are some repetitive tasks for all kinds of QAs. Writing the same testcases most of the time for different pieces of software, writing locators for automated tests and a few others. These tasks can safely be done by AI and then checked and used by QAs.
Using AI as another QA tool will create more space for exploratory and strategic testing. Focusing on what really matters. AI is now what automation was for manual when it first appeared in the 1990s. It’s cool, yes, but you still need the other types of testing, and you still need to double check sometimes, maybe now more than ever. Imagine you have a junior QA that you need to supervise and guide, you can push some tasks his way.
How AI Actually Helps QA Today
We talked about the why’s (why AI won’t replace QA, why QA should use AI as a sidekick, why AI needs to be double-checked) but let’s also talk about the how’s. How can AI help QA today.
AI can do self-healing. This means it can fix the broken scripts when UI changes occur. This will reduce flaky tests. Tools that offer this are ACCELQ or mabl.
AI can spot visual differences across devices. It can compare screenshots of older builds and the one that is most recent. Applitools Eyes and Percy from BrowserStack are offering this kind of solution. For example, Percy can be integrated in the pipeline, and react to PRs with visual changes, approving or rejecting them.
AI can analyze logs and analytics to detect anomalies, generate insights and summarize reports. You can use Chat GPT/Gemini and similar tools for this.
You can’t talk about QA with talking about test-cases. AI can generate test-cases as a starting point of you testing strategy. You can use jira plugins but don’t expect too much.
What I am trying to say is that AI doesn’t make the QA role meaningless. It makes it stronger. A QA working with AI is a better QA than one that’s refusing to see it’s value. But also, an AI trained for QA is not better than a QA working without AI. The partnership between smart tools and smarter testers will deliver a new level of quality.
