Is QA still relevant with agile and AI coding agents?
Agile made QA continuous; AI coding agents make it urgent. Why human QA judgment matters more in 2026, and where it fits in a fast release cycle.
In this article
Originally published 2012. Rewritten October 2026.
In 2012 the question was whether agile teams still needed QA once developers wrote their own tests. In 2026 it's sharper: if an AI agent writes the code and the tests, who checks either of them? The answer is the same as it was, but the stakes are higher. Quality assurance is still relevant. What changed is where it sits in the cycle.
What agile changed
Agile moved testing earlier and spread it across the team. Developers write unit tests, every commit runs through continuous integration, and "done" includes "tested". That made a separate test phase at the end of a long project mostly disappear, which was a good thing.
What agile didn't remove was the need for someone to look at the whole product the way a user does. Unit tests prove that each piece does what its author intended. They don't prove that the intentions add up to an app that works on a customer's phone.
What AI coding agents change
Coding agents now write features, refactor code and generate test cases in minutes. That speeds everything up, and it moves the risk:
- Agents test what they were told to test. If the prompt missed a case, so did the tests.
- Their self-tests run in clean environments. A headless browser or an emulator won't show how the app behaves on a three-year-old budget Android with a full disk and a weak signal.
- They sound equally confident when they're wrong. A passing test suite written by the same agent that wrote the code is weak evidence.
So the more code an agent writes, the more valuable an independent check becomes: someone, or something, that didn't write the code, judging whether it actually works.
Where QA fits in a 2026 sprint
| Stage | What happens | Who |
|---|---|---|
| Planning | Agree acceptance criteria and the riskiest flows | Product, engineering, QA |
| Building | Automated tests on every commit, many written with AI help | Engineers and agents |
| Before release | Human sessions on real devices: the key flows plus exploratory charters | QA or a crowd testing provider |
| After release | Watch crash rates, reviews and support tickets; feed issues back | QA and support |
The step that most teams skip is the third one, because it can't be fully automated. It's also where the bugs users notice tend to show up.
What QA people do now
The job has moved away from clicking through scripted cases and towards judgment:
- Reviewing AI-generated test cases for what they missed
- Deciding which flows carry the most risk this release
- Running exploratory sessions and writing clear, reproducible reports
- Saying "not ready" with evidence when it isn't
Those are human skills, and they're getting scarcer as teams cut manual testing.
Signs your team has cut QA too far
- Your app store reviews mention crashes, failed payments or OTPs that never arrive.
- You ship a hotfix within days of most releases.
- Customers and support staff find bugs before your team does.
- "It works on my phone" is a common answer in bug discussions.
Where 99tests fits
If your team writes tests with AI and runs them in CI, 99tests adds the step that's hardest to automate: human testers running those test cases on real phones and browsers in India, with every finding verified before it reaches you. See how we test AI-generated test cases and the packages.
