Jun 16, 2024

AI as QA Agent

#programming, ai, testing

~ views | ~ words

A really useful way to apply AI could be to automate usage of your web app to catch bugs. This is typically something engineers write code to do or humans write test plans to do. But I wonder if, given a corpus of normal usage of your site (via web traffic analytics and client-side instrumentation), AI could simulate a real person using your app. It would arguably be better than hardcoded tests, because humans don’t behave predictably when using software either, so AI’s unpredictable nature would be a feature rather than a bug (unlike when we ask AI to do, say, math, where predictability is a bit more important).

An AI agent could presumably simulate usage across regions and languages and browsers and devices and create an incentive to build or integrate observability tools instead of testing tools. This kind of QA is expensive to do and it could be a perfect task for AI: low risk, high reward.

Maybe someone’s already working on this.

If you like this post, please share it on Twitter and/or subscribe to my RSS feed. Or don't, that's also ok.