The debate in the software testing industry over “checking versus testing“ has always annoyed the right people. The ones who say it’s just semantics. The ones who say it is elitist. The ones who say normal people don’t care about a “distinction without a difference”.
I mean, I get it, nobody likes being told that the thing they’ve been calling testing is the equivalent of a green check, but that distinction is about to become an economic reality that most testers won’t survive.
And the sad part won’t be that AI is taking that work, it’ll be that so many testers were told by the “anti-checking” crew that what they were doing was testing.
Let’s face it: AI can already do checking.
Not perfectly. Not without some risk. But well enough to replace a great deal of the current population in our business. It doesn’t even need to be excellent.
It only needs to become cheaper, faster, and plausible enough to kick the teeth in of a large part of the software testing job market.
Once that problem gets close to being solved (a problem being worked on RIGHT NOW by your friends in the test tool industry), those jobs are going away and not coming back.
And a big part of why they aren’t coming back is because that money is moving.
Research and Markets values the AI model evaluation market at $2.36 billion in 2026 and forecasts it to reach $6.24 billion by 2030, with a 27.5 percent compound annual growth rate.
The market described in that report includes model validation, monitoring, benchmarking, compliance and governance, trending towards continuous validation, bias and fairness auditing, and explainable AI assessments.
The irony of all this is that this should have been the testers time to shine. Who should be better placed to test complex systems that behave unpredictably, depend on data, interact with humans, produce variable outputs, degrade over time, and fail in ways that are subtle, contextual, and expensive?
Testers!
Testers are investigators. Testers are system thinkers. Testers are people who understand that risk emerges from relationships between the system, the user, the environment, the business, the data, and time.
That’s why the “checking versus testing” argument was never about vocabulary, it was asking permission to stop thinking. If you define testing as checking, then AI is your competitor. If you define testing as skilled investigation, AI is a tool. And it only gets worse for the checking business from here.
A VentureBeat article on AI agents and chaos engineering describes a category of incident enterprises are not tracking well yet: autonomous agents taking technically reasonable actions from incomplete context that cause cascading problems in their infrastructure.
The article gives examples of agents restarting services, rerouting traffic, scaling resources, or modifying configurations in response to anomalies. In isolation, the action may make sense.
In the actual system state, with dependencies under pressure and other teams making changes, the action becomes a chaos event you can’t “check”.
Which brings me back to my point: most testers will not be laid off because AI is too smart. They will be laid off because they accepted a stupid definition of their own craft by people who don’t even work in our business!
The “checking versus testing” debate is dead. Bury it next to “best practices” and whatever has become of “shift left”. AI gets checking and testers get the future.
Layoffs in testing have only been nibbling round the edges, so take a class in RST and buckle up, the next leg down is going to be bumpy!
Discover more from Quality Remarks
Subscribe to get the latest posts sent to your email.