The mission of testing is not confidence.

That was true years ago when testing charlatans tried to convince the world that confidence could actually be measured from test case counts, requirements coverage, defect closure rates, and all their blardee, blardee, blar.

It’s still true, and even more important to push back against this nonsense now that confidence is being repackaged in AI language like agentic test workflows, autonomous testing, self-healing test automation, and machine-extruded test evidence.

At a recent software testing conference I participated in, the cognitive surrender was like a colorless, odorless gas that seeped into the room and before you know it, everyone was asleep. One of the arguments I heard was that making sure AI “works” is so hard we just need to give in to the AI, it all works better when we don’t try to test it, we can’t explain or understand it so just let AI do its thing.

Don’t think that’s happening?

Read the latest blog post (or press release in light of their IPO) by Anthropic to see just how real that world is.

What was most surreal was not the “what are we gonna do when AI builds itself” speculation. It was the internal quote about days when everything works and the person feels irrelevant, followed by days when everything breaks and they realize they no longer understand what they’ve been doing.

“On days where everything works well, I can’t help but think nothing I do matters, everything is automated and better and faster than I ever will be. But then there are days where everything breaks and I don’t understand why and I realize I have no idea what I’ve been up to anymore.”

That’s the entire sociotechnical risk these massive Ai companies represent in a nutshell.

Because if AI can generate code, tests, fixes, experiments, reports, and release recommendations faster than humans can meaningfully review them, then “testing for confidence” becomes actively dangerous. Confidence becomes easy to manufacture – and weaponize.

The mission of testing is not confidence.

The mission of testing is to provide information that helps people make decisions about risk. Confidence may be an outcome of that information. It may also be reduced by that information. Both are legitimate but neither are the goal.

As I’ve said before, the problem with “testing for confidence” is that it changes the job of the tester. Instead of asking, “What do we need to learn to manage risk to our business?” teams start asking, “What can we produce to feel like we’re ready to release?”

That goal displacement matters a lot.

It moves testing toward confirmation to support an outcome rather than investigation by rewarding the production of evidence that supports the desired decision. It makes uncertainty look like failure instead of a normal and necessary part of responsible decision-making, and that’s the core elements of that sleeping gas.

And the AI tech bros in testing would have you grab that pipe with both hands and take a long, deep breath.

This is the confidence game being played in testing right now: vendors are increasingly replacing older language with any combination including “confidence”: release confidence, testing confidence, ship with confidence, and that testing’s future is confidence as if that can somehow be engineered.

The enshittification of the testing industry has arrived draped in lazy marketing.

Harry Collins summarized what’s happening to testing beautifully in his book Artifictional Intelligence by warning that humans surrender too easily to machines when we mistake performance for understanding.

AI systems aren’t socially grounded in understanding, judgement, and tacit expertise that humans develop through participation in a community, so society has to adapt to fill the void that AI can’t produce.

And thats the void those telling you confidence is the mission of testing are attempting to fill.

Look, I realize AI is not going away and there are some benefits to using AI in testing and its usage creates new challenges for testers that must be addressed. But in the age of AI in testing, we have to become more serious about uncertainty, not less.

The mission of testing is not confidence.

If someone is telling you it is, they are either knowingly trying to deceive you, dangerously naïve and shouldn’t be involved in testing, or most likely trying to sell you something.