Previously I had talked about the idea of personal marketability when it came to learning AI. That was in the context of an AI and Testing series, so I trust what was being learned was obvious. I’m seeing an interesting trend developing with ardent opponents of any and all AI, particularly on social media platforms like LinkedIn. This is my attempt at a social response to that.

There’s a reality around AI that I see on LinkedIn, Reddit, and other venues, quite a bit: certain vocal testers or developers opposed to the current crop of generative AI seem to base their critiques solely on using ChatGPT, Claude, Grok, Gemini (or whatever) to generate code or tests. Or maybe seeing if one of the model’s can finally count the number of the letter “r” in a word like “strawberry.” There are valid critiques there, to be sure, but it’s also very surface-level compared to what a wide swath of the industry is actively pursuing.

AI didn’t arrive with ChatGPT. It’s been embedded in the software we test and use for decades: spam filters, recommendation engines, fraud detection systems, loan permission systems, search ranking algorithms, predictive text, image recognition pipelines, and materials science. The difference is that most of that AI was invisible to the average user and largely invisible to the average tester. You didn’t have a conversation with your spam filter. It just worked, or it didn’t, and diagnosing the problem required knowing how it was built.

What’s happened in the last few years is democratization, not invention. Generative AI gave everyone a direct interface to a capable model, which is genuinely new, but it also means that many people formed their entire mental model of “AI” from that interface. Critiquing all of AI based on a few ChatGPT sessions is a bit like critiquing all of databases based on a few frustrating experiences with a spreadsheet. The surface is related to the thing underneath, but the conclusions don’t transfer cleanly.

Now, that being said, a recalibration is starting to show up in the market itself around generative AI patterns. Microsoft debuted its Gaming Copilot for Xbox in March 2025, pitching it as an AI sidekick that could offer real-time tips, game recaps, and coaching. It lasted roughly fourteen months before the new Xbox CEO cancelled it on mobile and halted its planned console launch entirely. The feature was widely described as a solution in search of a problem: players hadn’t asked for it, and most didn’t notice when it was gone.

This isn’t an isolated case. McKinsey’s 2025 State of AI survey found that two-thirds of companies remain stuck in pilot phases, and only about 39% report AI having any measurable effect on earnings. The honest reading isn’t that AI has failed; it’s that generative AI features bolted onto existing products, without clear user need, tend not to survive contact with actual users. Which means the question worth asking isn’t whether generative AI is overhyped — after all, it clearly is to some degree — but whether you understand it well enough to know the difference between hype and substance. The gap between hype and substance is precisely where skilled testers and developers have something real to contribute.

The testers and developers who will be most valuable in this space aren’t the ones with the loudest opinions about whether AI is good or bad. They’re the ones who understand enough about how these systems are actually built (the pipelines, the retrieval layers, the evaluation frameworks, the failure modes) to say something precise about where they hold up and where they don’t. That’s a higher bar than “I tried it and it was non-deterministic.” It’s also a more interesting one.

If you really do care about where this industry is going with the widespread dissemination of AI, and I for one am very concerned across a broad array of fronts, then do this: Practice with it in depth. Experiment. And then show your experiments. Make those experiments repeatable for others. Not only will this make you more career relevant and career competitive, regardless of the direction AI goes, but it will make you an active agent of attempting to steer the technocracy towards better rather than worse outcomes, rather than somebody complaining (even if validly) about certain forms of AI.

You will not combat uninterpretable, untrustable, unethical, or worrisome AI by sounding off on social media. You will stand a chance of combatting it if you work from within, understand exactly how and where it can fail (while also understanding exactly how and where it can succeed), and showing people paths to interpretable, trustable, ethical, and less worrisome AI.