Is overusing AI causing a decline in professional tech skills?

Recently, I went to visit my father-in-law in an aged care facility. When I approached him in the lounge, he was waving a tablet in front of another resident, trying to teach them how to play Mahjong. It wasn’t hard for him to get them interested in playing the brain-teaser.

“Use it or lose it,” his new friend said, as my father-in-law tried to teach him how to swipe a touch screen.

Later that week, I loaded up the code for a game I published three years ago. I hadn’t touched the engine for a long time. Since then, users had sent me a long list of bug reports, and I decided it was time to rip that band-aid off. But when I tried to compile the changes, errors were thrown up all over the place! No stranger to debugging, I tried to isolate the problem.

It didn’t take long. The game engine had been updated in the time I’d been gone. As a result, the API calls and methods I’d used to make it had been made obsolete, along with some of my knowledge of that engine.

Because I hadn’t used it, I’d lost it. Time to read through the docs again…

Tech skills decline faster than you’d think (or like)

What I experienced was hardly uncommon—tech skills are in constant depreciation, and it can sneak up on you at any moment. IBM research suggests tech skills have a “half life” of just two and a half years, which lines up nicely with me losing engine-specific knowledge in three years. 

The key term is “half life,” here. It’s unlikely a front-end developer would forget fundamentals in just three years, but frameworks and libraries certainly evolve, and you might go to use something that is depreciated or obsolete. Think back to the ancient year of 2022, when AI agents were not even a thing. Now, they’re used by over 30% of professional developers. 

Things move fast around these parts.

Of course, usually these tech skills are maintained by two things: using them on a daily basis (unlike me, in the above scenario) or engaging in further learning specific to those skills (which I also didn’t do). For tech professionals in hands-on roles (e.g. software developers), you’d historically tick the first box at least. You’re using it, so you’re not losing it.

But does AI change this calculus? Let’s look at the case of professional developers, even though I know tech is so much broader than that.

Most developers are using AI on a daily basis now

84% of devs use AI in their development process, and half use AI tools on a daily basis, according to a study by StackOverflow. That’s not unsurprising: the day I installed Supermaven (a code completion tool) into VSC, I was blown away by how eerily well it could read my mind. After typing just three letters, it would conjure the entire line of code I had in my mind, incorporating all my functions, variables, and weird logic. 

There was no way something like that wasn’t going to be widely adopted. Of course, then the temptation becomes not to engage your mind at all. 

A while back, I was building a demo for a 3D space sim. However, to pull it off, I needed to understand linear algebra well. Since I hate math, this was an incredibly unattractive proposition, and I didn’t know where to start. In a moment of sheer laziness, I asked ChatGPT if it could make something. 

It did. Sort of. The code ran, but it had lots of functionality issues. Because I didn’t understand linear algebra, I couldn’t really actually understand the code, so I asked ChatGPT to fix the issues without any true comprehension. (I now know this is called vibe coding!) In the end, it couldn’t offer a solution that was completely right, so I sighed, opened up 3Blue1Brown, and started watching math videos. 

AI sucks at complex tasks, but for some people, it’s good enough

What I’d discovered was the single biggest reported frustration cited by 66% of developers: “AI solutions that are almost right, but not quite,” which tends to lead to the second-biggest frustration: “debugging time-consuming AI-generated code.” It’s one of the reasons that developer sentiment of AI is pretty low: more developers distrust the accuracy of AI tools (46%) than trust it (33%). How can you trust something that produces junk code?

However, if I were a lazier person, I could have pushed the buggy code that ChatGPT had made anyway. That sounds crazy, but there is no shortage of lazy people in this world. 

A small percentage of developers are engaging in vibe coding as part of their professional development work (15%), getting LLMs to produce code and trusting it without human review. That’s not the majority by any means, but 15% is a lot when you consider two things:

• That means 15% of the code going into professional environments and products is now unvetted by a human, and;

• People are unlikely to admit to vibe coding (Hint: It’s still not popular to admit to using AI at work), so this percentage is likely higher.

That said, there are people who do this not because they’re lazy, but because their workplace rewards or encourages it, such as putting them under considerable pressure to deliver with limited resources, or not rewarding good work over fast work (“Move fast and break things”).

Of course, all this AI use isn’t specific to developers. According to Google’s DORA research division, 90% of tech industry workers are using AI in their job, up 14% from last year.

How this all loops back to tech skills decay

If you don’t use skills, you lose them, and tech skills depreciate faster than any other kind. This makes sense when you think about how most people treat AI tools: like an intern or junior tech professional, but sometimes requires guidance and gets things wrong. Naturally, if you asked said intern to do a key task for you, you’re not doing them, and so your skills would start to atrophy. (It wouldn’t be surprising to find that intern went on to take your job, but that’s an entirely different issue.)

Of course, you could argue that for many tech professionals, AI is far too unreliable to be trusted enough to do large chunks of your work. I would argue that people are already using it for this to a degree, and that these tools are either going to plateau for the rest of time or become slowly more functional. 

In the next few years, the temptation to push tasks to AI will grow, and as a result fighting skills decline will become a concern not just for individual tech professionals, but for organizations as a whole. 

Further reading

To learn more about trends like this one that may affect your organization in the year ahead, read Pluralsight’s 2026 Tech Forecast, a report based on predictions from 1,500+ tech insiders, business leaders, and Pluralsight Authors.