The Great AI Deskilling has begun

• AI can boost performance at work but then quietly erode workers' core skills over time.
• AI researchers said AI creates an illusion of expertise and inflated confidence.
• Overreliance on AI may weaken judgment, learning, and long-term resilience, they said.

Does using AI have you feel like you're getting faster at, but worse at, your job? You're not alone.

For Josh Anderson, a software consultant with 25 years of coding experience, the first couple of weeks felt like magic.

From June to August, Anderson livestreamed himself on YouTube as he built a software product using AI without writing a single line of code himself.

The app, called Road Trip Ninja, was designed to help families find reliable stops on long drives — places with clean bathrooms, decent food, and space for kids to run around. Users could log and rate locations along major routes, turning trial-and-error road trips into something more predictable.

As he let AI build his app, features appeared in minutes. Progress came fast.

But as the livestream progressed over a few weeks and the codebase expanded to around 100,000 lines, the pace slowed.

The back-and-forth with the chatbot stretched from minutes to hours. Plans drifted from the standards he'd set. And fixing problems turned into a "never-ending wrestling match," he said.

By the time Anderson stepped in to make changes himself, something else had shifted: his confidence.

Anderson, assumed that if he needed to take the wheel during his experiment, he'd slide back into the work without thinking about it.

He's spent years building and maintaining systems meant to survive real users and real chaos. Instead, when he opened the code, he hesitated.

"It wasn't like I was completely frozen," he told Business Insider. "But there was just hesitation with every move."

That moment captures a risk that a growing group of workplace researchers say companies and workers aren't paying enough attention to: that AI is quietly deskilling people.

Early warning signs

He's not alone. When Anthropic's Claude went down earlier this month, some developers said they struggled to continue their work. Tasks that had become routine with AI suddenly felt harder without it.

"Claude outages hit way harder when you realize you've outsourced half your brain to it," one Redditor posted. Another joked: "I guess I'll write code like a caveman."

Moments like this point to a shift already underway: AI is boosting output while quietly chipping away at the skills behind it.

John Nosta, founder of innovation and tech think tank Nosta Lab, calls this the "AI rebound effect" — when better performance masks declining ability. "The skill set actually falls below baseline," he said. The danger isn't only dependency — it's regression.

Because AI systems deliver fast, polished answers, they can also distort how people judge their own abilities. "We have an overinflated sense of ability through AI," Nosta said.

Part of the problem, Nosta said, is that AI reverses how humans normally think. In traditional reasoning, people move from confusion to exploration to structure — and only then to confidence. AI reverses that sequence.

"Coming to the answer first is an inversion of human cognitive process," Nosta said.

If that inversion becomes the norm, the stakes are larger than productivity. "Human cognition is on the obsolescence chopping block," he added.

The illusion of expertise

At work, fluency is often mistaken for competence.

Rebecca Hinds, head of the Work AI Institute at Glean, said AI can create the illusion of expertise. It's becoming harder to tell where the worker's knowledge ends, and where the technology begins, she said.

Her research details two potential outcomes.

Used intentionally — especially in areas where a worker already has expertise — AI can create a "cognitive dividend," freeing up time and sharpening judgment.

Used reflexively as a shortcut, it creates "cognitive debt," making people faster while quietly eroding their skills.

The difference is in whether AI supports thinking or replaces it.

Deskilling at the beginning of your career

The risk of deskilling is especially acute for early-career workers, Hinds said.

Junior roles have traditionally been training grounds to learn how to break down a messy problem, fix what's broken, and defend your thinking when someone challenges it.

Without that experience, workers can appear competent without ever developing real expertise.

The full impact of that shift may take years to fully show up. But the early signs are already visible, and those most at risk are the ones early in their careers.

"Right now, most professionals learned their craft before AI, so they have the baseline," Jan Tegze, author of "Job Search Guide" and "How to Talk to AI," told Business Insider. "The risk is with those who never build that baseline at all."

Ben Eubanks, chief research officer at human capital advisory firm Lighthouse Research & Advisory, said the gap between learning concepts in school and applying them in real-world work has always existed, but AI is widening it.

"You don't have to go put your hands on a problem, wrap your mind around a solution, or challenge the way things have always been done," he said. "You get to ask AI instead, and it gives you a neat, tidy solution instantly."

That shift, he added, is making it harder for early-career workers to build resilience.

At the same time, some companies are beginning to reinforce that behavior.

Employees are increasingly being evaluated on how often they use AI tools, Hinds said, rewarding speed and output over deep understanding.

Ashley Herd, CEO of Manager Method in Atlanta, which offers practical training for leaders, said she has seen the push to include AI usage in performance reviews increase over the past six months, especially in tech firms.

However, problems show up when something breaks, or when workers have to think through a problem on their own, said Sara Gutierrez, chief science officer at talent assessment firm SHL.

Anderson felt that shift firsthand when he stepped back into his work.

"I was aware of how it worked," he said. "But I didn't get those [workout] reps over those three months."

That's prompting some leaders to rethink training. Mehdi Paryavi, CEO of digital economy think tank the International Data Center Authority, said companies may need "mental gyms," spaces where employees can deliberately practice AI-free problem-solving in the same way you use physical gyms to build muscle.

'My swing was off'

Anderson's experiment shows what happens when those mental workouts don't happen.

By the end of his experiment, he was proud of what he'd built. But when he stepped back in to make changes himself, something felt off.

He compared it to watching someone play golf. You can study the swing. You can understand the mechanics. You can say, "That's how I should do it."

But that doesn't mean your body knows how to do it. And when he finally took the club back into his own hands, he felt it.

"My swing was off," he said of writing the code himself again. "I'm like, 'but I know how to do this,' but I couldn't get my body to move the way I wanted it to."