According to Business Insider, a new report released Wednesday from the Work AI Institute, produced with researchers from Notre Dame, Harvard, and UC Santa Barbara, warns that AI is creating an “illusion of expertise” in office workers. Rebecca Hinds, head of the institute and the report’s coauthor, says workers feel smarter and more productive while their fundamental skills slowly erode. She compares the effect to the early days of search engines, but argues it’s more powerful and risky with generative AI. Hinds specifically flagged that leaders who stack-rank employees based on AI tool clicks are exacerbating the problem, incentivizing shallow use over real understanding. The report concludes that AI can create a “cognitive dividend” or “cognitive debt,” depending on whether it’s used as a deliberate partner or a reflexive shortcut.
The sneaky skill drain
This is one of those reports that feels obvious the second you read it, but we’ve all been avoiding thinking about it. The core idea is terrifyingly simple: if a tool does the hard, messy, thinking part of your job for you, you never actually learn how to do that thinking yourself. Your brain gets a free pass. Hinds uses the perfect example of using AI to beat the “blank page” and generate a first draft. Sure, it’s faster. But you skip the entire internal struggle of structuring arguments, finding the right words, and wrestling with ideas. That struggle is where the real learning and ownership happens. As she puts it, “The more you poke holes in it, the more it feels yours.” Without that process, your skills atrophy. You become a manager of outputs, not a crafter of ideas.
Why early-career roles are most at risk
Here’s the thing: this isn’t a uniform threat. The report nails it by saying the highest exposure is in early-career, apprenticeship-style jobs. Think about a junior developer who never has to debug gnarly code because Copilot fixes it, or a marketing associate who never learns campaign structure because ChatGPT drafts everything. These are the foundational years where you build the mental models and muscle memory for your career. If you automate that foundational work away, what are you building your senior expertise on top of? A shaky foundation of prompt engineering. You might look productive, but you’re not building durable, transferable skills. That’s a huge long-term risk for both the individual and the company.
How leaders are making it worse
And now for the really frustrating part. Hinds points out that leadership metrics are often making this illusion-of-expertise problem worse. In some companies, AI tool usage clicks are tied to performance reviews. Let that sink in. Employees are literally incentivized to click a button more often, not to use the tool thoughtfully or to achieve a better business outcome. It’s the digital equivalent of measuring productivity by how many hours someone sits at their desk. This approach encourages the most shallow, checkbox-style use of AI possible. It completely misses the point. As Hinds argues, companies should tie AI to real business goals—quality, innovation, customer satisfaction—not vanity metrics. But that’s harder to measure, so we default to counting clicks.
Shifting from debt to dividend
So, what’s the fix? The report isn’t anti-AI. It’s about being deliberate. The goal is to achieve a “cognitive dividend”—where AI frees up your mental bandwidth for higher-order judgment—and avoid the “cognitive debt” of weakened skills and bloated confidence. Hinds recommends workers and leaders ask three key questions about any AI use: Is it improving the quality of my core work? Is it helping me develop new skills? Is it making the organization better? The blunt truth, which Hinds states perfectly, is that AI “does not magically transform you as a leader. More often, it amplifies what already exists.” If you have a culture that values quick wins over deep understanding, AI will supercharge that. It’s a mirror, not a magic wand. We need to be honest about what it’s reflecting.
