The constraint here is cognitive atrophy. Offload too much judgment to an AI tool and the underlying skill stops compounding. Sandra Matz, a professor at Columbia Business School, frames the career risk in specific terms: workers who deploy AI incorrectly risk "gradually falling behind and becoming complacent," a dynamic she describes as becoming professionally "boring."
How the skill gap opens
Complacency is the surface symptom. The structural issue is that professional skills develop through repeated decision-making, and AI short-circuits that loop when it substitutes for judgment rather than assisting it. A worker who lets AI draft every analysis, frame every argument, and flag every risk stops practicing those operations. Over months and quarters, the gap between the worker's raw capability and the tool's output widens, and the worker grows dependent on the tool to perform at a level they can no longer reach independently.
Matz's frame points to something the productivity discourse usually misses. The standard argument for AI at work is throughput: more output, faster. What that argument skips is the question of what kind of worker is left at the end of the process.
Where this sits in the stack
The workers most exposed are those in roles where judgment is the primary deliverable. Writing, analysis, strategy, client counsel. These are tasks where AI assistance is most tempting and output is hardest for a manager to benchmark against the worker's actual underlying capability.
The risk Matz identifies is cumulative, and quiet. A worker who over-relies on AI may hit short-term performance targets. The lag is the problem. Skill atrophy compounds slowly. By the time a manager or the worker notices the gap, catching up requires deliberate practice that most workloads don't budget time for.
What the framing implies for adoption pattern
Matz's argument is about deployment pattern. Where AI replaces the hard cognitive work instead of handling the administrative or mechanical layer beneath it, the worker pays the career cost.
The phrase "gradually falling behind" is doing real work in that framing. It is a rate-of-change argument. Peers who use AI to amplify judgment they continue to develop will accumulate advantage over time. Workers who use it to skip that judgment accumulate a deficit instead.