Does automating the boring stuff in DS actually make you worse at your job long-term

Been thinking about this a lot lately after reading a few posts here about people noticing their skills slipping after leaning too hard on AI tools. There's a real tension between using automation to move faster and actually staying sharp enough to catch when something goes wrong. Like, automated data cleaning and dashboarding is genuinely useful, but if you're never doing, that work yourself anymore, you lose the instinct for spotting weird distributions or dodgy groupbys. There was a piece from MIT SMR recently that made a decent point that augmentation tends to win over straight replacement in the, long run, partly because the humans who stay engaged are the ones who can actually intervene when the model quietly does something dumb. And with agentic AI workflows becoming more of a baseline expectation in 2026, that intervention skill matters even, more since these pipelines are longer, more autonomous, and way harder to audit when something quietly goes sideways. The part that gets me is the deskilling risk nobody really talks about honestly. It's easy to frame everything as augmentation when really the junior work just disappears and, the oversight expectation quietly shifts to people who are also spending less time in the weeds. The ethical question isn't just about job numbers, it's about whether the people left are, actually equipped to catch failures in automated pipelines or whether we're just hoping they are. Curious if others have noticed their own instincts getting duller after relying on AI tools for, a while, or whether you've found ways to keep that hands-on feel even in mostly automated workflows.