Rewiring Work: AI Isn't Killing Skills, It's Reskilling Them

Give me a break — the headline in Business Insider gets this only half right. Calling this moment “The Great AI Deskilling” captures a visible trend: tools are making routine cognitive work easier and cheaper. But the claim feels like a baseball card — big, shiny, and missing context. Here’s what nobody tells you: deskilling is an effect, not a destiny. It shows up where incentives and labor markets let it.

The piece frames deskilling as a broad wave, implying that lowering the skill floor is the dominant story. That’s true in narrow pockets: templated writing, simple data entry, basic legal forms — tasks that were once gatekept by credentialed workers are getting automated and standardized.

But the deeper pattern is uneven. Some jobs get deskilled at the task level while the role itself gets upskilled. A salesperson using AI for lead prioritization may spend less time on manual research and more time on negotiation strategy; the tool removes grunt work and creates demand for judgment, persuasion, and relationship management.

That nuance matters because treating deskilling as a universal law produces blunt fixes. Training programs that only teach “how to use AI” will leave people behind if firms quietly reprice roles downward. Conversely, assuming every displaced worker can just “retrain into AI” ignores real frictions: time, childcare, geography, credit constraints. From running large ops teams, I saw the same pattern with past tech waves: companies strip tasks first, redesign jobs second, and rarely pay a premium for new human skills unless competition, regulation, or customer pressure forces their hand.

Look at how this has played out before. When spreadsheets turned finance departments upside down, basic number-crunching work evaporated. Entry-level roles changed, but we didn’t see “The Great Finance Deskilling” — we saw the bar rise for analysis, storytelling with numbers, and cross-functional communication. The risk now is different: AI tools don’t just speed up work; they can hide complexity so well that nobody in the building actually understands what’s going on.

Say Business Insider is right that AI lowers the bar for many tasks. The immediate employer choice is simple: cut headcount, or redeploy. Most businesses prefer the short-term ledger line that improves margins. And the labor market doesn’t magically correct for lost craftsmanship. If wages fall and employers stop investing in training, you get lasting deskilling: fewer people learning the deep versions of a craft because the job no longer requires it and no one wants to pay for it.

But spare me the pure doom narrative. New specialties will keep emerging around prompt design, model oversight, integration testing, and AI risk management. Governance roles will appear to manage hallucinations, bias, and safety trade-offs. Those jobs aren’t equivalent to the old craft positions; they sit in different parts of the org chart, pay differently, and require different career pathways. That’s the squeeze point: some workers get upgraded roles, others get pushed into lower-paid, heavily monitored work that’s “AI-assisted” in name and algorithmically supervised in practice.

Here’s what nobody tells you about “market forces will fix it.” Critics of the deskilling argument say technology historically creates more jobs than it destroys. Fair point; sometimes true. But that’s a conditional claim: only if those new jobs are located where displaced workers live, only if they’re reachable without years of unpaid retraining, and only if employers are willing to fund transition periods. Those conditions do not automatically appear just because someone writes a think piece about innovation.

We’ve seen a version of this with airline pilots and fly-by-wire systems. Automation made flying safer and reduced routine workload, but regulators and airlines doubled down on training and recurring checks. Where that training lagged, incident reports started to flag pilots over-reliant on automation and slow to react in edge cases. The technology didn’t deskill people on its own; the mix of incentives, regulation, and culture did.

Policy should target levers, not vibes. Start with wage-linked training: subsidize on-the-job learning tied to actual advancement, not just attendance or online certificates. Update bargaining rules and job-classification standards so “AI adoption” doesn’t become a loophole for downgrading roles while keeping the same accountability. And get sector-specific: in healthcare, infrastructure, and safety-critical industries, require audit trails and human-in-loop thresholds so organizations can’t quietly hollow out expertise and then blame “the algorithm” when something breaks.

Wake up — the article also underweights how corporate accounting and human incentives shape outcomes. If managers are rewarded for immediate headcount cuts, they’ll automate first and treat human capital as an afterthought. If regulators, unions, investors, or customers demand resilience and accountability, those same managers will redesign workflows to keep meaningful skill in the loop. Change the scorecard and you change the behavior.

There’s a quieter danger here: cultural and institutional memory. When institutions stop requiring mastery, they also stop building deep bench strength. Processes that once carried tacit knowledge — the weird exception you only learn by troubleshooting it three times at 11 p.m. — become brittle flows glued together by prompts and templates. Training that used to happen through apprenticeship gets replaced by “just ask the bot,” which works right up until an edge case appears and no one remembers why the old workaround existed.

You want to blunt deskilling? Make the cost of losing skill visible. Require operational audits that flag where AI has replaced human judgment, mandate incident reviews that document when AI-made errors expose lack of human expertise, and tie procurement and compliance policies to demonstrated human oversight instead of checkbox “AI adoption” slides.

If Business Insider is seeing the first wave, the next headline won’t be about “The Great AI Deskilling” — it’ll be about the first big AI-driven failure where everyone realizes the humans left in the loop were never taught how to steer.