A common belief about workplace AI goes like this: stop doing a task yourself, and you lose the skill for it. It’s the same logic as muscle atrophy in the human body, and this belief has a name: skill atrophy.

Anthropic’s latest Economic Index report, based on a linked survey of 9,700 professionals, found the opposite. The people who handed AI their entire task, not just a draft, turned out to be the most confident about their careers and the most likely to say their skills were growing in value.

The Skill-Atrophy Myth Anthropic Just Debunked

The assumption behind skill atrophy is straightforward. Hand the thinking off to someone, or something, like an AI tool, and the muscle for it weakens. Anthropic’s data tells a different story once you look at the heaviest delegators specifically.

The people who delegate to Claude the most are the most optimistic about their future labor market outcomes, and feel their skills are growing in value.

Three findings carry that thesis:

• Heavy delegators report learning at the same rate as everyone else, not less.
• 68% say AI helps them learn more, a figure that barely shifts whether someone delegates a little or a lot.
• 57% say AI has made their skills more valuable, and that number climbs the more they delegate, not less.

Judgment Skills in the Age of AI: What the Heaviest Delegators Do Differently

Take a task most knowledge workers recognise: drafting an article.

Done through back-and-forth chat, it takes a median of 13 rounds of revision.

Done through full delegation, where someone gives clear instructions and lets AI run with the task, it takes one well-built prompt and a single round of checking.

That is not less effort. It is effort moved somewhere else, away from producing the draft and into deciding what good looks like before handing it off, then verifying what comes back. The heaviest delegators did not stop thinking. They shifted where they apply the thinking.

That shift has a name, and it might be the defining soft skill of this decade: judgment. Knowing what to hand off, what to keep, and how to verify what comes back is not a technical skill. It is a human one, and it compounds with practice rather than eroding with use.

Judgment is a non-technical skill that compounds with practice over time.

There is a pattern in who holds onto that judgment hardest, and it is not random. Workers with 15 or more years of experience rate AI’s ability to do their job about 10 percentage points lower than first-year employees do.

This is not denial. Experienced workers consistently point to the same gap: the trust-building and relationship management AI cannot touch. That is the exact judgment the heaviest delegators are practising every time they decide what is safe to hand off.

A Lesson From Steel Country

Charles Schwab was, by most accounts, the first executive in America ever paid close to a million dollars a year, more than the sitting US President at the time. He ran Carnegie Steel, and later US Steel.

Andrew Carnegie was direct about why he paid Schwab that much. Schwab did not know steel better than the engineers under him. What he had was rarer, the ability to walk into a mill, make a person believe in their own better work, and then get out of the way.

Carnegie knew exactly what he was paying for, because it is the same thing he wanted said about himself once he was gone. He wrote his own epitaph long before he died.

Here lies a man who knew how to enlist in his service better men than himself.

That line is the Delegation Paradox in one sentence. Judgment does not disappear when a task is handed off. It moves into knowing who, or what, has earned the trust to run with it, and into having the discipline to actually step back once that trust is given.

Carnegie hired Schwab, then got out of his way. Neither man became less relevant for doing this. It is the reason their names outlasted their companies.

Closing Thoughts:

The fear behind skill atrophy assumes that handing work to AI means handing away your thinking. Anthropic‘s research suggests the opposite happens for the people who delegate well. The thinking does not disappear, it relocates, into deciding what is worth handing off and verifying what comes back. That is judgment, and like any skill, it grows with deliberate practice, not avoidance.

If your team is still treating AI delegation as a productivity shortcut rather than a skill worth building on purpose, Radiant Institute’s AI enablement training is built around exactly this gap. Reach out if you would like help turning judgment into a trainable, repeatable part of how your people work with AI.