What AI Can Do for Us — and What We Still Need to Own
What happens when AI takes over more of the work? Insights from Anthropic’s Economic Index on delegation, experience, learning, and the human value that remains.
CAREER & CRAFT
Priyanjit Purkayastha
9/11/20264 min read
There is a particular anxiety that accompanies almost every conversation about AI and the future of work.
The more capable AI becomes, the more work it can potentially do. And if AI can do more of our work, the obvious question follows:
What happens to us?
Will our skills become obsolete? Will relying on AI make us less capable? Will the people who delegate the most eventually become the easiest to replace?
I recently went through Anthropic's latest Economic Index report, Cadences, and one finding caught my attention because it challenges that assumption:
The people who delegate the most to AI are also the most optimistic about their future.
The report combines millions of real-world Claude interactions with a survey of about 9,700 users, looking at both how people use AI and how they perceive its impact on their work and careers.
More delegation doesn't necessarily mean more fear
Anthropic distinguishes between “augmentation,” where humans remain actively involved, and “automation,” where AI is asked to complete a task with little or no input.
One might expect those handing over more work to AI to be more worried about their jobs.
The data suggests otherwise.
People with a higher share of automated Claude sessions are more optimistic about AI's impact on their future—including pay, job security, ability to find work, meaning, autonomy and human interaction.
This doesn't prove that using AI causes optimism. It could simply be that people who are already enthusiastic about AI are more willing to delegate to it.
But the relationship is still interesting.
Perhaps they aren't thinking, “AI is going to take my job.”
Perhaps they're thinking:
“This can take a lot of work off my plate.”
Experience may matter more, not less
Another finding really stood out to me.
Workers with 15+ years of experience estimate that AI can currently perform roughly 10 percentage points less of their work than people in their first year.
Experienced workers particularly point to judgment, contextual awareness, situational reasoning, trust-building and managing people as areas where AI struggles to replicate them.
This made me think about how we define a job.
We often describe it as a collection of tasks.
But a career is more than that.
Experience teaches you which task matters, what information is missing, when a technically correct answer is actually the wrong answer, whom to involve, what risks are acceptable, and how to get people aligned.
Much of this knowledge isn't written down. It is accumulated.
Perhaps, then, experience doesn't become irrelevant in an AI world.
Its value may shift—from knowing how to execute something to knowing what deserves to be executed in the first place.
So what should we do with AI?
The report also offers an interesting perspective on learning.
68% of respondents say they learn more with AI, while 57% say AI has made their skills more valuable. Heavier AI delegators report learning at roughly the same rate as other users, while their perception of the value of their skills increases with automation.
Of course, these are self-reported perceptions, so this doesn't rule out skill erosion.
But it raises a useful question:
What do we do with the capacity AI gives back to us?
If AI writes the first draft, do we simply accept it?
Or do we use the saved time to question the assumptions, understand the subject more deeply and make better decisions?
That choice remains ours.
And perhaps this is the real opportunity for someone 10–15 years into a career.
Don't just ask:
“How do I learn AI faster?”
Also ask:
“What should I stop doing myself?”
Let AI handle more of the execution.
Then spend the time it gives back on the things that are harder to delegate:
Judgment. Context. Decision-making. Relationships. Trust.
To me, that suggests a possible shift in the future of work:
AI increasingly handles execution.
Humans increasingly own direction, judgment and responsibility.
The people leaning into AI the hardest aren't necessarily the ones who feel most threatened by it.
Maybe they are simply the ones learning, through experience, what they can hand over—and what they still need to own.
One caveat: the Economic Index Survey isn't representative of the general population. Its respondents skew heavily toward knowledge-work occupations, particularly computer/mathematical and management roles.
Source: Anthropic Economic Index: Cadences, published June 26, 2026.
