On June 26, Anthropic released the sixth "Economic Index" report, which for the first time compared the questionnaire responses of approximately 9,700 users with actualClaudeUse data hooks. The most shocking numbers in the report:More than 35% of respondents expect AI to perform most or almost all of their work tasks within 12 months. This is more aggressive than most forecasts in the market.
Automation goes beyond augmentation for the first time
A key turning point in the report: In APIs (professional workflows),Automation has surpassed enhancement as the dominant usage mode. By “automation” I mean completely delegating tasks toClaudeCompleted independently, while "augmentation" refers to humans and AI collaborating to complete tasks.
This is no small change. In the use of AI in the past few years, "AI is the co-pilot and humans are the captain" has been a repeatedly emphasized safety narrative. But Anthropic’s actual data shows thatIn production environments, people are already letting AI fly on its own.
The experience gap is widening, not narrowing
The most actionable finding in the report is the gap between experienced users and novice users. experiencedClaudeuserConsistently achieve better task completion results, because they view AI as a collaborative partner—iterating, providing context, validating output, and determining which tasks are suitable for AI and which require human judgment. Newbies often try to entrust everything, and the success rate is even lower.
More importantly,The gap is widening rather than narrowing. Experienced users learn AI faster than novices can catch up. This means that the earliest organizations to deploy AI not only have more AI workflows, but more importantly, their people have accumulated better AI judgment.
About 49% of jobs: at least 1/4 of the tasks have been completed by AI
Another core indicator tracked by the report is AI task penetration rate. About 49% of jobs have at least a quarter of their daily tasks passedClaudeFinish. This number has stabilized in recent reports, indicating that the breadth of AI task coverage may have entered a plateau period——AI has already entered most of the jobs it can enter, and the next step is to deepen, not broaden.
Classic optimism bias: 40% worry about being replaced in entry-level jobs, but think they are safe
Another interesting finding in the report: Respondents believe that entry-level jobs have a 40% chance of being affected by AI, but their own unemployment risk assessment is much lower.This is the classic optimism bias that appears again and again in every wave of automation.. The report points out that those who least underestimate the impact of AI tend to be those workers who are most exposed.
Three suggestions for action for businesses
- Protect junior talent: Redesign the onboarding process and focus on the judgment tasks that AI is not good at (tacit knowledge, contextual understanding, interpersonal management) as the focus of new employee training
- Accelerate experience accumulation: Don’t let AI just replace humans, but establish a feedback loop of “human-machine collaboration” so that the team can learn how to use AI better
- Facing the Reality of Automation: Automation has moved beyond enhancement in API production environments, and many enterprises’ internal AI policies may already lag behind actual operations
The strongest signal sent by this report is:AI isn’t changing jobs, it has changed jobs. 35% expect AI to be doing most of the work within a year - a number that might have been considered science fiction a year ago is now a prediction based on actual usage data. access AI Dash Discover more AI tool reviews and usage guides.