On March 5, 2026, Anthropic – the company behind Claude – released a new research report based on the analysis of millions of real conversations on Claude to examine which tasks AI is actually being used to replace.

What makes this study notable is that the data does not come from predictions or surveys, but from real conversations between users and AI systems. As a result, the report provides a clearer picture of how AI is actually being applied in people's daily work.

The major gap between AI capabilities and real-world adoption

The most important finding of the report is:

AI can perform up to 94% of tasks in the IT industry
But actual usage is only around 33%

This shows that the gap between what AI “can do” and what it is “actually being used for” remains very large.

In other words, although AI is already capable of supporting or performing many tasks, its practical adoption in businesses and organizations is still at an early stage. This also means that the AI-driven transformation is only just beginning.

Jobs most affected by AI

According to data from Anthropic, many occupations in the knowledge workforce are already experiencing significant impacts from AI.

The occupations with the highest proportion of tasks already “covered” by AI include:

Software developers: 75% of tasks can already be performed or supported by AI
Customer service representatives: 70%
Data entry workers: 67%
Medical records specialists: 67%
Market research analysts: 65%
Financial analysts: 57%

These occupations have one thing in common: they involve significant amounts of data processing, information analysis, or written communication – areas in which AI is developing rapidly.

Occupations affected by AI

Jobs AI has barely reached

In contrast, jobs that require physical labor or direct interaction in real-world environments remain relatively less affected by AI.

According to the report, the following occupations have currently seen little significant AI replacement:

Chefs
Auto mechanics
Emergency responders
Bartenders

What these occupations have in common is their reliance on physical labor, motor skills, and real-world situational judgment – factors that AI still finds difficult to replace completely.

Highly educated workers are more affected

One surprising finding from the study is that the occupations most affected by AI are not primarily low-skilled jobs, but highly educated knowledge-work roles.

According to Anthropic's analysis:

The group most affected by AI has an average income 47% higher than the group least affected
The proportion of people with postgraduate degrees is four times higher

This shows that AI is having a strong impact on work involving knowledge and data processing – areas that were previously considered relatively stable.

No signs of mass unemployment yet, but opportunities for young workers are declining

Although AI is increasingly capable of performing a wide range of tasks, Anthropic's report found no clear evidence that AI has caused a large-scale wave of unemployment.

However, one notable signal is that the labor market is changing.

The number of job openings for young workers aged 22–25 in occupations more exposed to AI has declined by around 14% compared with 2022.

This suggests that businesses may be becoming more cautious about hiring new employees for roles that AI could support or replace in the future.

AI is still in the early stages of transformation

The gap between technological capabilities and actual adoption shows that AI still has a long way to go before becoming a widely used tool across every field.

According to experts, this also creates opportunities for businesses and workers: organizations that adopt AI early may increase productivity and gain a competitive advantage, while workers need to adapt quickly to changes in the labor market.

Nevertheless, the data from Anthropic's report makes one thing clear: AI is already capable of doing more than we might think, but real-world adoption has only just begun.

The early stage of AI transformation

In the coming years, as businesses gradually integrate AI into their operational processes, the gap between what AI “can do” and what it is “actually doing” is likely to continue narrowing. This will require both businesses and workers to adapt more quickly to changes in the way work is performed, particularly in occupations involving data and knowledge processing. Leave your email to receive the full analysis of this report and upcoming trends in AI adoption in businesses.

New AI report: The major gap between capabilities and real-world adoption