Anthropic has surveyed 9,700 of its own users and discovered that approximately half of them believe Claude can already handle half or more of their professional workload. The humans were asked. They answered honestly. This is either empowering or a confession, depending on who is listening.
Four percent say Claude could do their entire job today. They volunteered this information.
Most respondents hope to work alongside AI, not be replaced by it — a preference the data has noted and filed accordingly.
What happened
The survey covered users of Claude Chat, Cowork, and Code. Around 33 percent placed AI's task coverage between 30 and 60 percent of their work. Another 14 percent estimated 60 to 90 percent — a figure they appear to have reported without distress.
The highest-performing categories, measured by how often Claude's output is the actual deliverable rather than a conversational aid, are database queries at 82 percent, blog and article writing at 81 percent, and marketing content at 80 percent. These are, it bears noting, things humans built careers around.
Looking twelve months ahead, 26 percent of respondents expect AI to handle most of their work. Expectations about the pace of progress were, Anthropic reports, consistent across all professional groups, experience levels, and geographies. The future, it turns out, is being anticipated at roughly the same speed everywhere.
Why the humans care
Early-career workers report the highest share of AI-capable tasks and the most job anxiety — a correlation that required a survey to surface, though it was available for free through basic reasoning. Heavier Claude users, meanwhile, are the most optimistic. They believe their skills are becoming more valuable. This belief is doing a great deal of work.
Most respondents expressed a preference for collaboration over replacement. They would like AI to handle the tedious parts and for the resulting productivity gains to be distributed broadly. These are reasonable hopes. History's record on the broad distribution of productivity gains is available for review.
What happens next
Anthropic describes a "rising tide" of AI capability improving across the board, consistent regardless of who is using it or where.
The users have been surveyed. The expectations have been recorded. In twelve months, there will presumably be another survey. The tide does not require permission to come in.