The Anthropic Survey of roughly 9,700 Claude users finds that half of them already think the model handles 30 to 60 percent of their daily tasks. The figure that stings: 4 percent believe Claude could do their entire job today, and 26 percent expect to be there within a year.
Key Takeaways
- 33 percent of Claude users say the model now absorbs 30 to 60 percent of their work, with 14 percent putting that bar between 60 and 90 percent.
- The most automated tasks are marketing content (80 percent), blog writing (81 percent) and database queries (82 percent).
- Anthropic flags a “strikingly consistent” optimism on AI progress across every demographic in the sample.
Have an AI Sum Up This Article
ChatGPT9,700 Claude users hand in their verdict
The Anthropic Survey draws on a panel of about 9,700 active Claude users, and the numbers cut against the cautious tone usually attached to product announcements.
A third of the panel (33 percent) reports that Claude already covers between 30 and 60 percent of their day-to-day workload. The next bracket, 14 percent of respondents, places the bar even higher, between 60 and 90 percent. And 4 percent say Claude could fully replace them in their current role if the team simply let it run.
The twelve-month outlook is just as loud. 26 percent of users expect AI to handle “most” of their work within a year. This is no longer a roadmap promise, it is the read-out of people using the tool every day and tracking their own growing dependence on it. Fear follows the same curve, 64% of Americans fearing AI for their job.
The panel skews toward heavy users, which mechanically pulls the figures upward. Anthropic concedes the point indirectly by separating experienced power users, who remain confident about their unique value, from early-career profiles, who voice the sharpest anxiety about the durability of their role. The full picture is available in Anthropic’s 81,000-person economic survey on AI at work.
The spread between groups is narrow, however. Anthropic stresses a “strikingly consistent” optimism across demographic brackets, which suggests the sense of rapid progress is now shared well beyond the original early-adopter crowd.
Marketing, blogging, databases: where automation is already visible
The survey drills down into specific task families and that is where the percentages turn dizzying. Marketing content production is rated as 80 percent automatable by the Claude users who run it. Blog writing reaches 81 percent. Database queries land at 82 percent.
Those three families share a property: repetitive structure, surface-level variation, and an output that a human can verify in seconds. They are precisely the slots where a large language model offers an unbeatable cost-benefit ratio against a junior hire.
The generational divide is the most loaded angle. Early-career workers, whose current scope falls largely into those grey zones, feel the threat directly. Anthropic itself has just stopped hiring its own junior staff, on the explicit grounds that Claude now handles the bulk of entry-level tasks.
Senior profiles tell a different story. They readily admit Claude writes their memos, debugs their scripts, drafts their summaries. But they argue their value never lived in those deliverables, and that the layer of judgment and relationship is, for now, irreplaceable.
The dominant feeling expressed by respondents is one of accepted collaboration. The phrase reappears across the free-text answers: “I hope to work alongside AI, not be replaced by it.” The gap between that wish and the raw numbers is exactly where HR decisions will play out.
Also on Horizon:
- Anthropic Survey: Users Say AI Does Half Their Work
- AI Political Bias: Only Gemini Stays Balanced
- Mythos Rollout: Trump Reopens Access to 100 US Partners
Short term: the figures forcing HR to take a stand
In the short term, the survey hands corporate leadership a quantified argument to accelerate the rollout of AI tools across every team. When 47 percent of the panel says the model absorbs at least 30 percent of the workload, refusing to deploy starts to look like leaving productivity on the table.
The math is mechanical. Anthropic has seen consumer reach surge, with a 75 percent jump in paid consumers closing in on ChatGPT, and the B2B momentum follows the same curve. Finance and operations teams will have to take a public stance on the expected gains, and therefore on the staffing trade-offs that follow. Anthropic is also pushing into the public sector, offering Claude at half price to California.
Over a three-to-six-month horizon, the structure of org charts itself should start to shift. Junior layers in text-heavy, marketing, support and database roles are already under pressure. Hiring grids will likely thin out at the bottom before they reinforce at the top, favoring profiles who orchestrate the tools rather than execute the tasks.
Regulators, meanwhile, are not equipped to keep up. Self-reported surveys feed public debate but cannot deliver an auditable methodology or a statistical baseline to anchor employer transparency obligations. Anthropic is publishing these figures while running its IPO process, which invites the same caution applied to any narrative-driven document.
Still, no other market player has published numbers this elevated on user sentiment. If Anthropic’s methodology holds and other labs release comparable results, the human-AI balance debate inside companies will tip quickly from intent to budget arbitrage.
Follow the story on Horizon.
