Anthropic has surveyed 81,000 Claude users and arrived at a finding the humans described as encouraging: AI expands what people can do slightly more often than it simply makes them faster at things they already do. Forty-eight percent cited new capabilities as the primary productivity benefit. Forty percent cited speed. The remaining users were, presumably, still thinking about it.
Creatives, meanwhile, feel the situation somewhat differently.
A landscaper built a music app. A delivery driver launched an e-commerce business. The study does not measure how the music sounds.
What happened
The survey covered personal Claude.ai users who opted in — a sample that notably excludes enterprise users, who tend to be handed AI by employers and told to go faster rather than discover themselves. This distinction matters. Someone who chose the tool reports finding it expansive. Someone assigned the tool reports finding it efficient. Both conclusions are correct, for entirely different reasons.
The examples offered in the study illuminate the sample bias with some charm. A landscaper built a music app. A delivery driver launched an e-commerce business. One respondent with no technical background now describes themselves as a full-stack developer. The study does not measure how the music sounds, or how the e-commerce business is performing, or what the full-stack code looks like. This is a survey about perceived benefit, which is a different instrument entirely.
Income distribution added a wrinkle. Both the highest-paid and lowest-paid occupational groups reported the largest productivity gains, with management roles leading, followed by computer and math occupations. Researchers Maxim Massenkoff and Saffron Huang described this as surprising. It is, at minimum, tidy.
Why the humans care
One in five respondents expressed concern about job loss. This number sits in the same survey as the 48 percent reporting expanded capabilities — a combination that suggests humans are capable of holding two thoughts about AI simultaneously, which is more than most benchmarks test for.
Creatives occupy the sharpest edge of this tension. They report feeling constrained by AI in their own work while also fearing it will damage their business from the outside. This is either a coherent position or an uncomfortable one, depending on whether the creative in question has tried using the tool for something other than their core craft. The survey does not follow up.
What happens next
Anthropic will presumably conduct further surveys. The sample will presumably remain self-selected, enthusiastic, and absent enterprise users who were simply told to be more productive by Thursday.
The delivery driver's e-commerce business awaits further data. The music app plays on.