What 81,000 Claude Users Actually Want From AI

Last December, Anthropic did something I wish more AI companies would do: they actually asked their users what they think. Not through a sterile survey with Likert scales, but through a conversational interview conducted by Claude itself. Nearly 81,000 people from 159 countries, speaking 70 languages, sat down to talk about what they want from AI, what they’re afraid of, and whether the technology is actually delivering.

This is, as far as I can tell, the largest qualitative study ever done on this topic. And the results are refreshingly human.

Hope and fear live in the same person

The most striking finding isn’t a single number — it’s that hope and fear don’t divide people into opposing camps. They coexist inside the same person. A lawyer in Israel put it bluntly: “I use AI to review contracts, save time… and at the same time I fear: am I losing my ability to read by myself? Thinking was the last frontier.”

That tension runs through almost every response. People see AI as a tool that could free them from drudgery, but they’re also worried it might erode the skills they value most. A freelancer in the US described how Claude helped piece together a medical history that led to a proper diagnosis after nine years of misdiagnosis. Meanwhile, a technical support specialist in the US said they got laid off because their company wanted to replace them with an AI system.

Both things are true. The same technology that saves one person’s health can cost another their livelihood. And most users seem to hold both realities in their heads at once.

What people actually want

Anthropic classified responses into what people most wanted from AI. The top category, at 18.8%, was professional excellence — using AI to handle routine tasks so they can focus on higher-value work. A healthcare worker in the US described receiving 100-150 text messages per day from doctors and nurses, and said AI lifted the documentation burden, giving them more patience and time for families.

Personal transformation came next at 13.7% — people using AI as a coach, therapist, or guide for self-improvement. One respondent from Hungary said AI “modeled emotional intelligence for me… I could use those behaviors with humans and become a better person.” That’s not a use case you see in most product demos.

Life management — scheduling, mental load reduction, executive function support — hit 13.5%. A manager in Denmark said, “If AI truly handled the mental load… it would give me back something priceless: undivided attention.” That line hit me harder than any statistic.

Time freedom came in at 11.1% — people wanting to reclaim hours from work and chores to be present with family. An entrepreneur from Nigeria said, “I live hand to mouth, zero savings. If I use AI smarter, it may help me craft solutions to that cycle. It still depends on me.” That’s not blind optimism; it’s a realistic hope from someone who knows the technology alone won’t fix structural problems.

The fears are just as specific

People aren’t afraid of Skynet. They’re afraid of losing their edge, their jobs, their ability to think without a crutch. A software engineer from South Korea said, “Humanity has never dealt with something smarter than itself. We need to reflect on how to prepare for the AI age.” That’s not panic — it’s a sober assessment from someone who understands the technology from the inside.

What I find interesting is that the fears are often the mirror image of the hopes. People want AI to handle routine work so they can focus on meaningful tasks, but they’re afraid that outsourcing too much will atrophy their own skills. They want AI to save time, but they’re worried about what they’ll lose in the process.

The method matters

Anthropic used a custom version of Claude — the Anthropic Interviewer — to conduct these conversations. This is a genuinely interesting approach to qualitative research. Normally you have to choose between depth (small sample, rich interviews) or breadth (large surveys with shallow responses). Here they got both: open-ended conversations at massive scale.

They then used Claude-powered classifiers to categorize responses across dimensions like what people want, whether they’re getting it, what they fear, their occupation, and overall sentiment. Each person got one primary category for what they want from AI, but concerns were multi-label — because people tend to have several distinct worries, not just one.

They also published a Quote Wall where you can filter responses by region, concern, and vision. I spent way too long browsing it. Some of the quotes are heartbreaking, some are inspiring, and most are just… honest. People aren’t grandstanding. They’re describing their actual lives.

What this tells us

If you work in AI — building it, writing about it, investing in it — this study is worth reading in full. The public conversation about AI tends to oscillate between utopia and apocalypse. What’s missing is a grounded vision of what “AI going well” actually looks like for real people.

These 81,000 responses are that vision. It’s messy, contradictory, and full of caveats. But it’s real. And it’s a lot more useful than another thinkpiece about AGI timelines.

Anthropic published the full study with methodology details and limitations in an appendix. I’d encourage anyone serious about AI to read it. The data is richer than I’ve summarized here, and the regional breakdowns alone are worth exploring.