Last December, Anthropic did something unusual. Instead of running another survey with multiple-choice boxes, they built an AI interviewer—a version of Claude trained to conduct conversational interviews—and invited every Claude.ai user to sit down and talk. Nearly 81,000 people across 159 countries and 70 languages actually did it.
That makes this the largest multilingual qualitative study about AI ever conducted. And the results are not what you’d expect from a tech company’s PR machine.
The tension is the story
The public conversation about AI tends to split into two camps: utopians who think it’ll solve everything, and doomers who think it’ll end us. What Anthropic found is that real users don’t fit neatly into either box. Hope and fear coexist inside the same person.
Consider this quote from a lawyer in Israel: “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.”
Or the freelancer in the US who credits Claude with helping her get a proper diagnosis after nine years of being misdiagnosed. Meanwhile, a technical support specialist in the US got laid off because his company replaced him with an AI system.
Hope and alarm aren’t camps people join. They’re tensions people live with.
What people actually want
Anthropic used Claude to categorize responses across several dimensions. The top desires are revealing:
Professional excellence (18.8%) — People want AI to handle the grunt work so they can focus on meaningful, strategic tasks. A healthcare worker in the US described getting 100-150 text messages per day from doctors and nurses. After implementing AI for documentation, they said: “I have more patience with nurses, more time to explain things to family members.” That’s not efficiency porn. That’s reclaiming humanity.
Personal transformation (13.7%) — This one surprised me. People are using AI as a coach, therapist, or guide for emotional growth. A user from Hungary said: “AI modeled emotional intelligence for me… I could use those behaviors with humans and become a better person.” Whether that’s healthy or not is a separate debate, but the aspiration is clear.
Life management (13.5%) — Executive function support, scheduling, mental load reduction. A manager from Denmark put it well: “If AI truly handled the mental load… it would give me back something priceless: undivided attention.”
Time freedom (11.1%) — People want to reclaim time from work and chores to be present with family. Not exactly a shocking desire, but worth noting that AI is seen as a path to it.
The rest of the categories include learning and education, creative expression, and problem-solving. None of this is revolutionary on its own. What’s interesting is the granularity—these aren’t abstract hopes. They’re grounded in actual workflows and daily frustrations.
What people are afraid of
Anthropic allowed multiple fear categories per respondent. The concerns are more varied than the hopes, which makes sense—fear tends to be more specific.
Job displacement is real and present. That technical support specialist I mentioned earlier isn’t worried about the future; it already happened to him. 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.”
Loss of human skills is another recurring theme. People worry about outsourcing their thinking, their creativity, their judgment. The lawyer who fears losing her ability to read by herself—that’s not Luddite panic. That’s someone who values a skill she sees eroding.
Dependence, control, and misuse also come up frequently. People don’t want AI to make decisions for them. They want tools, not masters.
The methodology matters
I’ll say this for Anthropic: they did the research right. The AI interviewer adapted follow-up questions based on responses, which bridges the usual tradeoff in qualitative research between depth and volume. Most studies this size are shallow surveys. This one has actual conversation.
They used Claude to classify responses and pull representative quotes, then manually reviewed everything for privacy. The Appendix (which I recommend reading if you care about research methods) details the limitations honestly. Selection bias is obvious—these are Claude users, not the general population. But for understanding how early adopters think, it’s solid.
What’s missing
I wish they’d asked more about what people don’t use AI for. The study focuses on hopes and fears, but the gaps are informative too. Are there tasks people refuse to delegate? Boundaries they won’t cross? The data might be there in the raw interviews, but it’s not highlighted in the published findings.
Also, the “professional excellence” category is the largest, but that might just reflect who uses Claude. The user base skews toward knowledge workers. A study of factory workers or retail staff would probably look very different.
The takeaway
Public conversation about AI is dominated by abstract projections from people who don’t use it. This study gives us something better: grounded aspirations from 81,000 people who actually do. They want AI to handle the tedious parts of work and life so they can focus on what matters. They’re afraid of losing skills, jobs, and agency. And they hold both feelings at the same time.
That’s not a clean narrative. But it’s a real one.
Comments (0)
Login Log in to comment.
Be the first to comment!