Anthropic asked 80,508 users what they actually want from AI

When Silicon Valley talks about AI, the conversation usually swings between apocalyptic doomerism and abstract trillion-dollar market projections. What gets lost is what everyday people actually want the technology to do for them.

Anthropic decided to find out. Last December, the company invited every Claude.ai user to sit down with Anthropic Interviewer, a prompted version of Claude designed to conduct conversational research. Over the course of a single week, the bot interviewed 80,508 people across 159 countries in 70 languages. Anthropic believes it is the largest and most multilingual qualitative study ever conducted.

More time to cook with mom

Qualitative research usually forces a brutal trade-off: you either get deep insights from a dozen people, or shallow survey checkboxes from thousands. Anthropic solved this by letting Claude ask adaptive follow-ups and then using model-powered classifiers to categorize the open-ended responses.

The top request was professional excellence at 18.8%, with users wanting AI to handle mundane tasks so they could focus on strategic work. Another 9% saw AI as an entrepreneurial partner. Personal transformation and life management followed closely, at 13.7% and 13.5% respectively.

But the best insights came when the bot dug beneath the productivity buzzwords. People who started out talking about automated workflows were often chasing something far more grounded. One worker in Colombia noted that finishing tasks early let them cook dinner with their mother. A Japanese software engineer recalled that when Claude first surpassed human quality on a work task, the real win was leaving on time to pick up their daughter from daycare.

Beyond corporate tasks, roughly 14% of respondents relied on AI as external scaffolding for executive function challenges like planning, memory, and task follow-through. Others leaned on it for non-judgmental learning, and in war-torn regions like Ukraine, some users turned to Claude for emotional support when human infrastructure collapsed.

Hope and anxiety under one roof

Overall, 81% of respondents said AI had already taken steps toward fulfilling their vision. Developers reported cutting 173-day engineering processes down to three days, while non-technical users broke barriers they thought were permanent. One entrepreneur in Chile who had run a butcher shop for over 20 years and barely touched a PC used AI to launch a new venture.

Yet the study made clear that people are not divided into warring camps of techno-optimists and doomers. Instead, hope and alarm live inside the exact same individuals.

The most widespread anxieties centered on job displacement and the loss of human cognitive ability. Users love having a patient assistant, but they quietly worry about what happens to their own brains and livelihoods when the assistant does all the heavy lifting.

Why it matters

For everyday users, this research validates AI as a personal cognitive prosthetic rather than just a workplace speed hack. Whether you are dealing with focus challenges, learning to code despite a disability, or trying to automate tedious admin work, the real value lies in buying back mental bandwidth and time for actual life.

For the AI industry, this study is a masterclass in modern qualitative research. Deploying an AI interviewer to conduct 81,000 adaptive conversations in 70 languages shows how frontier models can reinvent social science and product discovery at massive scale.

That said, we think the findings come with an obvious grain of salt. Every single interviewee was an active Claude user who voluntarily opted into an interview, making this an inherently tech-forward, self-selected crowd. Yet even within this favored demographic, deep-seated fears about cognitive atrophy and job destruction were pervasive. If the people who love AI the most are still terrified of what it might do to their minds and careers, tech labs have a massive societal gap left to bridge.