Anthropic has released a dataset of 1,250 interview transcripts documenting how professionals across industries integrate artificial intelligence into their work. The conversations, conducted by an AI interviewer rather than human researchers, represent one of the largest qualitative studies examining AI's real-world impact on employment. The complete dataset is now available on HuggingFace under a CC-BY license.
A New Approach to Understanding AI's Societal Impact
The project introduces Anthropic Interviewer, a Claude-powered tool designed to conduct in-depth qualitative research at unprecedented scale. Traditional interview-based research faces inherent constraints: human researchers can only conduct so many conversations, transcription takes time, and scaling to thousands of participants proves prohibitively expensive.
Anthropic's solution delegates the interviewing role to AI while human researchers maintain control over research design, question formulation, and analysis. The tool operates across three distinct phases. First, it generates an interview rubric aligned with specific research objectives. Then it conducts adaptive, real-time conversations lasting 10 to 15 minutes each. Finally, it analyzes transcripts to identify patterns and extract representative quotations.
The system already shows signs of participant acceptance. Post-interview surveys found 97.6% of participants rated their satisfaction at 5 or higher on a 7-point scale, and 99.12% said they would recommend the format to others.
What 1,250 Professionals Revealed About Working With AI
The study drew from three distinct populations: a general workforce sample of 1,000 participants, 125 creative professionals, and 125 scientists across more than 50 disciplines.
General Workforce: Productivity Gains Meet Career Anxiety
Among the general workforce sample, 86% reported that AI saves them time and 65% expressed satisfaction with AI's current role in their work. The findings suggest most professionals view AI as a net positive for their day-to-day productivity.
Yet optimism coexists with significant tension. Some 69% of participants mentioned social stigma associated with AI use at work. One fact-checker described staying silent when a colleague expressed hatred for AI, explaining that many people hold strong negative opinions about its use in professional settings. This social pressure appears to create a hidden economy of AI adoption, where workers benefit from the technology privately while maintaining appearances publicly.
Career anxiety also figured prominently. While 41% of interviewees felt secure in their positions and believed human skills remain irreplaceable, 55% expressed worry about AI's long-term impact on their careers. The responses to this anxiety varied: some deliberately limited AI use in areas they considered professionally defining, others pursued additional responsibilities or specialized skills, and a small percentage had already begun planning career transitions to entirely different fields.
Creatives: Embracing Efficiency Despite Peer Judgment
Creative professionals reported even stronger productivity benefits, with 97% saying AI saves them time and 68% indicating it improved their work quality. A photographer described reducing turnaround time from 12 weeks to approximately three, while a web content writer reported more than doubling daily output from 2,000 to over 5,000 words of polished content.
The efficiency gains come with professional complications. Some 70% of creatives mentioned managing peer judgment around AI use. A map artist expressed concern about brand reputation, noting a desire to avoid association with AI stigma in their business.
Economic anxiety ran deeper among creatives than in other groups. A voice actor noted that certain sectors of their profession had essentially disappeared due to AI adoption. A composer worried about platforms using AI to flood markets with cheap alternatives to human-produced music. One artist captured the tension directly, acknowledging that their gains from AI likely meant losses for other creatives who previously received their business.
Perhaps most revealing was the disconnect between stated preferences and actual practice. All 125 creative participants said they wanted to maintain control over their creative outputs. Yet many acknowledged moments where AI drove creative decisions. One artist estimated the split at 60% AI-driven concepts to 40% personal ideas.
Scientists: Desire for Partnership, Barriers to Trust
Scientists presented the most ambivalent relationship with AI. The technology had not yet proven capable of handling core research tasks like hypothesis generation and experimental design. Instead, researchers primarily used AI for peripheral work: literature review, coding assistance, and manuscript writing.
Trust emerged as the dominant barrier, cited in 79% of interviews. An information security researcher articulated the fundamental problem: if verifying every AI output takes as much time as completing the task manually, the efficiency proposition collapses. A chemical engineer raised concerns about sycophancy, noting that AI systems tend to shift answers based on how questions are phrased.
Despite current limitations, 91% of scientists expressed desire for more AI assistance in their research. The aspirational vision went beyond writing help to include critiquing experimental design, accessing scientific databases, and generating novel hypotheses. One medical scientist articulated the goal as finding an AI that could identify relationships and interactions not immediately evident to human researchers.
The Gap Between Perception and Practice
The research surfaced an intriguing discrepancy. When asked to characterize their AI use, 65% of participants described it as augmentative (AI collaborating with users on tasks) rather than automative (AI performing tasks directly). But Anthropic's separate analysis of actual Claude conversations showed a nearly even split: 47% augmentation versus 49% automation.
Several explanations present themselves. Users might refine AI outputs after conversations end, making interactions appear more automative than they functionally are. Self-reported usage could reflect aspirational preferences rather than actual behavior. Social desirability bias might push participants toward describing more collaborative relationships with AI than their usage patterns suggest.
What Comes Next
Anthropic has launched a public pilot inviting Claude.ai users to participate in similar interviews about their vision for AI's role in their lives. The company plans to analyze anonymized insights and publish findings as part of ongoing societal impact research.
The company is also pursuing targeted partnerships. Collaborations with cultural institutions including the LAS Art Foundation, Mori Art Museum, and Tate aim to understand AI's role in creative augmentation. A partnership with the American Federation of Teachers will explore AI education and gather teacher perspectives on the technology's development. Scientists funded through Anthropic's AI for Science grants will participate in interviews about their research needs and expectations.
The limitations of this initial study warrant acknowledgment. Participants were recruited through crowdworker platforms rather than representing a random sample of the workforce. Text-only interviews cannot capture tone, facial expressions, or body language that might affect interpretation. And the sample primarily reflects Western workers, limiting global generalizability.
Still, the dataset offers researchers an unusual resource: candid, lengthy conversations about a transformative technology, conducted at scale and released openly for independent analysis. As AI continues reshaping professional work, understanding how people actually experience that transformation becomes increasingly valuable. Anthropic is betting that asking an AI to conduct those conversations might be the most practical way to gather answers at the scale the question demands.
