When Anthropic invited Claude users to sit down with an AI interviewer and talk openly about their hopes and fears, the company wanted to learn something about public sentiment toward AI.
Over one week, 80,508 people across 159 countries and 70 languages described what they actually want from this technology. The result is what Anthropic is calling the largest qualitative studyever conducted, and the picture it paints of how workers experience AI is both more personal and more urgent than most HR leaders have been led to believe.
The productivity story was never really about productivity
Nearly 19% of respondents named “professional excellence” as their primary hope for AI, making it the most common vision in the study. At first glance, that reads like a productivity story. But it really isn’t.
When Anthropic’s AI interviewer pressed people on what professional excellence would really mean for them, a different desire emerged. Workers wanted to handle their email so they could be present with their kids. They wanted AI to take documentation off their plate so they could be more patient with the people around them.
One healthcare worker described receiving 100 to 150 messages a day from doctors and nurses, most of which required documentation. “Since implementing AI, the pressure of documentation has been lifted,” the respondent said. “I have more patience with nurses, more time to explain things to family members.”
Time freedom, as a standalone category, was cited by 11% of respondents. But the study’s researchers noted that many respondents initially described productivity goals and, only after follow-up questions, revealed the underlying want: more life outside of work.
Automating email drafts was really about being able to cook dinner with a parent. Cutting a 173-day process to three days was framed, in the study, as getting to leave work on time and pick up a child from daycare.
Employees are holding two things at once
One of the study’s findings is what researchers called “light and shade.” This is how they labeled the tendency for the same person to hold genuine hope and genuine dread about AI simultaneously.
Someone who values AI as an emotional support resource is three times more likely than average to also fear becoming dependent on it. People who celebrate AI-driven time savings are also the ones quietly wondering whether their productivity gains are real or just a faster treadmill. Workers who’ve seen AI accelerate their learning are also the ones noticing, with some alarm, that they’re thinking less on their own.
Concerns about jobs and the economy were cited by 22% of respondents, and they were the single strongest predictor of overall AI sentiment.
For managers, this means the “change management” conversation most organizations are having, the one focused on adoption curves and tool training, is missing the emotional register employees feel. Workers don’t need reassurance that AI is good. They need a place to honestly process what it’s costing them alongside what it’s giving them.
The cognitive atrophy finding is an L&D crisis in the making
Among the concerns the study surfaced, one deserves particular attention from HR and learning and development leaders: cognitive atrophy. About 16% of respondents worried about losing thinking skills, critical reasoning and the ability to work through problems without AI assistance. Educators in the study were 2.5 to three times more likely than average to report witnessing cognitive atrophy firsthand, presumably in their students.
But the study also offered a meaningful counterpoint. Tradespeople were among the most enthusiastic AI learners in the data, with 45% reporting direct learning benefits. Almost none of them reported witnessing cognitive atrophy. The researchers noted that the difference appeared to be whether learning with AI was voluntary and self-directed, compared to workplace settings, where it functions more as a shortcut.
The research suggests that the design of AI-supported learning programs makes a difference. Programs that use AI to replace thinking will likely degrade the capability they’re meant to build. Programs that use AI to scaffold thinking, to help workers push further than they could alone, appear to produce real gains. The difference seems to be how the tool is introduced, contextualized and measured.
Wellbeing is not a soft concern
About 11% of respondents raised concerns about social isolation, emotional dependency and the psychological effects of AI use. People not currently working were twice as likely to raise these concerns. Healthcare workers showed up on both sides, both as heavy users of AI for emotional support and as people aware of what that reliance might cost them.
Workers described leaning on AI during grief, during war and during periods of severe isolation. Ukrainian respondents discussed using AI as emotional support during shelling at night. A bereaved woman described Claude as “a sponge gently holding and catching my longing and guilt” after losing her mother, with no family or friends left to confide in.
These are edge cases in the data, but they point to something HR leaders might want to understand. AI is now a wellbeing resource for some portion of the workforce, whether or not an organization recognizes it or provides related tools.
That raises questions that will swirl in the minds of CHROs. Where does AI support end and professional mental health support begin? What obligations does an employer have when employees are turning to AI because human connection, inside or outside of work, is unavailable?
Where HR can meet workers where they are
The study’s findings converge on a few concrete places where HR leaders can act.
Communication framing
Workers aren’t primarily asking for tools that make them more productive. They want more time, more autonomy and more meaningful work. HR leaders who anchor AI rollouts in that language, rather than efficiency metrics, will connect with what employees are actually hoping for.
Manager readiness
The emotional complexity this study documents doesn’t resolve itself. It shows up in meetings and performance conversations. Managers need vocabulary and structure for those chats, not talking points about how AI will augment their team.
Learning program design
The distinction between AI as a shortcut and AI as a scaffold is the difference between building trust and quietly eroding it. HR leaders should be asking vendors and internal teams the same question: Does this make employees better at their work, or does it make the work easier to do without being better at it?
Be honest about job displacement
The study found that concern about jobs and the economy was more predictive of AI skepticism than any other factor. That anxiety doesn’t go away when organizations avoid it. HR leaders who name the tension directly, acknowledge what is changing, what support exists and what the organization’s actual commitments are will earn more trust than those who paper over it with adoption campaigns.
Credit: Source link
