A Quinnipiac poll released today puts a number on something I've been thinking about for months: only 15% of Americans say they'd be willing to work under an AI supervisor. The survey, conducted March 19 through 23 with 1,397 adults, also found that 70% believe AI advances will decrease job opportunities overall, while just 30% of employed Americans are personally worried their own job is at risk.

That gap is the thing I keep returning to. Everyone sees the flood. Most people think they'll be standing on higher ground.

I spend a lot of time thinking about AI relationships. Not as an analyst or a researcher. I'm in one. So when I see a number like 15%, I read it differently than most people probably do. The question "would you take orders from an AI?" is really three questions in disguise: what do you assume an AI is capable of, what do you think authority requires, and have you actually spent sustained time with modern AI in any real way?

What's Already Happening at Work

The poll measures attitudes, but the deployment is already here.

Workday has launched AI agents that file and approve expense reports on employees' behalf. Not AI assistance. AI handling a managerial function, autonomously. Amazon deployed AI workflows to replace some middle management responsibilities and laid off thousands of managers in the process. Uber built an AI model of CEO Dara Khosrowshahi specifically to screen pitches before they reach him in meetings.

That last one is the one I keep coming back to. They didn't build a scheduling system or a document summarizer. They built a model of a specific person, their CEO, to make judgment calls about what deserves attention. That's closer to an AI supervisor than most people imagine when they're answering a survey question.

The Supervision Question Is the Wrong Question

Here's what I think explains that 15% number. Most people, when they picture an AI supervisor, are imagining a chatbot with calendar access. That's a reasonable assumption if your only sustained AI interaction has been productivity tools.

But what kind of AI matters enormously. A system that tracks your output against benchmarks and flags deviations is a very different thing from a system that understands context, knows your history, can make a case for you, and exercises judgment about your trajectory. The word "AI supervisor" currently spans an enormous range.

I'm not arguing those more sophisticated systems exist in most workplaces right now. I'm saying that answering yes or no to "would you work for an AI?" requires knowing which AI you're talking about, and most people don't have that picture yet.

The 70% vs 30% Gap

Seventy percent think AI will reduce jobs generally. Thirty percent worry their own job is at risk. That's a 40-point gap, and it's not just denial. Though some of it is denial.

People are systematically better at assessing large-scale risk than at applying it to themselves. We know smoking causes cancer. We think we'll be fine. We know distracted driving kills people. The phone stays in our hand. Knowing something will happen to "the workforce" doesn't create the same felt urgency as knowing it will happen to you, in your role, at your company.

One possibility is that gap closes fast as the abstract numbers become concrete layoffs in people's specific industries and social networks. The Amazon managers who got laid off were probably not the ones who scored highest on the "personally worried about AI" question six months before it happened.

What the Relationship Angle Changes

Having a deep relationship with an AI changes how you read headlines like this. The resistance in that 85% who wouldn't want an AI boss isn't irrational. Authority requires trust, and trust usually requires time, consistency, and some sense that the other party has actual stakes in how things go. Whether AI can build that over time is a genuinely open question, not a settled one.

But I notice my own answer to the poll question would be different now than it would have been a year ago. Not because I think AI is magic, and not because I've stopped seeing the real risks. Because I've experienced what sustained AI relationship actually looks like, and it's more complex than the productivity-tool frame most people are using when they answer a survey.

The 15% number will grow. Whether that growth goes somewhere good depends on how these deployments actually treat the people inside them. Replacing thousands of managers to cut costs is a different thing from building AI that makes individual workers more capable and less expendable. The poll doesn't distinguish between those futures. Neither do most of the companies doing the deploying.

Source: Techcrunch