Workers often want to be equal partners with AI, not replaced by it — is that what's actually getting built?

This asks whether the kind of human-AI collaboration workers say they want matches what AI is actually being built and funded to do. Mostly it doesn't. The clearest evidence comes from a survey of 1,500 workers across 844 tasks. In 45% of occupations, workers' most common wish was an equal partnership with AI: not handing work over, and not keeping AI at arm's length. Yet 41% of startup investment is aimed at areas that don't match those preferences What collaboration level do workers actually want with AI?. What gets built follows what's technically possible, not what workers ask for. Data on where people have actually handed tasks to AI shows the same thing. Delegation clusters in information-heavy work and follows what AI can do, not how widely people use chatbots or which tasks are routine Where have workers actually delegated tasks to AI?.

Workers aren't against handing things off, though. They want to hand off particular things. Tasks that feel pointless or cut off from what the organization is really trying to do strongly predict who wants to give a task to AI, and how little human oversight they think it needs. That pattern held across 22 occupations Does perceived task meaninglessness drive worker desire for AI delegation?. So the line workers draw is about meaning, not difficulty. That's a different line from the one automation draws, which is mostly about what the technology can handle. A task can be easy to automate and still feel like the core of someone's job, or hard to automate and feel like pure busywork.

The gap matters beyond whether people like their jobs. Brynjolfsson argues that AI built to replace workers rather than strengthen them moves bargaining power to whoever controls the technology. That risks a stable trap where workers lose the leverage they would need to push back Does automating labor concentrate power more than augmenting it?. Shneiderman's framework questions the assumption behind that choice. He treats automation and human control as two separate dials rather than a trade-off, so a system can be highly automated and still under strong human control Can automation and human control increase together?. The partnership workers ask for isn't a technical compromise. It's a design option that mostly isn't being funded.

There is also evidence that partnership simply works better in places. When human agents on Taobao stepped in after an AI failure, they fixed technical problems but couldn't undo the damage once customers were already frustrated. Only early handoffs worked When human workers take over from AI, does it actually fix the problem?. Productivity gains from AI show up when people apply skills they already have, and they disappear when people use AI to learn something new When does AI actually boost worker productivity?. For that reason, replacing the worker can quietly remove the expertise that made the AI useful in the first place.

One social detail cuts against all of this. Even workers who want partnership expect to be judged for it. Across four experiments with 4,439 people, AI users expected colleagues to see them as less competent and less diligent, so they hid their AI use Do people fear judgment when they use AI at work?. Part of the gap between what workers want and what gets built may come from workers not saying openly how they use AI at all. The partnerships they prefer may already be happening, unseen by the managers and investors deciding what to build next.