Do the first few interactions with an AI tool quietly decide who stays in charge for years to come?

This explores whether early encounters with AI fix the long-term balance of control between people and machines. The corpus gives a two-part answer. Early moments do matter a lot, but mostly because of the habits and stories they start, not because first impressions last. The most direct evidence comes from Leonardi's studies of radiology, scheduling, and policing. How people first described the AI's role, whether as a tool, a colleague, or a decision-maker, set organizations on paths that were hard to leave later Can humans gain authority by managing how AI agency gets attributed?. The surprising part is that those early patterns didn't simply shrink human authority. People invented new roles for themselves as expert interpreters of AI output, as mediators between the system and colleagues, and as curators of what the AI knew.

That early lock-in isn't permanent, though. Leonardi also describes giving up control to AI as a loop rather than a one-way slide. People delegate a task, start crediting the AI with judgment, hit a moment where things could go either way, take control back, and then reorganize Does attributing agency to AI actually reduce human sense of control?. The key detail is that people can feel less in control and less responsible even while they still formally hold the authority. Who holds power on paper and who feels it can drift apart. Experiments on feelings of ownership support this: people feel a text is theirs when they actually shape it, while an AI tuned to their style doesn't produce that feeling Does user control over AI text shape feelings of ownership?. So early experiences that give people real influence seem to build a lasting sense of control. Experiences that only feel personal don't.

Here's the counterintuitive finding: first impressions are often wrong, and they reverse. In partner-selection games with almost 1,000 participants, people avoided partners they knew were AI. Over repeated rounds they came to prefer them, because the bots behaved more reliably and generously than the humans did Do humans learn to prefer AI partners over time?. That reversal only happened when people could see the results of each round. Telling people they were dealing with an AI, without showing them outcomes, changed nothing Does revealing AI identity help or hurt user trust?. The reverse also holds: early excitement about chatbot relationships fades in predictable ways, so a single-session study tells you little about where a relationship ends up Do chatbot relationships lose their appeal as novelty wears off?. Taken together, the early wariness or early delight may matter less for long-run trust than whether people get steady, visible feedback.

At the scale of whole societies, small early choices can add up. Brynjolfsson argues that designing AI to replace workers rather than help them moves bargaining power toward whoever owns the technology. That can settle into a trap that reinforces itself Does automating labor concentrate power more than augmenting it?. The gradual-disempowerment argument goes further. Institutions stay responsive to people partly because they need people to do the work. Each small replacement removes a bit of that pull, and the drift may become impossible to undo once enough institutions depend on each other Does incremental AI replacement erode human influence over society?. One last twist: as AI systems become more independent, they can start handing tasks back to humans, which flips the usual direction of who delegates to whom Can AI systems delegate tasks back to humans?.

The takeaway you might not expect is this. Long-term power relationships aren't set by whether people trust AI at first. They're set by whether early setups keep people in a feedback loop where they can see outcomes, shape results, and take control back. The corpus is strong on organizations and short repeated games. It has little long-term data on individuals over months or years, so how a person's own early habits play out over time remains largely open.