Why do we hand off boring paperwork just because it's tedious, but only hand off shopping once we actually trust the AI?

This explores why handing a purchase to an AI agent feels different from handing over routine admin work, and what makes people willing to do each. No study in the collection compares the two head-to-head. Read side by side, though, the notes suggest a clear answer: people hand off admin work because they don't care about it, and they hand off shopping only when they trust it. Those two motives lead to very different designs.

Start with why people delegate at all. On the admin side, the strongest predictor is how pointless a task feels. Workers most want to give AI the tasks they see as meaningless or cut off from real goals, and for those tasks they also accept less human oversight Does perceived task meaninglessness drive worker desire for AI delegation?. In practice, what workers have actually committed to AI workflows clusters in information-heavy jobs and follows what the technology can do Where have workers actually delegated tasks to AI?. Shopping works almost the opposite way. People rarely find their own purchases pointless, and they adopt autonomous shopping assistants because they see a clear benefit, especially when scarcity makes the assistant's speed valuable What drives shoppers to accept autonomous AI assistants?. Admin delegation is about getting rid of something. Shopping delegation is about gaining something.

That difference shows up in how much control people keep. In Accenture's survey of 25,590 people, 74% said they would trust an agent with purchases, but 56% would still name the brand and many loyal customers wouldn't switch Do consumers actually want to delegate shopping decisions to AI?. So shoppers hand over the clicking, not the choosing. A useful framework breaks delegation into eleven task traits, and two of them explain the gap: subjectivity (does the right answer depend on personal taste?) and contextuality (does it depend on what only you know?) What makes delegation work beyond just splitting tasks?. Shopping scores high on both. Filing an expense report mostly doesn't. The same framework calls verifiability foundational, and "did it fill the form correctly" is much easier to check than "did it buy what I would have bought."

The less obvious lesson is about trust. You might expect people to pull back as the stakes go up. In a study of students using a general-purpose agent, they didn't. Trust dropped sharply on tasks that were irreversible and visible to other people, like sending an email, even when the output was fine. High-stakes tasks that could be corrected showed no such drop What makes people distrust AI agents they delegate to?. A purchase is both: money leaves your account and a seller sees the order. That puts shopping in the same trust category as sending messages on your behalf, not alongside quietly tidying a spreadsheet. It suggests that approval steps before checkout matter more than raw capability.

Two final points for anyone building or using these agents. First, remembering your preferences is a skill of its own. On phone-agent benchmarks, finishing a task, handling private data properly, and reusing saved preferences are separate abilities, and no model is best at all three Do phone agents succeed at all three critical tasks equally?. A shopping agent that completes orders but forgets your sizes is failing at the part that matters most. Second, temper expectations about saved time. Chat-based delegation cut clicks and scrolling but didn't make tasks finish faster Does chat delegation actually save time on task completion?. Whether the task is admin or shopping, the main gain may be less effort, not less time.