When a colleague hands you AI-generated work, does sorting it out eat into your own working day, and by how much?

This explores the receiving end of AI at work: what happens to your time when a colleague hands you AI-generated work, rather than when you use AI yourself. The corpus suggests the answer is often yes, and that the cost is bigger and harder to see than most productivity numbers capture. A September 2025 survey of 1,004 U.S. desk workers found that people estimate 15.4% of the work they receive is AI-generated but unhelpful. They also report spending an average of 1 hour 51 minutes dealing with each instance, which is longer than the sender would have needed to just do the work themselves How much work that employees receive is actually unhelpful AI content?. In other words, the time the sender saved doesn't disappear. It moves downstream to the recipient.

This matches a broader pattern even when you use AI yourself. Studies find that AI often doesn't cut total task time. Instead it shifts time away from doing the work and toward writing prompts and figuring out what came back Does AI really save time, or just change how we spend it?. Receiving someone else's AI output is the same evaluation burden with one extra handicap: you didn't write the prompt, so you have less context for judging whether the output is any good. The reviewing, decoding and redoing still happens, just on someone else's clock.

Why would senders keep passing along weak output? Two findings help explain it. The first is the 'LLM Fallacy': when AI output is fluent and seamless, people tend to treat it as evidence of their own skill and lose track of where their contribution ended and the machine's began Do AI-assisted outputs fool users about their own skills?. This is a self-perception error, separate from whether the output is accurate How does AI-assisted work reshape how people see their own abilities?. A sender can sincerely believe they handed over finished work. The second is that people who use AI expect to be judged as less competent and less diligent, so they're less willing to tell managers and colleagues they used it Do people fear judgment when they use AI at work?. That means recipients often don't know they're looking at AI output, which makes it slower to spot and costlier to fix.

There's also a quieter structural shift. Heavy generative AI users increased their document-production activity by 21.2% but their communication activity by only 7.1%, a tilt toward solo output and away from coordinating with others Does generative AI shift knowledge workers away from communication?. More documents with less conversation around them is exactly the setup in which badly framed AI work lands on someone's desk without the context they'd need to use it. Separately, AI's measured productivity gains show up mainly when people apply skills they already have When does AI actually boost worker productivity?. A sender working outside their expertise is therefore more likely to produce output that looks done but isn't.

The takeaway you may not have expected: AI productivity can be a local illusion. Measured at the sender, it looks like time saved. Measured across the team, some of that saving is a transfer of work to whoever has to make sense of the result. The corpus is thin on direct measurement of this effect. The main evidence is one self-report survey, and no study here tracks recipients' time objectively. The size of the cost should therefore be treated as an estimate, though its direction is consistent across the related findings.