Two jobs equally exposed to AI can split on pay, depending on whether AI takes a slice of the work or most of it.

This explores why AI hurts pay in some exposed jobs and not in others. One caution first: the collection has no study that tracks earnings falling in one named occupation and staying flat in another. What it does have is a set of mechanisms that explain the split, and they point somewhere unexpected. How much AI touches a job matters less than how that exposure is spread across the job's tasks. Firm-level data from 2010 to 2023 show that when AI hits only a few of a job's tasks, workers move their time to the tasks AI can't do, and the net loss of jobs is small. When exposure is spread across most of the work, there is nowhere to move, and demand for labor falls Does concentrated AI exposure enable workers to adapt and reallocate?. So two occupations with the same 'exposure score' can end up in very different places, depending on whether AI takes a slice of the job or the whole thing.

Timing is the second answer, and it's why so many studies find nothing yet. Danish administrative records show employers adopting chatbots widely and workers taking on new AI-related tasks within two years of ChatGPT's launch. Earnings and hours stayed within 2% of where they were Does AI chatbot adoption change worker pay and hours?. Jobs get reorganized first, and pay may follow later. An occupation showing 'no effect' may simply be early in that sequence.

Third, the damage may show up in who gets hired, not in what current workers are paid. U.S. payroll data find no economy-wide job losses. But young workers in AI-exposed occupations face hiring rates 19% lower than their peers, while experienced workers show no such gap Is generative AI displacing workers at economy-wide scale?. Average earnings in an occupation can look stable while the entry door quietly narrows. Exposure also lands differently depending on who holds the jobs. In male-dominated fields it concentrates among high-paid, high-skilled workers. In female-dominated fields it spreads evenly down to lower-paid workers, who have fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?.

Employers matter as much as occupations. Firms with heavy AI exposure replace freelance workers with AI faster and more cheaply than other firms, which suggests that a firm's in-house AI capability drives the pace, not the technology alone Do firms substitute labor for AI at different rates?. Real delegation of work to AI is also concentrated in information-heavy occupations. It follows what the technology can actually do, not older predictions about automating routine jobs, and the wage pattern reverses at advanced-degree levels Where have workers actually delegated tasks to AI?. The productivity side has a matching split: AI boosts workers applying skills they already have, but the gains disappear when they use it to learn something new When does AI actually boost worker productivity?.

The bigger point is that average wages can rise while particular groups lose. Anthropic's scenario model projects average wages going up while knowledge-worker wages stall or fall, because gains flow to whoever owns the capital and workers can't easily switch occupations Does AI growth inevitably shift wealth away from workers?. That is a projection, not a measurement. Still, it reframes the question. Whether earnings fall may depend less on what AI can do in a job than on whether workers can move to other tasks, other employers, or other occupations before the reorganization reaches their pay.