Of the 22 tasks O*NET lists for payroll and timekeeping clerks, 4 could be done in half the time with a chat window today. 4 cannot be sped up at all. Here is which.
Exposure means published research judged an AI could halve the time a task takes, at the same quality. It does not mean the task stops needing a person. How to read this
Almost two thirds of this job sits in the middle band, waiting on software rather than on capability. That is the highest share of any People occupation, and it fits: payroll is rule-based, high volume and repetitive, which is precisely the shape of work that purpose-built tools handle well.
Much of that software already exists and has for years. The gap between 18 per cent reachable today and 64 per cent reachable with tooling is largely a question of what your organisation has bought and configured, not what is possible.
The work that is protected here is checking. Verifying hours, reviewing timesheets, catching what does not reconcile, and knowing which anomaly is a data-entry slip and which is somebody being underpaid. That is judgement, and payroll rarely gets described that way.
These are O*NET's task statements for this occupation, exactly as written, sorted into the three bands by the ratings behind this report. Nothing has been reworded.
Nobody surveyed an employer. Tasks are counted equally, so a percentage here is a share of the list rather than a share of your week. And the ratings were made in 2023, which makes every figure a floor rather than a ceiling.
It must never be used to select anyone for redundancy. The full limits are here.
This is the standard task list for the title. The useful next question is which of the tasks you actually do should go near AI, and which should not.
Exposure ratings from Eloundou, Manning, Mishkin and Rock, GPTs are GPTs: Labor market impact potential of LLMs, Science 384, 1306–1308, 2024, used under the MIT licence. This page includes information from the O*NET Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. People Team AI has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications. Employment figures from the US Bureau of Labor Statistics.