Data Entry Keyer
Data entry keying is about as close to a pure text-in, text-out task as any occupation on this site, and the data reflects that.
Based on Microsoft's "Working with AI" study of real Copilot conversations mapped to O*NET tasks (arXiv 2507.07935): Data Entry Keyers scored 0.32 on AI applicability, well above the cross-occupation mean (0.159, stdev 0.098) across all 785 SOC codes studied — banded here as High real-world AI usage relative to other occupations.
Corroborating signal from the Anthropic Economic Index (Claude.ai/API conversations, a different product and userbase than the Microsoft data above): this occupation shows observed exposure of 0.67 — roughly the top 1-2% of the 756 occupations AEI tracks (dataset mean 0.077, 95th percentile 0.385). As with customer service, this is a high-volume signal, not a low one — treat it as directional confirmation of the Microsoft finding.
What a data entry keyer actually does
Data entry keyers take information from paper forms, scanned documents, and other source materials and enter it into computer systems, spreadsheets, and databases. The job is built around accuracy and speed: reading a source document, typing or scanning the relevant fields, checking the result against the original, and correcting any mismatches before moving to the next record.
Keyers compile and sort documents before entry, verify data for completeness, resolve discrepancies by checking alternate sources, and maintain logs of what's been processed. Some specialize in a particular kind of document — insurance claims, medical records, inventory counts — but the core task is the same across settings: transcribe structured or semi-structured information from one format into another, quickly and without error.
It's detail-oriented, repetitive, screen-based work, done at a keyboard for most of a shift, with little variation in the physical setting or the nature of the task from one day to the next.
Why it reads this way
Data entry keying is about as close to a pure text-in, text-out task as any occupation on this site, and the data reflects that. Microsoft's analysis of real Copilot conversations scored this occupation 0. 32 on AI applicability, well above the cross-occupation average of 0. 159, placing it in the High usage band. The Anthropic Economic Index shows an even sharper signal: an observed exposure score of 0.
67, putting it in roughly the top 1-2% of all 756 occupations AEI tracks for measured AI usage — this is not a marginal or speculative finding. The tasks that define the job — reading a document, extracting the relevant fields, and transcribing them accurately into a system — are also exactly the tasks that optical character recognition and language models have gotten dramatically better at handling directly, without a human retyping anything.
BLS projects roughly flat-to-declining employment through 2034 for this occupation, with the smallest number of projected annual openings of any occupation profiled on this site, consistent with software increasingly handling this work end to end.
Skills this role draws on
Source: U.S. Bureau of Labor Statistics OEWS wage data (May 2024), accessed via O*NET OnLine wage report, SOC 43-9021.00
Safer, skill-adjacent careers
These aren't generic "consider retraining" suggestions. Each one shares real skill or task overlap with data entry keyer work, and each one scores meaningfully higher on our structural exposure scale, with the reasoning shown below.
Both roles involve close, rules-based checking of information against a standard, but inspectors verify physical parts and materials on a production floor, not text in a database — a hands-on task that keeps its score at 78 versus 9 for data entry.
Data entry keyers who value precise, procedure-following work can find the same structured, checklist-driven discipline in surgical technology, but applied to physically preparing instruments and assisting in the operating room — score 89 versus 9.
This path trades a keyboard for hands-on personal care — record-keeping and routine-following carry over, but the core work of assisting a real person with daily physical needs is far harder to automate — score 92 versus 9.
Data sources & methodology
Salary data: U.S. Bureau of Labor Statistics OEWS wage data (May 2024), accessed via O*NET OnLine wage report, SOC 43-9021.00.
Task descriptions: Based on O*NET occupational analysis (43-9021.00).
Real-world AI usage band: Microsoft's "Working with AI" study of Bing Copilot conversations mapped to O*NET tasks (arXiv 2507.07935), and corroborating data from the Anthropic Economic Index.
Growth projections: -1% (2024-2034), based on BLS Occupational Outlook Handbook.