Data Entry Keyer

What you'll do

This is benchmark authoring, not keying. You design evaluation tasks that put an AI agent in the position of a data entry clerk facing a realistic mess, and then you decide exactly what "done correctly" looks like.

Concretely, that means assembling large, awkward datasets (CSVs, PDFs, spreadsheets and technical documents) and deliberately seeding them with the errors real work throws up: malformed records, missing fields, inconsistent formats, silent truncations. For each task you define the correct final state and list every error case and reconciliation step the agent must handle. Then you write a grading rubric of 35 or more criteria covering both accuracy and how completely the errors were caught.

The standard to aim for is a regulated or audit-sensitive environment, so tasks should feel like a claims queue or a back-office reconciliation, not a textbook exercise. Collaboration is asynchronous and in writing.

Who fits

• Substantial data entry, QA or data validation experience in a regulated domain such as healthcare claims, finance back office or legal operations
• A documented record of high accuracy, ideally with measurable error rates
• A sharp eye for inconsistencies and formatting problems across different file types
• Written English good enough to draft complex tasks and long rubrics clearly
• A disciplined, process-driven way of working
• Comfort working independently and communicating proactively in a remote team

No AI experience is required.

What it pays

$20–35/hour, micro1's published band, but the ad says compensation is output-based: you are paid per task that meets the project specifications, and a minimum number of tasks per week applies. The hourly figure is therefore a guide to effective earnings, not a guaranteed rate.

A task here is not small. Building a seeded dataset plus a 35-item rubric takes real time, especially at first, and a rejected task earns nothing. The ad does not publish the per-task fee, the weekly minimum or the acceptance rate.

Worth knowing

Good:

• Rewards experience that is rarely valued elsewhere in AI training: audit-grade accuracy on routine data work
• Remote and asynchronous
• 50 openings, flagged as high demand
• Fast start: micro1 says roles fill within 48 hours and first tasks begin 24–48 hours after onboarding

Less good:

• Per-task pay with an unstated weekly minimum
• Tasks are complex to build; early ones will pay below the band
• The rubric-writing load (35+ criteria per task) is heavy
• No project length stated
• Contractor engagement: no benefits, your own tax to manage

About this listing

Posted by micro1 on its own jobs portal and read on 24 September 2026. Location is Remote with no country list. The output-based pay, the weekly minimum and the start timeline are from the ad; the per-task fee is not published. See micro1.