What the work involves

You'll spend most of your time reading model output about CRM scenarios and deciding whether it would survive contact with a real Salesforce or HubSpot org. That means catching automation logic that would fire twice, workflow advice that ignores record ownership or field-level security, pipeline-stage recommendations that don't map to how a real sales team actually forecasts, and data-hygiene guidance that would create duplicates rather than prevent them. Alongside rating and ranking responses against a written rubric, you'll annotate CRM records and interaction data, contribute representative scenarios drawn from your own operational experience, and document current sales-ops practice — including the pain points models tend to gloss over.

A second, quieter part of the job is knowing when the rubric doesn't cover the case in front of you. Tasks land that the guidelines never anticipated: ambiguous prompts, output that's technically correct but operationally reckless, two responses that are wrong in different directions. Flagging those clearly and in writing is as valuable to the project as the ratings themselves.

What the platform screens for

  • Real CRM administration depth, not CRM usage. Expect follow-ups on objects, automation tooling, validation rules, deduplication, migrations, and why a particular design decision was made.
  • Prior paid human data work for AI. micro1 states this as a firm requirement — annotation, labeling, RLHF, AI response or model evaluation, or rubric-based grading, performed as paid work. Using ChatGPT at your day job does not count, and the screen probes for the distinction.
  • Rubric discipline. Whether you can apply someone else's criteria consistently, including when you personally disagree with them.
  • Written English at B2 or above, since feedback to model teams is the deliverable.

Logistics

Remote, contractor, asynchronous. Work is drawn from a task queue; throughput expectations vary by project phase and there is generally no fixed schedule, though the listing asks for availability to begin promptly. Observed pay for this listing spans $28–92/hr — a wide band that typically tracks platform depth, engineering-adjacent CRM experience, and prior evaluation history rather than a single posted rate. Nothing here is guaranteed; rates are set per engagement.