What the work involves

You will spend most of your time producing and critiquing real deliverables rather than writing abstract guidance. A typical task gives you a messy input — a raw support ticket export, a usage telemetry dump, a set of account notes — and asks you to build the artifact a CS leader would actually build from it: a health-score model in a spreadsheet with defensible weightings, a renewal risk scorecard, a QBR deck a customer executive would sit through without wincing, an escalation runbook with clear ownership and time-boxed triggers.

The other half is evaluation. You read model-generated versions of the same artifacts and judge them: is the churn signal real or a plausible-sounding invention? Does the health score double-count login frequency and feature adoption? Does the QBR narrative survive a CFO asking where the number came from? You write the reasoning behind the score, and that written rationale is often weighted more heavily than the score itself.

What the platform screens for

  • Ownership, not adjacency. Screens probe whether you built and defended a health-score model or renewal forecast, or whether you consumed someone else's. Expect follow-ups on inputs, weights, and what you changed after the model missed.
  • Artifact craft. Spreadsheet structure, formula hygiene, slide logic, and the ability to explain why a chart choice is wrong.
  • Evaluation judgment. Whether you can distinguish a confidently wrong output from a merely unpolished one, and rank failures by consequence.
  • Written clarity under time pressure, since rationales are the deliverable.

Logistics

Fully remote and asynchronous. Contributors set their own hours; 5–20 hours per week is the stated band, with more available to people whose work scores well. Tasks are claimed from a queue, so there is no fixed schedule, but sustained low throughput tends to reduce access. The interview process includes an AI voice screen — a spoken conversation where follow-up questions adapt to your answers, so specifics matter more than a rehearsed narrative. Pay is observed at $70/hour; rates on these programs vary by task type and are not guaranteed.