Guides

Practical guides for managed data and operations teams

Step-by-step guides and playbooks, written from what we actually see running pilots and managed teams, not generic advice.

• How to write annotation guidelines people can follow

  A step-by-step method for writing labeling instructions: structure, decision rules, examples, edge cases, agreement checks and version control.

  Published 2026-10-02

• Product data enrichment checklist

  A field-by-field checklist for enriching product data: identifiers, titles, categories, attributes, variants, images and QA, with a worked example.

  Published 2026-10-02

• Amazon Mechanical Turk shutdown: questions and answers

  Amazon Mechanical Turk closed on 2026-09-30. The dates that still matter, what happens to HITs and balances, and what changed in Ground Truth and A2I.

  Published 2026-10-02

• Managed team vs freelancers: how to choose for repeatable data work

  Compare a managed team, marketplace freelancers and in-house hires for repeatable data work: who trains, who checks quality, and the full cost.

  Published 2026-10-02

• Human review of LLM outputs: a practical playbook

  How to run human review of LLM and agent outputs: rubrics, gold sets, reviewer calibration, agreement, QA sampling and escalation, with a worked example.

  Published 2026-10-02

• Document extraction QA: how to check OCR and AI output

  A practical QA method for OCR and AI document extraction: field specs, validation rules, targeted human review, sampling and an error taxonomy.

  Published 2026-10-02

• How to set up an Amazon A2I private workforce

  Create a private workforce and work team, rebuild your A2I human review workflow, and test it before full volume. Checked against AWS docs on 2026-10-02.

  Published 2026-10-02

• MTurk closed: keep Ground Truth and A2I running

  What changed in SageMaker Ground Truth and Amazon A2I, every workforce option, and five steps to move off the Amazon Mechanical Turk workforce.

  Published 2026-09-30