Customer Support AI Task Author

The Work

You will create and calibrate tasks that train and evaluate AI agents on enterprise customer support work. Each task should show whether an agent can understand a customer issue, inspect account settings, use the right runbook, and choose a suitable resolution or escalation.

• Write realistic customer issue triage and support decision tasks.
• Build scenarios using ticket queues, account records, runbooks, and system states.
• Define clear routing decisions, resolution paths, and escalation criteria that can be checked objectively.
• Solve each task yourself to check its realism, answer clarity, and scoring rubric.
• Adjust ticket ambiguity, configuration complexity, and runbook detail to set the right difficulty.
• Review and correct AI-written task prompts and scoring rubrics.

What It Pays and Takes

This is remote, flexible, hourly contract work for freelance contributors. The role focuses on creating and evaluating support-task content in English.

• Pay: $20 to $50 per hour.
• Time: 20 or more hours per week.
• Location: You must be based in the Netherlands.
• Language: English.
• Work type: Freelance contractor and part-time.
• Experience: The listing is marked entry level, but you need hands-on experience as a support engineer, technical support specialist, or customer success engineer.
• Knowledge: Understand product failure modes, escalation paths, ticketing systems, knowledge bases, and runbooks.
• Skills: Reason through ambiguous support cases, explain routing decisions, and write precise criteria that can be applied consistently.
• Helpful background: Experience with enterprise or industrial software, runbooks, escalation playbooks, or internal knowledge bases.

How It Works

You apply on OpenTrain with your resume and then complete the application on the hiring site.

About AI Training Work

AI training work uses human-written examples, ratings, and reviews to help AI systems produce more useful results. People with practical support experience are paid to turn their judgment into clear tasks and evaluation rules that models can learn from.