Customer Support AI Task Author

The work

You will create and review training and evaluation tasks that teach AI agents how to handle enterprise customer support cases. Your work will turn practical support judgment into clear prompts, realistic system states, and objective scoring rules.

• Write customer issue triage prompts based on real product support workflows.
• Define rubrics for routing decisions, resolution paths, and escalation criteria.
• Build scenarios using ticket queues, account records, runbooks, and system states.
• Solve each task to check that the scenario and expected reasoning are realistic.
• Adjust ambiguity, configuration complexity, and runbook depth to set the right challenge.
• Review and correct AI-drafted prompts and scoring rubrics.

What it pays and takes

This is a part-time contractor role for someone with hands-on customer or technical support experience. The role is listed as entry level, but you must already understand support workflows and be able to explain and evaluate decisions precisely.

• Pay: $20-$50 USD per hour.
• Time: 20+ hours per week.
• Location: Australia.
• Language: English.
• Work type: Part-time contract.
• Experience: Support engineer, technical support specialist, or customer success engineer experience.
• Knowledge: Product failure modes, escalation paths, ticketing systems, knowledge bases, and runbooks.
• Skills: Reasoning through ambiguous support cases, explaining routing decisions, and defining verifiable correctness criteria.
• Helpful background: Enterprise or industrial software support, or writing and maintaining runbooks, escalation playbooks, or internal knowledge bases.
• Requirement: Precise written communication for tasks evaluated without interpretation.

How it works

Apply on OpenTrain with your resume, then complete the application on the hiring site.

About AI training work

AI training work uses human-written examples and careful reviews to improve how artificial intelligence handles real tasks. People with practical expertise are paid to create realistic examples and judge whether an AI system reaches a sound result.