AI and Automation

Responsible workflow

1. Define the educational or operational problem.
2. Classify the data before selecting a tool.
3. Keep a human accountable for decisions and approvals.
4. Test with synthetic data before connecting real services.
5. Record prompts, models, versions, and evaluation results.
6. Monitor failures and provide a manual fallback.

Platform roles

• marimo supports reproducible computational explanations and prototypes.
• n8n coordinates approved low-risk workflows and notifications.
• GitHub records source, reviews, tests, and deployment history.
• NocoBase holds low-risk operational records and committee workflows.

Automation must not send confidential university or health data to an external AI service without an approved data-processing arrangement.