YonSuite AI Assistant Tasks: Inputs, Confirmation and Human Escalation
This tutorial shows how to split a complex request into bounded tasks, check input sources and authority at each stage, require human confirmation before important actions, and review facts and exceptions in the output.
Learning objective
Help teams design an AI-assistant workflow with traceable inputs, authority, results and responsibility.
03
Steps
01Split the request into gathering, analysis, recommendation and execution, listing required inputs for each stage.
02Separate read-only advice from execution and require an authorised person to confirm before any state change.
03Check facts, dates, entities and sources; stop and escalate when inputs conflict or a tool fails.
- 01
Split the request into gathering, analysis, recommendation and execution, listing required inputs for each stage.
- 02
Separate read-only advice from execution and require an authorised person to confirm before any state change.
- 03
Check facts, dates, entities and sources; stop and escalate when inputs conflict or a tool fails.
04
Implementation notes
- Keep irreversible actions outside the default automated path.
05
References
- YonSuite深度整合GPT后，或将孕育出超凡“数智员工” ↗用友YonSuite · Published: 2023-02-20 · Accessed: 2026-08-14