Best practices for AI Employees
Last updated: 7 May 2026
What this does
Use this guide to introduce AI Employees with clear scope, review expectations, and entitlement control.
Current access and functionality
- Verified against the current
impact-web,impact-client, andimpact-marketplacecode on 2026-05-07. - AI Employee creation and management are handled in
impact-web. - Management requires workspace owner or admin access and the AI Advanced entitlement.
- AI Employee capacity is controlled by subscription or entitlement limits.
- Staff may interact with AI Employees through permitted workflows but do not manage setup unless granted admin access.
When to use AI Employees
- Use them for bounded support roles with repeatable context.
- Use them where outputs can be reviewed by a human owner.
- Use them to reduce drafting, summarising, routing, or support workload.
- Avoid using them as the sole owner of ambiguous or sensitive decisions.
Setup practices
- Define the AI Employee’s purpose in one sentence.
- Assign a human owner responsible for review and correction.
- Limit workspace or team scope to where the AI Employee is needed.
- Use clear names so staff can distinguish AI Employees from human members.
- Review capacity and entitlement before adding more AI Employees.
- Revisit configuration after process or subscription changes.
Review practices
- Check early outputs more often after setup.
- Keep humans responsible for approvals, sensitive messages, and exceptions.
- Document what the AI Employee should not handle.
- Remove or adjust AI Employees that no longer have a clear purpose.
Common issues
- Issue: AI Employee creation is blocked. Fix: Check AI Advanced entitlement and available capacity.
- Issue: Staff misunderstand what the AI Employee does. Fix: Rename it and clarify its scope in the workflow.
- Issue: Output quality varies. Fix: Narrow the scope and add human review checkpoints.