A useful custom AI Agent request explains what the AI Agent should do, its required inputs, expected output, and success criteria.
Describe what the AI Agent should do
- The support problem your team wants to solve.
- The ticket categories, fields, policies, or integrations involved.
- What should trigger the AI Agent.
- What the AI Agent should produce, suggest, update, or create.
- Examples where the AI Agent should act and should not act.
Make the request reviewable
- Use real task language your support team already uses.
- Include known exceptions and edge cases.
- Identify any required tags, statuses, priorities, or custom fields.
- Describe how your team will know the AI Agent is successful.
- Note whether the support task should start in Supervised Mode.
What to avoid
- Do not ask for a broad AI Agent that handles several unrelated support tasks.
- Do not leave out the data source the AI Agent needs.
- Do not rely on vague goals such as "make tickets better" without an expected action.
When the request details are ready, request a custom Autopilot AI Agent.