Agent
Define instructions, tools and the required outcome.
PRACTICAL GUIDE
An agent performs a bounded task; automation determines when an eligible task runs. Design both around permissions and review.
WHAT TO CONSIDER
Define instructions, tools and the required outcome.
Choose a schedule, condition, manual action or verified event.
Keep consequential external actions behind their own authority.
STEP BY STEP
Start with a small example. Clarify requirements and review the result before using it.
Name the output the process should produce.
Define when a run is appropriate.
Limit context, tools and budget.
Check activation, failure, stopping and repeated runs.
EXAMPLE WORKFLOW
An illustrative example using fictional data. It demonstrates an approach, not a real customer result.
Generate a weekly draft from approved notes.
Schedule triggers a bounded drafting task. A person reviews the output; sending it is a separate action.
QUESTIONS AND ANSWERS
Name the output the process should produce.
Check activation, failure, stopping and repeated runs.
Not every agent supports every trigger. The current automation pilot does not enable arbitrary SQL, shell commands or unrestricted external writes.
EXPLORE FURTHER
START WITH ONE TASK
Guest chat needs no registration. Advanced capabilities depend on plan and configuration.