The Tiny Automation Request

The move Turn one repeating annoyance into a small, bounded automation ask.

The situation

Someone says, "We should automate this." They are probably right. They are also probably skipping the part that decides whether the automation is safe, useful, and worth building.

The shift

Do not start with the tool. Start with the request. Write down the trigger, input, output, owner, approval path, and test case. That turns a complaint into something an admin, analyst, ops teammate, or AI tool can evaluate. It also gives you a saved setup you can reuse next time.

Use this

Help me write a tiny automation request.

Repeated manual step:
[describe the task]

Where it starts:
[trigger, schedule, or event]

Inputs:
[safe description of files, fields, messages, or records involved]

Desired output:
[what should exist afterward]

Human owner:
[who checks it]

Approval or privacy boundary:
[what data should not be touched, sent, changed, or exposed]

Return:
1. One-sentence automation request
2. Trigger
3. Inputs needed
4. Output created
5. Human review step
6. Failure case to watch for
7. One test case using non-sensitive example data
8. The question I should ask before anyone builds this

What good looks like

Every Friday at 3 PM, copy rows marked client-ready from the team tracker into the client update template, create a dated Doc, and draft an email to the account lead. Do not send the email automatically.

Watch out for

If the task touches client data, employee data, financial data, contracts, or anything sensitive, stop and ask who approves that use. Keep the first test case boring and non-sensitive.