# What to automate first: a worksheet

Pick one job that keeps coming back. Look at three to five recent examples, ideally with the people who did the work. Leave gaps visible instead of guessing.

## 1. Start and finish
What starts the job, and what should be true when it's done?
- What starts it:
- What "done" looks like:
- Recent examples:

## 2. What repeats
What stays the same across the examples, and what changes? Where does the work wait or come back?
- Same every time:
- Changes between examples:
- Where it waits or comes back:

## 3. What kind of work
For each part of the job: does it follow a rule, need an expert, or need an experiment?
- Follows a rule (clear):
- Needs an expert (complicated):
- Needs an experiment (complex):
- Who decides:

## 4. The agent's part
What could an agent do on its own? What should it only prepare? Who handles the exceptions?
- On its own:
- Prepares for a person:
- Waits for approval (anything that sends, pays, deletes, or publishes):
- Who handles exceptions:

## 5. The check
How will you know it worked, and how will you notice if it causes harm?
- What you will measure:
- How you will notice harm:
- When you will review it:

By Abdullah Faruk Gönüllü. The article that goes with it: https://afaruk.dev/blog/cynefin-ai-workflow-automation
Inspired by the Cynefin framework; not an official Cynefin method.
