A Day 1 CEO Kit field guide
Map the work before you automate it
You do not need to pick an AI tool to get started. First, write down one real process the way it happens now: what kicks it off, what people do, where they make choices, and what “done” looks like. That small map gives you something concrete to improve and check.
A quick check before you map
A process does not need to be exciting to be a good first candidate. It needs to be steady enough that you can tell when something went wrong.
- It happens often enough that you can describe the usual path from memory.
- The trigger, inputs, and result are fairly consistent from one run to the next.
- A mistake is easy to notice before it causes a bigger problem.
- You can pause the process and finish it by hand if needed.
The workflow map worksheet
Copy these prompts into a note or answer them on paper. Describe the current manual process first; do not smooth over the awkward bits because those are often where a tool will need a person's help.
1. Trigger
What event starts the work, and where do you notice it?
Write: This workflow starts when …
2. Inputs
What information or files are needed? Where does each one come from, and which details must be present?
Write: The person doing this needs … from …
3. Repeated steps
Write the steps in the order they happen today. Include the small checks and handoffs people often leave out.
Write: First … Then … After that …
4. Decision points
Where does someone choose between paths? Write the rule if it is clear. If it depends on judgment, name who should decide.
Write: If …, do …; if it is unclear, pause and ask …
5. Output
What should exist when the work is done? Say where it is saved or handed off and what makes it complete.
Write: The finished result is … in …
6. Human review
Who checks the result, what do they check, and what must wait for their approval?
Write: Before …, [person] checks …
7. Failure cases
What if information is missing, duplicated, out of date, or unusual? What should make the process stop instead of guessing?
Write: Stop and hand this back when …
8. Baseline
Note the usual time per run, how many items it handles, and how often you need to correct or chase something.
Write: For a typical run: ___ minutes, ___ items, ___ corrections or follow-ups.
A small example: the weekly internal update
Imagine you gather a few approved team notes into a Friday update for your own review. This is a draft for an internal audience, not a process that publishes or sends itself.
- Trigger and inputs
- The notes for the week are ready. The inputs are those notes, with the owner and current status attached to each item.
- Steps and decisions
- Gather the notes, group related updates, and mark anything blocked. If an item has no owner or its status is unclear, leave a question for the person who owns it instead of guessing.
- Output and review
- Save a short draft in the team document. Before sharing, the owner checks names, status, dates, and whether anything should stay private.
- Failure cases
- A note may be missing, duplicated, or out of date. Pause the draft and check the source rather than filling a gap with a guess.
Keep a simple baseline
Record what the manual version takes
For the next three normal runs, note the minutes from start to finished review, the number of items handled, and any corrections or follow-ups. Later, compare three similar runs using the same measures. Count human review time too; otherwise the comparison leaves out part of the work.
A plain baseline: Typical run: ___ minutes · ___ items · ___ corrections or follow-ups.
Now decide whether a tool belongs here
If the work follows an exact rule every time, a normal rule, template, or reminder may be simpler. If the inputs vary and the task involves sorting language or preparing a draft, an AI-assisted step might help—but keep the review and stop conditions in the map. If the process is still changing every week, keep learning it by hand before automating it.
For more on choosing a low-risk place to begin, read our guide tochoosing your first AI automation.