Map an Agent-Worthy Workflow

What you will build

A short, honest opportunity map. Not a strategy deck. A one page scoring of a few real processes from your own work, ending in one chosen use case you can defend.

Why this lab matters

Most agent projects fail before a line of code is written, because the wrong work was chosen. Picking work with real value, clear boundaries and a checkable outcome is the single highest leverage decision you make.

Expected output

• Three to five candidate processes listed.
• Each scored on value, boundary clarity, input quality and how easily the outcome can be verified.
• One selected first use case, with two or three sentences on why it wins.

Prerequisites

• You have read Chapter 1 and Chapter 2.
• You can name a few processes you or your team repeat often.

Steps

1. List three to five processes you do more than once a week.
2. For each, write the goal in one sentence.
3. Score each from one to five on value if automated.
4. Score each on how clear the boundaries are. Fuzzy scope is a red flag.
5. Score each on input quality. Messy or missing inputs raise the cost.
6. Score each on how easily you could check the result was correct.
7. Add the scores. The highest total is your candidate, not your final answer.
8. Sanity check the top candidate against one question. If the agent got it wrong, would you notice before it caused harm? If not, either add a verification step or pick another.

Verification checklist

• Every candidate has a one sentence goal.
• Every candidate has all four scores.
• The chosen use case has a written reason.
• You can state how you would check the agent’s output is correct.

Common failure modes

• Choosing the most exciting task instead of the most checkable one.
• Picking work with no clear “done” state.
• Ignoring how messy the real inputs are.

Extension ideas

• Run the same scoring with a colleague and compare where you disagree.
• Add a rough cost per run estimate to each candidate.

Related chapter

Chapter 2, Pick Work Worth Automating.