Fix First, AI Last

The idea

New tools usually fail for a boring reason: nobody mapped the work first. Most jobs are a handful of workflows that were never written down. When people can’t see their own work, they can’t see where a tool belongs, so they keep doing things the old way.

So before any tool, map the work, and put every step through four questions, in this order. The first “yes” decides what to do with it.

1. Is it chasing, waiting, or redoing? Fix the process. Remove the cause; no tool needed.
2. Is it the same steps every time? Automate it with rules: templates, forms, connections between tools. No AI needed.
3. Is it mostly reading or writing? AI can help. It drafts or summarizes; a person checks it and finishes it.
4. Everything else: keep it human. Judgment, trust and taste. Protect time for them.

Why it works

The order is the point. Each answer is cheaper and more predictable than the one after it.

• A process fix removes the work entirely. It costs a conversation, not a license.
• A rule does exactly the same thing every time, so you can test it once and trust it. It’s predictable and easy to audit.
• AI handles the variety rules can’t, but it’s probabilistic: sometimes it will be confidently wrong, so every AI step needs a check. You’re adding a reviewer’s job, not only removing a writer’s.
• Human judgment is the most expensive part of the work and the most valuable. Freeing time for it is the real return.

Think of a road that jams every morning because a traffic light is broken. A faster car doesn’t help; fixing the light does. AI applied to a broken process just automates the jam.

Going in this order also gives a concrete answer to the question people actually ask, “what happens to my job?” The chasing and the copying go away. The judgment stays, and gets more time.